Ecography Virtual Special Issue: The patterns and causes of elevational diversity gradients by Wiley-Blackwell

The patterns and causes of elevational diversity gradients Nathan J. Sanders and Carsten Rahbek

A major focus of research in spatial ecology over the past 25 years has been to understand why the number of species varies geographically. The most striking, and perhaps best documented, pattern in spatial ecology is the latitudinal gradient in species diversity in which the number of species, for most taxa, declines with increasing latitude. Understanding the underlying cause(s) of the latitudinal gradient has proven challenging, perhaps because there are really only two latitudinal gradients (in the northern and southern hemispheres), and because it is often difficult to perform experiments at latitudinal scales. Elevational gradients in species diversity are nearly as ubiquitous as latitudinal gradients, and they offer many characteristics that make them perhaps more suitable for uncovering the underlying cause(s) of spatial variation in diversity. First, there are many replicates of elevational diversity gradients – essentially each mountain or mountain range is a replicate, so it is possible to test for the generality of the underlying cause(s). Second, it is possible to carry out manipulative experiments along elevational gradients. Third, field data can be collected more readily along elevational gradients than along latitudinal gradients, simply because the

spatial extent of elevational gradients is small relative to latitudinal gradients. Finally, many of the potential underlying causes that covary along latitudinal gradients (history, climate, time since glaciation, area) do not covary along elevational gradients (Körner 2007). Given the benefits of elevational gradients relative to latitudinal gradients, it seems clear that they can be useful tools to understand the underlying cause(s) of diversity gradients. And, in fact, there is a growing appreciation of the utility of elevational gradients as tools to uncover the mechanisms that shape both patterns of biodiversity and the functioning of ecosystems (Fukami and Wardle 2005, Nogues-Bravo et al. 2008). Ecography has played a major role as an outlet for many studies of elevational gradient studies, and in fact such studies are one of the strengths of the journal. Since its inception, Ecography has published more than 25 papers that have explicitly focused on elevational diversity gradients. The papers highlighted in this Virtual Issue indicate that Ecography has been, and will continue to be, an important outlet for papers at the cutting edge of documenting and explaining elevational gradients in diversity. Here, our goal is to highlight some elevational diversity gradient papers published in Ecography (bold-face in reference list) that we feel have made long-lasting contributions to the study of spatial ecology. This Virtual Issue (http://tinyurl.com/cr2lkew) is about elevational diversity gradients, though we recognize that a number of key papers have been published in Ecography on topics ranging from montane diversity at regional or continental scales (Parra et al. 2004, Ricklefs et al. 2004, Ruggiero and Kitzberger 2004, Ruggiero and Hawkins 2008), population dynamics (Ramriez et al. 2006, Gimenez-Benavides et al. 2011), interactions among species (Fuentes et al. 1992, Mazia et al. 2004), adaptation (Berner et al. 2004), and climate change (Dollery et al. 2006).

The patterns Nearly 20 years ago, one of us (Rahbek 1995) asked whether the conventional wisdom about elevational diversity gradients – that they mirrored the latitudinal gradient and declined with elevation – was supported by the data. Early View (EV): 1-EV

Examining all of the literature (at the time, 97 papers) on elevational diversity gradients showed that the answer was, for the most part, ‘no’. Most studies, when sampling effort was corrected for, showed hump-shaped diversity gradients, with diversity peaking at mid-elevations. The quantitative review of published studies by Rahbek (1995) to document the generality (or lack thereof ) of the pattern was illuminating. The studies in Rahbek’s paper were from various mountain ranges, and on various taxa. One reason that different patterns of elevational diversity might occur in different systems may be that the scale and extent of the elevational gradients varied among studies (Rahbek 2005, Nogues-Bravo et al. 2008) or because different mountain ranges are embedded in different regional climatic areas with different evolutionary histories. This is an under-appreciated fact in comparative studies of elevational diversity gradients. Another approach to examine generality of elevational diversity gradients is to focus on several replicate elevational gradients within the same region, so that species occuring along the gradient might come from the same regional species pool and share similar evolutionary histories. This was the approach of Grytnes (2003), who sampled plant diversity along seven transects in northern Norway, Wang et al. (2009) who sampled tree and herb communities along six elevational gradients in northeast China, and of Sanders (2002) who compiled regional lists of the ants of Colorado, Nevada and Utah. In those studies, the patterns differed slightly among replicate samples, but the underlying causes were similar within each gradient. These results contrast with a study on non-volant mammals in several mountain ranges in Utah by Rowe (2009). In that study, the patterns of diversity with elevation were similar, but the underlying mechanisms differed among mountain ranges. But most of the elevational diversity gradient studies that have been published in Ecography have come from investigators who have compiled empirical data for a given taxon in a particular mountain range. These studies might differ in the extent and scale at which diversity is sampled, ranging from Herzog et al.’s (2005) data on bird diversity in 250 m elevational bands in the Andes to Grytnes’s (2003) 25 m2 plots in Norway. Regardless of the differences in sampling and extent among studies, most agree with the results from Rahbek’s (1995) review of the literature: in most instances, diversity peaks at mid-elevations, with a few notable exceptions (Brehm et al. 2003, Machac et al. 2011).

The underlying causes A number of factors have been implicated as underlying causes of elevational diversity gradients. Some of the most frequently tested are climate and productivity (Rahbek 1995, Odland and Birks 1999, Grytnes 2003, Fu et al. 2006, Rowe 2009, Wang et al. 2009), sourcesink dynamics (Kessler et al. 2011), area (Rahbek 1995, Sanders 2002, Jones et al. 2003, Bachman et al. 2004, Herzog et al. 2005, Romdal and Grytnes 2007), disturbance (Escobar et al. 2007, Bunn et al. 2011), geometric constraints (Sanders 2002, Bachman et al. 2004, Herzog 2-EV

et al. 2005, Fu et al. 2006, Rowe 2009) and evolutionary history (Machac et al. 2011). The diversity of results among studies, and even within studies, suggests that no single mechanism is responsible for all elevational diversity gradients. Future studies, many of which are likely to be published in Ecography (we hope), will move the field forward, perhaps by examining the interplay between contemporary and past climate (Hortal et al. 2011), integrating ecology and evolution (Graham et al. 2009, Machac et al. 2011), employing new tools (Levanoni et al. 2011) and demonstrating the effects of climatic change on current (Forister et al. 2010) and future patterns of biodiveristy (Colwell et al. 2008). Papers published in Ecography have been some of the first to test explicitly many of these mechanisms, and their generality. As the number of studies on elevational diversity gradients continues to grow (more than 300 as of 2011), Ecography will continue to play a role in shaping the field and helping to uncover the mechanisms which shape broadscale variation in species richness, especially along elevational gradients.

References Bachman, S. et al. 2004. Elevational gradients, area and tropical island diversity: an example from the palms of New Guinea. – Ecography 27: 299–310. Berner, D. et al. 2004. Grasshopper populations across 2000 m of altitude: is there life history adaptation? – Ecography 27: 733–740. Brehm, G. et al. 2003. Unique elevational diversity patterns of geometrid moths in an Andean montane rainforest. – Ecography 26: 456–466. Bunn, W. A. et al. 2011. Change within and among forest communities: the influence of historic disturbance, environmental gradients, and community attributes. – Ecography 33: 425–434. Colwell, R. K. et al. 2008. Global warming, elevational range shifts, and lowland biotic attrition in the wet tropics. – Science 322: 258–261. Dollery, R. et al. 2006. Impact of warming and timing of snow melt on soil microarthropod assemblages associated with Dryas-dominated plant communities on Svalbard. – Ecography 29: 111–119. Escobar, F. et al. 2007. From forest to pasture: an evaluation of the influence of environment and biogeography on the structure of beetle (Scarabaeinae) assemblages along three altitudinal gradients in the Neotropical region. – Ecography 30: 193–208. Forister, M. L. et al. 2010. Compounded effects of climate change and habitat alteration shift patterns of butterfly diversity. – Proc. Natl Acad. Sci. USA 107: 2088–2092. Fu, C. Z. et al. 2006. Elevational patterns of frog species richness and endemic richness in the Hengduan Mountains, China: geometric constraints, area and climate effects. – Ecography 29: 919–927. Fuentes, M. 1992. Latitudinal and elevational variation in fruiting phenology among western-European bird-dispersed plants. – Ecography 15: 177–183. Fukami, T. and Wardle, D. A. 2005. Long-term ecological dynamics: reciprocal insights from natural and anthropogenic gradients. – Proc. R. Soc. B 272: 2105–2115. Gimenez-Benavides, L. et al. 2011. Demographic processes of upward range contraction in a long-lived Mediterranean high mountain plant. – Ecography 34: 85–93.

Graham, C. C. et al. 2009. Phylogenetic structure in tropical hummingbird communities. – Proc. Natl Acad. Sci. USA l06: 19673–19678. Grytnes, J. A. 2003. Species–richness patterns of vascular plants along seven altitudinal transects in Norway. – Ecography 26: 291–300. Herzog, S. K. et al. 2005. The elevational gradient in Andean bird species richness at the local scale: a foothill peak and a high-elevation plateau. – Ecography 28: 209–222. Hortal, J. et al. 2011. Ice age climate, evolutionary constraints and diversity patterns of European dung beetles. – Ecol. Lett. 14: 741–748. Jones, J. I. et al. 2003. Area, altitude and aquatic plant diversity. – Ecography 26: 411–420. Kessler, M. et al. 2011. The impact of sterile populations on the perception of elevational richness patterns in ferns. – Ecography 34: 123–131. Körner, C. 2007. The use of ‘altitude’ in ecological research. – Trends Ecol. Evol. 22: 569–574. Levanoni, O. et al. 2011. Can we predict butterfly diversity along an elevation gradient from space? – Ecography 34: 372–383. Machac, A. et al. 2011. Elevational gradients in phylogenetic structure of ant communities reveal the interplay of biotic and abiotic constraints on diversity. – Ecography 34: 364–371. Mazia, C. N. et al. 2004. Interannual changes in folivory and bird insectivory along a natural productivity gradient in northern Patagonian forests. – Ecography 27: 29–40. Nogues-Bravo, D. et al. 2008. Scale effects and human impact on the elevational species richness gradients. – Nature 453: 216–219. Odland, A. and Birks, H. J. B. 1999. The altitudinal gradient of vascular plant richness in Aurland, western Norway. – Ecography 22: 548–566.

Parra, J. L. et al. 2004. Evaluating alternative data sets for ecological niche models of birds in the Andes. – Ecography 27: 350–360. Rahbek, C. 1995. The elevational gradient of species richness – a uniform pattern. – Ecography 18: 200–205. Rahbek, C. 2005. The role of spatial scale and the perception of large-scale species–richness patterns. – Ecol. Lett. 224– 239. Ramirez, J. M. et al. 2006. Altitude and woody cover control recruitment of Helleborus foetidus in a Mediterranean mountain area. – Ecography 29: 375–384. Ricklefs, R. E. et al. 2004. The region effect on mesoscale plant species richness between eastern Asia and eastern North America. – Ecography 27: 129–136. Romdal, T. S. and Grytnes, J. A. 2007. An indirect area effect on elevational species richness patterns. – Ecography 30: 440–448. Rowe, R. 2009. Environmental and geometric drivers of small mammal diversity along elevational gradients in Utah. – Ecography 32: 411–422. Ruggiero, A. and Hawkins, B. A. 2008. Why do mountains support so many species of birds? – Ecography 31: 306– 315. Ruggiero, A. and Kitzberger, T. 2004. Environmental correlates of mammal species richness in South America: effects of spatial structure, taxonomy and geographic range. – Ecography 27: 401–416. Sanders, N. J. 2002. Elevational gradients in ant species richness: area, geometry, and Rapoport’s rule. – Ecography 25: 25–32. Wang, X. P. et al. 2009. Relative importance of climate vs local factors in shaping the regional patterns of forest plant richness across northeast China. – Ecography 32: 133–142.

3-EV

ECOGRAPHY 27: 299 /310, 2004

Elevational gradients, area and tropical island diversity: an example from the palms of New Guinea Steven Bachman, William J. Baker, Neil Brummitt, John Dransfield and Justin Moat

Bachman, S., Baker, W. J., Brummitt, N., Dransfield, J. and Moat, J. 2004. Elevational gradients, area and tropical island diversity: an example from the palms of New Guinea. / Ecography 27: 299 /310. The factors causing spatial variation in species richness remain poorly known. In this study, factors affecting species richness of palms (Palmae/Arecaceae) were studied along the elevational gradient of New Guinea. Interpolated elevational ranges were calculated from a database of all known collections for 145 species in 32 genera. The amount of land area at different elevations greatly affects the species richness gradient. If assessed in equal-elevation bands species richness appears to decline monotonically, but when assessed in equal-area bands species richness shows a pronounced midelevation peak, due to the large proportion of lowlands in New Guinea. By randomising species ranges within the total elevational gradient for palms and accounting for area, we found the mid-elevation peak to be consistent with a middomain effect caused by the upper and lower limits to palm distribution. Our study illustrates the importance of accounting for area in macroecological studies of richness gradients and introduces a novel yet simple method for doing this through the use of equal-area bands. Together, the effect of area and the mid-domain effect explain the majority of variation in species richness of New Guinea palms. We support calls for the multivariate assessment of the mid-domain effect on an equal footing with other potential explanations of species richness. S. Bachman, Dept of Geography and Earth Sciences, Brunel Univ., Uxbridge Middlesex, U.K. UB8 3PH. / W. J. Baker (correspondence: w.baker@rbgkew.org.uk), N. Brummitt, J. Dransfield and J. Moat, The Herbarium, Royal Botanical Gardens, Kew, Richmond, Surrey, U.K. TW9 3AB.

Along with the well-known latitudinal gradients in species richness, patterns of biotic diversity along elevational gradients have been studied for centuries (Willdenow 1805, Darwin 1859, Wallace 1876, Whittaker 1960, Brown 1971, Rahbek 1995, 1997). Many of these studies have attempted to correlate observed patterns of diversity with various environmental gradients such as precipitation, temperature, humidity and productivity. It has been generally recognised that species richness declines with elevation. However, monotonic declines in richness are less typical than are unimodal peaks or patterns where species richness plateaus before decreasing (Rahbek 1995). Monotonic declines in species richness are thought by some authors to mirror a

decrease in productivity (Rahbek 1997 and references therein, Kaspari et al. 2000), although productivity gradients have also been used to explain mid-elevation peaks in species richness (Rosenzweig 1992, 1995, Rosenzweig and Abramsky 1993). However, to date, no universal relationship between productivity and richness has been elucidated. In contrast, the relationship between area and number of species is well known (Arrhenius 1921, Williams 1943, Rosenzweig 1995). Traditionally, species richness is thought to increase with increasing area following a power-law model (Williamson 1988; but see also Connor and McCoy 1979). In regions with diverse landscapes that range from sea level to high mountains, the land

Accepted 16 December 2003 Copyright # ECOGRAPHY 2004 ISSN 0906-7590 ECOGRAPHY 27:3 (2004)

299

area in different elevational bands varies greatly. Commonly, land area decreases as elevation increases such that lowlands account for the highest proportion of the total land area of a given region (MacArthur 1972); mountainous islands such as New Guinea display this feature particularly well. It seems strange then that, when examining species richness along elevational gradients, some authors have not considered the influence of area (e.g. Patterson et al. 1998, Ohlemu¨ller and Wilson 2000). Several studies show that when area has been controlled species richness peaks at mid-elevations (e.g. Lawton et al. 1987, Rahbek 1997), although unimodal peaks have been found without accounting for area (Koleff and Gaston 2001, Sanders 2002, Grytnes and Vetaas 2002). It has also been shown that area alone could explain a large proportion of the variance in the observed richness pattern (Rahbek 1997, Koleff and Gaston 2001, Sanders 2002). The mid-elevation peak is now recognised as a common pattern (Rahbek 1995, 1997, Heaney 2001, Kessler 2001, Nor 2001, Rickart 2001, Sa´nchez-Cordero 2001), but despite the recent resurgence of interest in studies of species richness along elevational gradients, a generally-accepted explanation for this mid-elevation peak has not yet been found. Recently, the potential influence of non-biological factors on species richness patterns has been highlighted (Colwell and Hurtt 1994, Colwell and Lees 2000). When species ranges are randomly placed within a geographically defined domain, a peak in richness inevitably arises towards the centre of this domain (Colwell and Lees 2000). In real terms, this domain may be an island or continent bounded by its shores, a mountain bounded by its summit and the lowlands at its base, or an ocean bounded by the surface and its floor. A gradient in elevation is a simple, one-dimensional domain. Within the domain, species with a range mid-point at midelevations will have a greater potential elevation range, or amplitude, as they can extend further both upwards and downwards than can those species whose range midpoint is found near the top or the bottom of the domain. Since species ranges are constrained geometrically in this way (they cannot extend either above the highest mountain top or below sea level), the number of overlapping elevational ranges will be greater at mid-elevations and, thus, so will species richness. This is due simply to species of moderate to large elevational range being more likely to cross the middle of the domain than to coincide at either boundary. The phenomenon, termed the ‘‘mid-domain effect’’ (MDE), is therefore dependent upon the underlying frequency distribution of relative species range-sizes. The mid-domain effect is attributed to these simple geometric constraints of the domain or, more accurately, to geometrically constrained stochastic processes affecting the underlying range-size frequency distribution (RSFD; Colwell et al. in press). 300

Despite growing evidence and support for the existence of the MDE (Colwell and Hurtt 1994, Lyons and Willig 1997, Pineda and Caswell 1998, Willig and Lyons 1998, Colwell and Lees 2000, Jetz and Rahbek 2001, Sanders 2002, Grytnes and Vetaas 2002), the concept has yet to gain widespread acceptance as a plausible factor contributing to richness patterns (Bokma and Mo¨nkko¨nen 2000, Koleff and Gaston 2001, Diniz-Filho et al. 2002). Several studies (Lees et al. 1999, Jetz and Rahbek 2001, Koleff and Gaston 2001, Sanders 2002) have investigated the combined influence of area and MDE on species richness patterns and in each case richness was explained well by the two combined factors. To date, however, few MDE studies have focused on vascular plants (but see Grytnes and Vetaas 2002), and in addition, except for a pioneering study of butterflies in Madagascar (Lees et al. 1999), no MDE studies have focused on large tropical islands such as New Guinea. New Guinea offers a perfect template for MDE studies being rich in endemic species, relative to total species richness, mountainous, and clearly defined in terms of a domain. Furthermore, the deepening biodiversity crisis should surely direct such research to those areas, such as tropical islands, where the greatest threats to biodiversity exist. This study aims to investigate the influence of available area and MDE on species richness in palms (family Arecaceae or Palmae) along the elevational gradient of New Guinea and surrounding islands, one of the most biodiverse areas in the Asia-Pacific region. We use two methods, one published (Sanders 2002) and one novel, to ask the following questions: 1) Does available area affect palm richness at different elevations in New Guinea? 2) Do geometric constraints affect palm richness along this elevational gradient? 3) Do genera of palms also show the same patterns as species do? The methods and data used in this study will be critically examined and may help to highlight important issues for other studies aiming to investigate similar themes.

Methods Study area and empirical data New Guinea is the largest tropical island in the world with a surface area of 808 510 km2 and an elevational range extending from sea level to 5030 m, the highest peak being Puncak Jaya (also known as Mt. Carstenz). The study area (Fig. 1) comprises the Indonesian province of Papua (formerly Irian Jaya) and the independent country of Papua New Guinea, and is bounded in the east by the Louisiade Archipelago and in the west by the islands Misool and Waigeo. In addition to mainland New Guinea, other major islands included in the study area are Yapen, Biak, Manus and the Bismarck Archipelago. The island of Bougainville (a part of Papua ECOGRAPHY 27:3 (2004)

Fig. 1. New Guinea and surrounding islands; the dark area represents the study region, the grey scale shading indicates elevational change. The island is divided into two political entities, the independent country Papua New Guinea to the east and the Indonesian province of Papua (formerly Irian Jaya) to the west. The islands to the north-east form the Bismarck Archipelago. Dots indicate the distribution of georeferenced collecting localities.

New Guinea) has been excluded because its biogeographic links are closer to the Solomon Islands than to New Guinea. The study area stretches over a relatively small latitudinal range (from 08 to 11840ƒS), thus minimising any possible effect of a latitudinal gradient. The primary dataset for this study was extracted from a database of /3000 herbarium specimen records from the specified area. The data were gathered from herbaria at six institutions (Royal Botanic Gardens, Kew; the Papua New Guinea Forest Research Inst., Lae; Herbarium Bogoriense, Bogor, Indonesia; the National Herbarium of the Netherlands [Leiden branch]; the Univ. of Aarhus, Denmark and the Queensland Herbarium, Brisbane, Australia). The database has been compiled as part of a current taxonomic research project on the palm flora of New Guinea (Baker 2002). Each record in the database relates to a single herbarium collection, including all known duplicates of any given collection. As well as taxonomic data, each record includes all locality and elevation data from the data labels of each specimen; collecting localities are shown in Fig. 1. The variation in the number of collections along the elevational gradient, with amount of land area in successive equal-elevation bands, is depicted in Fig. 2. Elevation ranges were calculated from the database records using two separate methods. The first method used empirical elevation readings recorded in the field by the collector; this will be referred to as the ‘‘field data method’’. This is regarded as a primary source of elevation data because it comes directly from the collector’s field measurements. These data will most likely have been gathered using a barometric altimeter, but may also have been estimated using topographic maps. In the second method, elevation ranges were derived from a digital elevation model (DEM) produced by the United States Geological Service (Anon. 2001); this will be referred to as the ‘‘DEM method’’. This is considered to be a secondary method because the elevation data are ECOGRAPHY 27:3 (2004)

indirectly derived from the locality data of the specimens through a DEM and are highly dependent on the quality of that locality data and the DEM. The DEM method was only applied to records that were geo-referenced (i.e. spatially referenced to a point on the surface of the earth), either with a GPS unit, from a gazetteer or with a reliable map estimate. Using only those species and genera for which both georeferences and field recorded elevation data were available (145 species, 32 genera) it was possible to compare the influences of the two data types on the analysis of species richness patterns. It should be noted that the two data types (field and DEM) are not wholly statistically independent. However, we include both since field and DEM data may differ slightly for the same record, and more importantly, in some cases, individual specimen records may have either georeferences or field-recorded elevation data, but not both. This comparison of field and DEM records may therefore expose potential biases in our analysis due to these differing data collection techniques.

Fig. 2. The relationship between available area and number of collections in successive equal-elevation bands. Lowlands make up by far the largest area in New Guinea; the amount of available area decreases dramatically with increasing elevation. The number of collections is not uniform over the elevational gradient but it remains proportional to the amount of available area.

301

When conducting a MDE analysis no taxon should have a range that extends beyond the domain being studied; in the narrowest sense, the species and genera should be endemic to the domain (Colwell and Lees 2000). In this study the lower limit of the domain is sea level and the upper limit is the maximum empirical elevation of palm records (Field data /2800 m, DEM / 2803 m). Of the 145 species available for analysis, 21 were not endemic to the study region. However, the elevational ranges of these non-endemic species did not exceed the limits of the domain after the examination of additional non-New Guinea records. Therefore, all 145 species were used in the analyses. Of the 32 palm genera that occur in New Guinea, only 2 are endemic to the region as a whole. The elevational ranges of all genera within New Guinea were therefore compared with their ranges outside of the region and as a result 8 genera were excluded from the analyses because their ranges outside of the study region extended beyond the domain limits within the study region. Thus, 24 genera were used in the analyses.

Gaston 2001, Colwell et al. in press). The purely theoretical Models 1 /3 (Colwell 2000) were also not used because of the implicit biological assumptions they make regarding RSFDâ&#x20AC;&#x2122;s (Colwell et al. in press). The model was iterated 100 times to produce the null distribution. The observed richness values and area values for each band were log-transformed to account for the relationship between species number and area. The MDE predictions were also log-transformed so that they could be compared with the log transformed observed richness values and area values. Using simple linear regression, the independent variables of area and MDE predictions were separately tested against our empirical dataset to examine the amount of variance in the observed richness that could be conditionally explained. The combined effects of both area and MDE predictions were also tested against observed richness. Tests of significance are complicated by the fact that richness figures for the elevational bands are spatially-autocorrelated and hence not statistically independent (Colwell pers. comm.). Consequently, no p-values are presented here.

Analytical methods

Method 2 Method 2 accounts for area by measuring richness along the elevational gradient in equal-area bands rather than equal-elevation bands. This method, despite its simplicity, appears not to have been used in species richness analyses so far. The elevational gradient was classified into equal-area bands using GIS software ArcView 3.2 (Anon. 1999). The original Digital Elevation Model (DEM) data (Anon. 2001) is given as integers (whole numbers), but because of the extremely large area of lowland in New Guinea this meant that the majority of pixels were classified as either 1 or 2 m elevation only. This did not allow sufficient equal-area bands to be produced, since it was impossible to split the largest band ( B/1 m elevation) any further. To overcome this problem the original elevation data was converted to decimal numbers and a random number between /0.5 and /0.5 added to the elevation of each pixel for all elevations (in ArcView Spatial Modeller the equation would be {[DEM] /0.5 /Grid.MakeRandom}). This produced a grid that could be easily classified into equal-area bands. Unfortunately, however, in ArcView it is not possible to classify a floating number grid into equal areas. To determine the band boundaries, therefore, the new DEM data was exported to a database program (only land was exported / sea was treated as null, and removed) and pixels sorted by ascending value of elevation. The band boundaries were calculated from the number of pixels divided by the number of bands desired. The value for each band boundary was then simply read off from the row number within the database, for each of 5, 10 and 15 equal-area bands.

Two methods were used to investigate the influence of area and MDE on taxon richness along the elevational gradient. In order to explore the basic relationship between elevation and richness the gradient was divided into 100 m equal-elevation bands and total richness was calculated as the sum of all species or genera occurring within each band. Method 1 then specifically tests the separate and combined influences of area and MDE on observed richness patterns within these equal-elevation bands, using simple linear regression. Method 2 examines the amount of variance in observed richness that can be explained by MDE after area has been accounted for, by dividing the gradient into equal-area bands, also using simple linear regression. Method 1 Method 1 follows the procedure used by Sanders (2002). Observed richness was calculated by summing the number of species or genera in each 100 m elevation band. It was assumed that each taxon occurred in all bands between its minimum and maximum elevational limits. The amount of area in each band was calculated using the DEM. The MDE null distribution was calculated using RangeModel software (Colwell 2000). We used Model 4 (Colwell 2000) which selects (with replacement) ranges from the empirical range-size frequency distribution (RSFD) and randomly places them in the domain (the elevation gradient). Model 5 (Colwell 2000), which uses empirical midpoints and random range-sizes, was not used because the null model is too closely constrained by the empirical data (Koleff and 302

ECOGRAPHY 27:3 (2004)

With area kept constant at each band, richness can be looked at anew. However, there is no set rule regarding the number of equal-area bands into which a gradient should be split. To explore the effect of the number of bands on richness estimates, we split the elevational gradient into 5, 10 and 15 equal-area bands. The elevational ranges for all these equal-area bands are given in Appendix 1. The equal-area richness patterns were then plotted against those predicted by the MDE null model under Model 4 of RangeModel (Colwell 2000). The model was iterated 100 times using the empirical RSFD. In using equal-area bands we have created bands with unequal elevation, since the amount of area in New Guinea declines with increasing elevation (see Fig. 2). In other words, the higher the equal-area band, the broader is its elevational range. Therefore multiple linear regression was used to compare the observed richness against expected richness derived from MDE null predictions with elevational range for each equal-area band as an additional explanatory variable. The procedure was carried out for species and genera using both types of elevation data (DEM and field data). Slope values for each relationship were also calculated, under the assumption that the closer the slope is to 1, the more closely the MDE null model predicts the number of observed taxa (see also Jetz and Rahbek 2001).

Results Basic relationships between richness and elevation Without accounting for area both species and genus richness of New Guinea palms decreases as elevation increases (Fig. 3A). Generic richness peaks in the lowest elevation band (0 /100 m) for both DEM and field data methods. Species richness also peaks in the lowest elevation band using the field data method, but peaks in the second band (201 /300 m) using DEM data. The DEM-derived richness estimates for both species and genera are consistently higher than the field data estimates along the gradient, with the exception of species for the first elevation band where the field data estimate is greater than the DEM estimate. However, the overall patterns along the gradient are broadly similar for both data types. Method 1 The relationships between New Guinea palm richness, area and MDE along the elevational gradient, presented as in Sanders (2002), are given in Fig. 4A /D. All r2 values are shown in Table 1. Area alone conditionally explained around half of the variance in observed richness pattern for both species and genera irrespective of data type used (min r2 /0.45 [genera, DEM]; max r2 /0.52 [species, Field and DEM]). MDE alone conECOGRAPHY 27:3 (2004)

Fig. 3. The relationship between richness and elevation within the study region. (A) species and generic richness within 100 m equal-elevation bands; (B) species and generic richness within equal-area bands. The thick, solid line represents species richness using field data; the dot-dash line species richness with elevations calculated from a DEM; the thin, solid line generic richness using field data and the dashed line generic richness with elevations calculated from a DEM.

ditionally explained very little of the variation in observed richness regardless of data type and taxonomic scale (min r2 /0.04 [species, Field]; max r2 /0.33 [species, DEM]). However, the combined effects of area and MDE conditionally explained a great deal of the variation (min r2 /0.77 [genera, Field]; max r2 /0.89 [species, DEM]) of the observed richness pattern, with the highest r2 value for species DEM data (Table 1). In general, the combined effects of area and MDE gave higher r2 values for DEM data (species, 0.89; genera, 0.84) than for field data (species, 0.84; genera, 0.77).

Method 2 The removal of the effect of area on richness through the use of equal-area bands yielded a mid-elevation peak in both species and genus richness patterns (Fig. 3B). This contrasts markedly with the basic relationship between richness and elevation, which is roughly monotonic (Fig. 3A). The relationships between richness, area and MDE are given in Fig. 5A /D for 15 equal-area bands. All r2 values are shown in Table 2. Although the MDE null predictions consistently underestimate observed richness, the variation in observed patterns at both taxo303

Fig. 4. Comparison of species and generic richness with amount of area and MDE null-model predictions along the elevational gradient. (A) species richness with elevations from field data; (B) generic richness with elevations from field data; (C) species richness with elevations from DEM data; (D) generic richness with elevations from DEM data. For each graph the solid line represents species or generic richness, the solid line with hollow circles shows area and the solid line with solid circles shows MDE null-model predictions.

nomic levels and with both data types are generally comparable with the null model predictions. Linear regression shows that the variance in species richness with field data in most cases is conditionally explained remarkably well by the MDE null predictions irrespective of the number of equal-area bands (min r2 / 0.90 [5 and 15 bands]; max r2 /0.98 [10 bands]; see Table 2). However, for species richness from DEM data and generic richness with both field data and DEM data the number of equal-area bands does seem to have an influence; r2 values decrease in unison with number of

equal-area bands, and markedly so for generic richness with field data (r2 /0.70, 15 bands; 0.66, 10 bands; 0.39, 5 bands). In general the variance in richness is best explained when 15 equal-area bands are used (species richness from DEM data, r2 /0.98; generic richness from DEM data, r2 /0.89; generic richness from field data, r2 /0.70), although for species richness with field data the highest r2 value is with 10 equal-area bands. As with Method 1, DEM data consistently returns higher r2 values (species, 0.93 /0.98; genera, 0.82 /0.89) than does field data (species, 0.90 /0.98; genera, 0.39 /0.70). By

Table 1. Results from the simple linear regression analysis using Method 1. The separate and combined effects of area and MDE are shown. Note that in all cases the combined effects of area and MDE explain more of the variance in observed richness than either does alone. Data type

Elevation range (m)

Taxonomic level

Parameter

Field Field Field Field Field Field DEM DEM DEM DEM DEM DEM

0 /2800 0 /2800 0 /2800 0 /2438 0 /2438 0 /2438 0 /2803 0 /2803 0 /2803 0 /2796 0 /2796 0 /2796

Species Species Species Genera Genera Genera Species Species Species Genera Genera Genera

observed observed observed observed observed observed observed observed observed observed observed observed

304

vs vs vs vs vs vs vs vs vs vs vs vs

r2 area MDE MDE /area area MDE MDE /area area MDE MDE /area area MDE MDE /area

0.52 0.04 0.84 0.50 0.07 0.77 0.52 0.33 0.89 0.45 0.10 0.84

ECOGRAPHY 27:3 (2004)

Fig. 5. Comparison of observed richness patterns and MDE null-model predictions along the elevational gradient, split into 15 equal-area bands. (A) species richness with elevations from field data; (B) generic richness with elevations from field data; (C) species richness with elevations from DEM data; (D) generic richness with elevations from DEM data. For each graph the hollow circles show observed richness and solid circles show MDE null predictions.

adding elevational-width of equal area-bands as an additional explanatory variable in a multiple linear regression high r2 values increased marginally whereas low r2 values increased appreciably (see Table 2). Slope values for the relationship between generic richness and MDE null predictions for field and DEM data for 15 equal-area bands, and for field data for 10 equal-area bands were 5/1.09/0.1 (see Table 2). For all types of data, MDE predictions generally underestimated the observed numbers of taxa (see Fig. 5). Slope values for species-level data were all considerably further from 1 than for genus-level data (see Table 2), revealing a greater under-estimation of actual species richness

patterns by MDE null model predictions (see also Fig. 5).

Discussion Data quality With the field and DEM methods we have used two different techniques for obtaining elevation range data from specimen records. As one would expect, both methods have produced broadly similar patterns, although there are slight discrepancies with the max-

Table 2. Results from multiple linear regression analysis of richness against mid-domain effect predictions and equal-area elevational band-width using Method 2. MDE null predictions are correlated with observed richness patterns once area has already been accounted for. Data type

Elevation range (m)

Taxonomic level

No. equal area bands

Slope

r2 (MDE alone)

Multiple r2

Field Field Field Field Field Field DEM DEM DEM DEM DEM DEM

0 /2800 0 /2800 0 /2800 0 /2438 0 /2438 0 /2438 0 /2803 0 /2803 0 /2803 0 /2797 0 /2797 0 /2797

Species Species Species Genera Genera Genera Species Species Species Genera Genera Genera

5 10 15 5 10 15 5 10 15 5 10 15

1.39 1.61 1.63 0.48 0.92 0.90 2.04 1.97 1.99 0.38 0.88 1.06

0.90 0.98 0.90 0.39 0.66 0.70 0.93 0.97 0.97 0.82 0.84 0.86

0.92 0.98 0.95 0.99 0.93 0.91 0.94 0.98 0.97 0.95 0.85 0.87

ECOGRAPHY 27:3 (2004)

305

imum elevations. Assessing the accuracy of either method is problematic. The field data method may provide spurious results because pre-twentieth century herbarium records are less likely to have accurate elevation data due to imprecision of maps and inadequate measuring equipment; even modern altimeters can give misleading measurements. The DEM method may be inaccurate due to imprecise georeferencing or the inexactness of the DEM itself. A strength of this study is that it builds directly on current taxonomic expertise. Data for this analysis are derived directly from individual records for each species (i.e. herbarium specimens), which are primary observations. The database includes the vast majority of available specimen records for New Guinea palms and expert taxonomists have verified the identity of each record. Thus, we feel our data are impeccably sourced and any problem detected during the analysis can be readily investigated. Most other studies have had to rely on distribution data contained in published monographs and field guides, in the form of line maps or dot maps, or written descriptions of distributions. We have therefore avoided some of the additional assumptions associated with the use of such secondary data sources, such as poor coverage of specimens inadequately reflecting true geographic ranges, the incorrect identification of individual records, or outdated taxonomic concepts. As practising taxonomists, we encourage others to use a more explicit, specimen-based approach. Sampling procedure can have a significant influence on richness estimates (Wolda 1987, Rahbek 1995). When investigating Rapoportâ&#x20AC;&#x2122;s rule, for example, Colwell and Hurtt found that simply following a standard procedure, such as sampling with equal effort at points along a gradient, can produce a spurious Rapoport effect (Colwell and Hurtt 1994). The relationship between area and sampling effort in each elevational band in this study is illustrated in Fig. 2. Although the number of collections is not equal along the elevational gradient, the graph shows that collection intensity nevertheless varies in proportion with land area in each elevation band, as one would expect.

Interpolation It is an assumption inherent in this study that species occur in all elevational bands between the minimum and maximum observed values. Such interpolation is typical of analyses involving species richness estimates (Whittaker et al. 2001, Grytnes and Vetaas 2002). It has been suggested that species richness estimates based on interpolation may overestimate richness towards the centre of the gradient because species are only strictly observed in bands at the extreme ends of each species range (Grytnes and Vetaas 2002). In addition to being 306

overestimated in the centre of the gradient, richness may also be underestimated at the periphery since it cannot be interpolated beyond the range limits (Grytnes and Vetaas 2002). However, interpolating species ranges is a pragmatic solution to an intractable analytical problem, as we have no evidence that species are not found where the ranges have been interpolated, which may not noticeably alter the underlying trends in richness (Lees et al. 1999). Furthermore, after plotting the residuals from the linear regression of log area and log species/ genus richness against elevation (Rahbek 1995) the midelevation peak was still evident with all four analyses (graphs not reproduced here).

The influence of area The analysis of richness patterns of palm species and genera along the elevational gradient in New Guinea yields further evidence for the significant influence of area on these patterns. The species-area relationship has been universally acknowledged, although the exact structure of the relationship is still under discussion (Connor and McCoy 1979, Plotkin et al. 2000, Crawley and Harral 2001). Method 1 shows that area alone explains a good deal of the variation in observed richness (Table 1). Method 2 shows that by factoring out area the richness pattern changes from roughly monotonic to unimodal, irrespective of taxonomic scale or number of equal-area bands. This agrees with previous studies that have found a hump-shaped species richness pattern after accounting for area (Lawton et al. 1987, Rahbek 1997). However, if richness along the elevational gradient were entirely dependent on the effect of area, the richness pattern would vary little across the gradient when examined using equal-area bands. As can be seen in Fig. 3B this is not the case and richness peaks at mid-elevations when the effect of area is removed. The influence of area is completely removed using the equalarea band method and allows the area-controlled richness pattern to be directly compared with MDE null predictions. We suggest that all subsequent MDE analyses need to take into account the influence of area, and this can easily be achieved using the equal-area band methodology presented here. Using equal-area bands in regions where amount of available area declines steadily with elevation, such as New Guinea (see Fig. 2), means that equal-area bands increase in elevational breadth as elevation increases. This might mean that species richness for equal-area bands will increase with elevation simply because as the elevational breadth of each band increases so each band will include a greater number of species, assuming the RSFD remains roughly the same with elevation, and also because beta diversity (species turnover) tends to increase with elevation (Colwell pers. comm., see also ECOGRAPHY 27:3 (2004)

Rahbek 1997). In New Guinea palms, elevational ranges increase with elevation for both species and genera (Spearman’s rank correlation between mean elevational range and elevational breadth for equal-area bands: coefficient rs /0.72 and 0.79 for species field and DEM data, respectively, n /145; 0.85 and 0.98 for generic field and DEM data, respectively, n /24). This increase in mean elevational range accompanies a reduction in the number of taxa at higher elevations, so countering the possible artefacts from increases in the elevational breadth of equal-area bands. Furthermore, the addition of elevational width of equal-area bands to the regression model in most cases accounted for only a marginal increase in the amount of explained variation in the observed richness pattern (see Table 2).

The mid-domain effect in New Guinea palms Despite growing evidence, there is still great scepticism surrounding the importance of the mid-domain effect. So far only a handful of studies have considered the influence of area and MDE on richness (Rahbek 1997, Lees et al. 1999, Jetz and Rahbek 2001, Koleff and Gaston 2001, Sanders 2002). In these studies, the combined effects of area and MDE were found to explain a large proportion of the variance in observed richness. Our study also supports these findings. MDE predictions conditionally explained up to 98% of the variance in observed richness in this study after the effect of area had been removed (species field data, 10 equalarea bands, r2 /0.98). However, as the authors were themselves careful to point out, geometric constraints are an additional influence on patterns of species richness (hence the name mid-domain ‘‘effect’’; Colwell and Lees 2000); it is not claimed to be the sole explanation. Vascular plants have rarely been considered in MDE analyses until now (but see Grytnes and Vetaas 2002), but our results and others point to the general influence of MDE across a wide range of taxa such as birds, mammals, insects, and vascular plants. This study furthermore shows that MDE is also important at different taxonomic scales. However, despite the high r2 values (Table 2), MDE null predictions consistently underestimate observed richness patterns (Fig. 5A /D). This may be a result of the way in which the RangeModel software (Colwell 2000) calculates richness along the domain (see Fig. 1 in Colwell and Lees 2000). Setting the bin range effectively determines the number of times the domain is intersected. For instance, if the domain limits are 0 and 1, and the number of bins is 5, the domain would be intersected at 0.1, 0.3, 0.5, 0.7 and 0.9. Every time a range crosses one of these points, the richness total for that bin increases by one. If a species range extended from 0.4 to 0.8, it would intersect the bins at 0.5 and 0.7, thus ECOGRAPHY 27:3 (2004)

adding to their respective totals. Richness for each bin is calculated as the total number of ranges intersected at that point. Using this intersect method it is possible that very small ranges might lie between these points which will not be picked up in the richness totals for that bin. This is more likely to occur for RSFD’s that contain a high proportion of small ranges. Taxa are also more likely to be missed when the number of bins is small, due to the increased size of the gap between the intersection points. This may explain why observed richness is explained least well (e.g. genus field data, 5 equal-area bands, r2 /0.39) by the MDE when the gradient is split into fewer bins/equal area bands. Inspecting the slope values for all relationships, only those for generic field data for 10 and 15 equal-area bands and generic DEM data for 15 equal-area bands were within 0.1 of an ideal slope of 1.0. The slope when using only 5 equal-area bands for genera for both field data and DEM data was considerably smaller than for both 10 and 15 bands with the same data type, again suggesting that 5 equal-area bands is too few to accurately represent richness patterns over an elevational gradient of this size (see Table 2). Furthermore, the slope values are furthest from 1 for species data (especially species DEM data), showing that MDE predictions under-estimated observed numbers of taxa more when ranges were smaller, and lending weight to the supposition that small species ranges are being missed between the intersections of the domain. Analyses that have considered the combined effects of area and MDE (in one dimension), including this study, have shown that these factors can explain observed richness patterns well (Jetz and Rahbek 2001, Koleff and Gaston 2001, Sanders 2002). MDE null models have attracted criticism for only considering richness patterns across one dimension when, of course, species have ranges extending in two dimensions (Bokma and Mo¨nkko¨nen 2000, Koleff and Gaston 2001). A two-dimensional approach has also been developed (Jetz and Rahbek 2001) which indicates that MDE null predictions across two dimensions can still explain a significant amount of the variance in observed richness patterns. It must be noted, however, that these two-dimensional predictions did not explain as much as the one-dimensional models. In order to fully understand the importance of the MDE it needs to be considered alongside various other factors that may contribute to the pattern of species richness. Unfortunately, data were not available in this analysis for the investigation of other variables that may have contributed to the pattern (e.g. productivity, temperature). Two multivariate analyses have shown that the MDE can still be an important explanatory variable in studies of richness gradients (Lees et al. 1999, Jetz and Rahbek 2002); we support calls for the influence 307

of MDE to be further assessed on an equal footing with other determinants of species richness patterns.

The mid-domain effect and Rapoport’s elevational rule Rapoport’s rule states that there is a simple positive correlation between species latitudinal range-size and latitude (Stevens 1989), following an original observation by Rapoport (1982). This idea has also been extended by Stevens (1992) to elevational gradients, where species elevational ranges would increase with elevation, and so decrease in extent towards the lowlands. Rapoport’s elevational rule runs counter to the expectation of the mid-domain effect, which predicts greatest species richness at middle elevations due to the geometric constraints exerted by the upper and lower boundaries of the elevational gradient. For New Guinea palms, elevational range does indeed increase with midpoint of elevation, and, as Fig. 3A shows, species richness does appear to be greater at low elevations without accounting for the differing amount of available area at different elevations. However, it should be evident from Fig. 3B that, after accounting for area, the results presented here provide support for the expectations of the mid-domain effect model, and against those of Rapoport’s elevational rule (see also Rahbek 1997). This suggests that it is simply the greater available area of lowland regions which gives rise to their apparent diversity, and that the Rapoport rescue effect, which remains largely untested in tropical lowland environments, may not be the cause of such species richness gradients.

Taxonomic scale This study is one of the first yet completed to assess the mid-domain effect at both species and genus level. Although the use of indicator taxa remains contentious for biodiversity studies, genera are now well established as reliable higher-taxonomic surrogates for species-level diversity patterns (Williams and Gaston 1994, La Ferla et al. 2002). It is clear from Fig. 5 that patterns in taxon richness across the elevational gradient, including the results of mid-domain effect null model simulations, are similar and highly correlated between both species and genera, further demonstrating the generality of the middomain effect.

islands. Given the urgent need to understand the factors dictating the distribution of tropical diversity, it is vital that new ecological models such as the mid-domain effect are explored over the broadest taxonomic spectrum and across a wide range of geographical scales and locations. Islands such as New Guinea present significant challenges to our understanding of biodiversity being extremely rich biologically and yet poorly explored. Detailed knowledge of tropical plant diversity is still limited by a shortage of good collections, especially for plant families such as palms, which are difficult to collect and often under-represented in herbaria. Although collection densities of palms in New Guinea remain low, a one-dimensional approach yields a wellsampled gradient that can be analysed effectively using null models. It seems likely that many similar datasets exist and we encourage others to undertake analyses such as ours. This study adds significantly to the growing body of evidence supporting the influence of the mid-domain effect. This effect is undoubtedly an important consideration for any analysis of richness gradients. Using a simple yet novel method, we have demonstrated that it is essential to account for the influence of area on species richness patterns when exploring concepts such as the mid-domain effect. In our study, the observed richness patterns can be explained to a strikingly large extent by the combined role of the mid-domain effect and area. In future, these factors need to be considered in a multivariate context to understand their ultimate importance relative to additional environmental variables. While findings such as ours may continue to be controversial in the short term, we believe that, in time, the role of the mid-domain effect will become increasingly accepted as one of several factors determining species richness patterns. Acknowledgements / Many people contributed to the New Guinea palm database, in particular Roy Banka, Anders Barfod, Kate Davis, Anders Kjaer, Meesha Patel and Helen Sanderson. We thank the staff of the herbaria at Aarhus, Brisbane, Bogor, Kew, Lae and Leiden for kindly providing access to collections and data. We thank Robert Colwell, David Lees and John-Arvid Grytnes for valuable discussions and numerous comments on earlier versions of the manuscript. Carsten Rahbek made many suggestions that improved the paper. This work was supported by a student internship from the Royal Botanic Gardens, Kew to SPB, a studentship from the Bernard Sunley Charitable Trust to NAB and by funding from the BAT Biodiversity Partnership to the Palms of New Guinea project.

References Conclusions Few studies of this kind have focused on vascular plants and even fewer have been based on taxa from tropical 308

Anon. 1999. ArcView GIS. / ESRI, Redlands, California. Anon. 2001. GTOPO30 / Global topographic data. / USGS, B/http://edcdaac.usgs.gov/gtopo30/gtopo30.html /, last accessed: 30 Sept 2002. Arrhenius, O. 1921. Species and area. / J. Ecol. 9: 95 /99. ECOGRAPHY 27:3 (2004)

Baker, W. J. 2002. The palms of New Guinea project. / Fl. Males. Bull. 13: 35 /37. Bokma, F. and Mo¨nkko¨nen, M. 2000. The mid-domain effect and the longitudinal dimension of continents. / Trends Ecol. Evol. 15: 288 /289. Brown, J. H. 1971. Mammals on mountaintops: nonequilibrium insular biogeography. / Am. Nat. 105: 467 /478. Colwell, R. K. 2000. RangeModel: a Monte Carlo simulation tool for assessing geometric constraints on species richness. Ver. 3. / User’s guide and application, B/http://viceroy.eeb. uconn.edu/asn /. Colwell, R. K. and Hurtt, G. C. 1994. Nonbiological gradients in species richness and a spurious Rapoport effect. / Am. Nat. 144: 570 /595. Colwell, R. K. and Lees, D. C. 2000. The mid-domain effect: geometric constraints on the geography of species richness. / Trends Ecol. Evol. 15: 70 /76. Colwell, R. K., Rahbek, C. and Gotelli, N. J. in press. The middomain effect and species richness patterns: what have we learned so far. / Am. Nat. Connor, E. F. and McCoy, E. D. 1979. The statistics and biology of the species-area relationship. / Am. Nat. 113: 791 /833. Crawley, M. J. and Harral, J. E. 2001. Scale dependence in plant biodiversity. / Science 291: 864 /868. Darwin, C. 1859. On the origin of species by means of natural selection or the preservation of favoured races in the struggle for life. / John Murray, London. Diniz-Filho, J. A. F. et al. 2002. Null models and spatial patterns of species richness in South American birds of prey. / Ecol. Lett. 5: 47 /55. Grytnes, J. A. and Vetaas, O. R. 2002. Species richness and altitude: a comparison between null models and interpolated plant species richness along the Himalayan altitudinal gradient, Nepal. / Am. Nat. 519: 294 /304. Heaney, L. R. 2001. Small mammal diversity along elevational gradients in the Philippines: an assessment of patterns and hypotheses. / Glob. Ecol. Biogeogr. 10: 15 /39. Jetz, W. and Rahbek, C. 2001. Geometric constraints explain much of the species richness pattern in African birds. / Proc. Natl. Acad. Sci. USA 98: 5661 /5666. Jetz, W. and Rahbek, C. 2002. Geographic range size and determinants of avian species richness. / Science 297: 1548 / 1551. Kaspari, M., O’Donnell, S. and Kercher, J. R. 2000. Energy, density, and constraints to species richness: studies of ant assemblages along a productivity gradient. / Am. Nat. 155: 280 /293. Kessler, M. 2001. Pteridophyte species richness in Andean forests in Bolivia. / Biodiv. Conserv. 10: 1473 /1495. Koleff, P. and Gaston, K. J. 2001. Latitudinal gradients in diversity: real patterns and random models. / Ecography 24: 341 /351. La Ferla, B. et al. 2002. Continental scale patterns of biodiversity: can higher taxa accurately predict African plant distributions? / Bot. J. Linn. Soc. 138: 225 /235. Lawton, J. H., MacGarvin, M. and Heads, P. A. 1987. Effects of altitude on the abundance and species richness of insect herbivores on bracken. / J. Anim. Ecol. 56: 147 /160. Lees, D. C., Kremen, C. and Andriamampianina, L. 1999. A null model for species richness gradients: bounded range overlap of butterflies and other rainforest endemics in Madagascar. / Biol. J. Linn. Soc. 67: 529 /584. Lyons, S. K. and Willig, M. R. 1997. Latitudinal patterns of range size: methodological concerns and empirical evaluations for New World bats and marsupials. / Oikos 79: 568 / 580. MacArthur, R. H. 1972. Geographical ecology: patterns in the distribution of species. / Harper and Rowe. Nor, S. M. 2001. Elevational diversity patterns of small mammals on Mt. Kinabalu, Sabah, Malaysia. / Glob. Ecol. Biogeogr. 10: 41 /62.

ECOGRAPHY 27:3 (2004)

Ohlemu¨ller, R. and Wilson, J. B. 2000. Vascular plant species richness along latitudinal and altitudinal gradients: a contribution from New Zealand temperate rainforests. / Ecol. Lett. 3: 262 /266. Patterson, B. D. et al. 1998. Contrasting patterns of elevational zonation for birds and mammals in the Andes of southeastern Peru. / J. Biogeogr. 25: 593 /607. Plotkin, J. B. et al. 2000. Predicting species diversity in tropical forests. / Proc. Natl. Acad. Sci. USA 97: 10850 / 10854. Pineda, J. and Caswell, H. 1998. Bathymetric species-diversity patterns and boundary constraints on vertical range distributions. / Stud. Oceanogr. 45: 83 /101. Rahbek, C. 1995. The elevational gradient of species richness: a uniform pattern? / Ecography 18: 200 /205. Rahbek, C. 1997. The relationship among area, elevation, and regional species richness in Neotropical birds. / Am. Nat. 149: 875 /902. Rapoport, E. H. 1982. Areography: geographical strategies of species. The Fundacio´n Bariloche Series Vol. 1. / Pergamon Press. Rickart, E. A. 2001. Elevational diversity gradients, biogeography and the structure of montane mammal communities in the intermountain region of North America. / Glob. Ecol. Biogeogr. 10: 77 /100. Rosenzweig, M. L. 1992. Species diversity gradients: we know more and less than we thought. / J. Mammal. 73: 715 /730. Rosenzweig, M. L. 1995. Species diversity in space and time. / Cambridge Univ. Press. Rosenzweig, M. L. and Abramsky, Z. 1993. How are diversity and productivity related? / In: Ricklefs, R. E. and Schluter, D. (eds), Species diversity in ecological communities: historical and geographical perspectives. Univ. of Chicago Press, pp. 52 /65. Sa´nchez-Cordero, V. 2001. Small mammal diversity along elevational gradients in Oaxaca, Mexico. / Glob. Ecol. Biogeogr. 10: 63 /76. Sanders, N. 2002. Elevational gradients in ant species richness: area, geometry, and Rapoport’s rule. / Ecography 25: 25 / 32. Stevens, G. C. 1989. The latitudinal gradient in geographic range: how so many species coexist in the tropics. / Am. Nat. 133: 240 /256. Stevens, G. C. 1992. The elevational gradient in altitudinal range: an extension of Rapoport’s latitudinal rule to altitude. / Am. Nat. 140: 893 /911. Wallace, A. R. 1876. The geographical distribution of animals. / Macmillan. Whittaker, R. H. 1960. Vegetation of the Siskiyou Mountains, Oregon and California. / Ecol. Monogr. 30: 279 /338. Whittaker, R. J., Willis, K. J. and Field, R. 2001. Scale and species richness: toward a general hierarchical theory of species diversity. / J. Biogeogr. 28: 453 /470. Willdenow, K. L. 1805. The principles of botany, and vegetable physiology. / Blackwood, Cadell and Davies, London. Williams, C. B. 1943. Area and number of species. / Nature 152: 264 /267. Williams, P. H. and Gaston, K. J. 1994. Measuring more of biodiversity: can higher-taxon richness predict wholesale species richness. / Biol. Conserv. 67: 211 /217. Williamson, M. 1988. Relationship of species number to area, distance and other variables. / In: Myers, A. A. and Giller, P. S. (eds), Analytical biogeography. Chapman and Hall, pp. 94 /115. Willig, M. R. and Lyons, S. K. 1998. An analytical model of latitudinal gradients of species richness with an empirical test for marsupials and bats in the New World. / Oikos 81: 93 /98. Wolda, H. 1987. Altitude, habitat and tropical insect diversity. / Biol. J. Linn. Soc. 30: 313 /323.

309

Appendix 1. Ranges of equal-area bands for each of 5, 10 and 15 bands used. Minimum and maximum elevations in metres for each equal-area band along the elevation gradient are given (as used in Method 2). Ranges are shown separately for each data type (Field and DEM). Band

5 Bands 1 2 3 4 5

Field data, Species

Field data, Genera

DEM data, Species

DEM data, Genera

Min (m)

Max (m)

0.5 14.5 122.8 304.6 923.2

14.5 122.8 304.6 923.2 2800.0

0.5 12.7 113.3 296.6 855.4

12.7 113.3 296.6 855.4 2438.0

0.5 14.5 122.8 304.6 923.7

14.5 122.8 304.6 923.7 2803.0

0.5 14.5 122.7 304.6 922.6

14.5 122.7 304.6 922.6 2797.0

10 Bands 1 0.5 2 1.2 3 14.5 4 56.7 5 122.8 6 219.8 7 304.6 8 550.3 9 923.2 10 1631.8

1.2 14.5 56.7 122.8 219.8 304.6 550.3 923.2 1631.8 2800.0

0.5 1.2 12.7 52.7 113.3 206.2 296.6 487.0 855.4 1510.8

1.2 12.7 52.7 113.3 206.2 296.6 487.0 855.4 1510.8 2438.0

0.5 1.2 14.5 56.7 122.8 219.9 304.6 550.8 923.7 1633.5

1.2 14.5 56.7 122.8 219.9 304.6 550.8 923.7 1633.5 2803.0

0.5 1.2 14.5 56.6 122.7 219.7 304.6 550.0 922.6 1629.9

1.2 14.5 56.6 122.7 219.7 304.6 550.0 922.6 1629.9 2797.0

15 Bands 1 0.5 2 1.0 3 1.5 4 14.5 5 40.8 6 74.1 7 122.8 8 185.4 9 252.7 10 304.6 11 435.7 12 630.8 13 923.2 14 1467.3 15 2017.2

1.0 1.5 14.5 40.8 74.1 122.8 185.4 252.7 304.6 435.7 630.8 923.2 1467.3 2017.2 2800.0

0.5 1.0 1.5 12.7 37.4 69.4 113.3 172.4 238.5 296.6 376.7 587.8 855.4 1262.5 1722.5

1.0 1.5 12.7 37.4 69.4 113.3 172.4 238.5 296.6 376.7 587.8 855.4 1262.4 1722.5 2438.0

0.5 1.0 1.5 14.5 40.8 74.1 122.8 185.5 252.9 304.6 436.1 631.3 923.7 1468.7 2019.7

1.0 1.5 14.5 40.8 74.1 122.8 185.5 252.9 304.6 436.1 631.3 923.7 1468.7 2019.7 2803.0

0.5 1.0 1.5 14.5 40.8 74.0 122.7 185.3 252.6 304.6 435.4 630.3 922.6 1465.9 2015.0

1.0 1.5 14.5 40.8 74.0 122.7 185.3 252.6 304.6 435.4 630.3 922.6 1465.9 2015.0 2797.0

310

ECOGRAPHY 27:3 (2004)

ECOGRAPHY 27: 733 /740, 2004

Grasshopper populations across 2000 m of altitude: is there life history adaptation? Daniel Berner, Christian Ko¨rner and Wolf U. Blanckenhorn

Berner, D., Ko¨rner, Ch. and Blanckenhorn, W. U. 2004. Grasshopper populations across 2000 m of altitude: is there life history adaptation? / Ecography 27: 733 /740. Life history differentiation along climatic gradients may have allowed a species to extend its geographic range. To explore this hypothesis, we compared eleven Omocestus viridulus (Orthoptera: Acrididae) populations along an altitudinal gradient from 410 to 2440 m in Switzerland, both in the field and laboratory. In situ temperature records indicated a striking decline in available heat sums along the gradient, and field populations at high altitudes reached egg hatching and adulthood much later in the year than at low elevation. The reproductive period at high altitude is thus severely limited by season length, especially during a cool year. However, controlled environment experiments revealed that intrinsic rates of embryonic and juvenile development increased with the populations’ altitude of origin. This countergradient variation is largely genetic and conforms to predictions of life history theory. No corresponding differentiation in the overwintering egg stage, a pivotal determinant of phenology, was found. This trait seems conserved within the gomphocerine grasshopper subfamily. Although we found evidence for altitudinal adaptation in development, the potential of O. viridulus to adapt to cool alpine climates appears restricted by a phylogenetic constraint. D. Berner (daniel.berner@fal.admin.ch), Agroscope FAL Reckenholz, Swiss Federal Research Station for Agroecology and Agriculture, Reckenholzstr. 191, CH-8046 Zu¨rich, Switzerland. / Ch. Ko¨rner, Inst. of Botany, Univ. of Basel, Scho˝nbeinstrasse 6, CH-4056 Basel, Switzerland. / W. U. Blanckenhorn, Zoological Museum, Univ. of Zu¨rich, Winterthurerstrasse 190, CH-8057 Zu¨rich, Switzerland.

The seasonal recurrence of adverse climatic conditions is a principal force shaping ectotherm life cycles in temperate regions. Growth, development, reproduction and dormancy need to be coordinated and timed in relation to the available growing season (Taylor and Karban 1986, Danks 1994). The set of adaptations, which synchronizes the life cycle with the growing season, is reflected in the organisms’ phenology (Tauber et al. 1986). The length of the growing season generally declines with increasing latitude and altitude. Thus, geographically widespread species have to cope with a variety of climatic conditions, which can basically be achieved in two / not mutually exclusive / ways. Firstly, a generalist genotype may display plastic responses in

relation to environmental conditions (Gotthard and Nylin 1995, Schlichting and Pigliucci 1998). Phenotypic plasticity in traits relevant to seasonal timing has been documented and interpreted in adaptive terms in several insect species (Tanaka and Brookes 1983, Nylin 1994, Blanckenhorn 1997, Kingsolver and Huey 1998). Secondly, spatial variation in selection pressures may give rise to genetic differentiation between populations due to natural selection. Prerequisites are heritable genetic variation and restricted gene flow between local populations (Slatkin 1987). Both responses, local adaptation and phenotypic plasticity, may allow a species to extend its distribution across a range of altitudes and latitudes. Several studies of ectotherms report genetic life history

Accepted 30 July 2004 Copyright # ECOGRAPHY 2004 ISSN 0906-7590 ECOGRAPHY 27:6 (2004)

733

differentiation in relation to systematic geographic variation in climate (Masaki 1967, Berven and Gill 1983, Dingle et al. 1990, Ayres and Scriber 1994, Blanckenhorn and Fairbairn 1995, Telfer and Hassall 1999, Merila¨ et al. 2000; but see Lamb et al. 1987). However, almost all studies focus on latitude, whereas evidence for altitudinal life history adaptation in animals is exceedingly scarce. This distinction matters indeed: altitudinal changes in climate typically occur on a particularly small spatial scale, where continuous gene flow is likely to impede genetic differentiation unless strong local selection is acting. An insect species with a remarkable altitudinal distribution is the grasshopper Omocestus viridulus (Acrididae: Gomphocerinae). In Switzerland, it occurs in grasslands from below 400 m to above 2500 m (Thorens and Nadig 1997), making it a suited system for the study of altitudinal adaptation. Over the whole range, it displays an annual life cycle, which includes egg hatch in spring, four larval instars followed by adult molt in summer, and reproduction until autumn. The species overwinters as an egg in embryonic diapause (Ingrisch and Ko¨hler 1998). This dormant phase is characterized by suppressed development, reduced metabolism, and high tolerance to harsh environmental conditions (Danks 1987, Leather et al. 1993). The developmental stage during diapause strongly influences later phenology, as it determines how many developmental steps an embryo has to pass through before hatch in spring. In addition, seasonal timing can be achieved through adjustment of development rates (Danks 1987). Higher rates of embryonic and larval development allow reaching adulthood, and thus the subsequent reproductive phase, sooner. The diapause stage, as well as embryonic and larval development rates, can be identified as chief traits determining phenology, and are therefore of high significance to geographic life history adaptation. Along the altitudinal gradient, O. viridulus faces a decline in the length of the growing season, which is likely to require phenological adjustment. Moreover, the species is sedentary (Ingrisch and Ko¨hler 1998) and most populations are separated to some extent by migration barriers, suggesting rather low levels of gene flow. For these reasons, we hypothesize that local life history adaptation, rather than phenotypic plasticity, allowed the grasshopper to extend its distribution to the wide range of altitudes. In this case, the species would represent a fine-grained patchwork of local demes, which are differentiated in traits relevant to seasonal timing. We address the hypothesis by both field and laboratory approaches. In a first step, the natural temperature regimes and their effect on field phenologies along the gradient are explored. In a second step, we compare populations with respect to developmental rates and diapause characteristics in common laboratory environ734

ments. The latter approach serves to remove environmental variation and reveal genetic differences in life histories, if they occur.

Materials and methods Study populations The present investigation includes a total of eleven Omocestus viridulus populations. The study sites were chosen to form a transect from the Swiss lowland into the Alps, covering an altitudinal gradient of 2000 m (Fig. 1). Distribution data were provided by the Swiss center of cartography of the fauna (CSCF). We considered only sites where large populations had been reported over several years. Some 110 km separate the furthermost sites. Although this spatial scale is relatively small, the study populations can be viewed as reasonably independent, as most populations of this widespread species are isolated to some degree by natural and human dispersal barriers (e.g. forests or farmland).

Field studies To estimate the length of the growing season in the field, temperatures were recorded hourly during the 2002 season at sites 1, 5, 7 and 10 (see Fig. 1) by means of data loggers (‘‘StowAway TidbiT’’, Onset Computer Corporation, Bourne, MA, USA). We were primarily interested in the conditions the embryos (eggs) experience. Since O. viridulus lays its clutches into the top soil layer or at the base of grass tussocks (Ingrisch 1983, Berner unpubl.), we positioned the loggers’ sensors at 1 cm soil depth under natural vegetation cover. Two loggers were used per site, and their measurements averaged for all calculations. Two different indices of

Fig. 1. The rectangle in the outline map shows the location of the study area in Switzerland. The sampling sites and corresponding altitudes are 1) Neerach 410 m, 2) Birchwil 540 m, 3) Scho¨nenberg 670 m, 4) Ba¨retswil 830 m, 5) Bendel 1055 m, 6) Na¨fels 1350 m, 7) Speer 1610 m, 8) Flumserberg 1850 m, 9) Gamserrugg 2060 m, 10) Pizol 2215 m, 11) Ho¨rnli 2440 m. The triangle denotes the city of Zu¨rich (47822?N, 8831?E). ECOGRAPHY 27:6 (2004)

season length were computed: one index uses the date at which 118C was exceeded for the first time. This date roughly corresponds to the initiation of postdiapause embryonic development, which is inhibited at temperatures below ca 118C (Wingerden et al. 1991). The second index uses the cumulative degree hours above 148C between the first appearance of larvae at each site and the end of the year. This approximates the season length for larvae and adults. 148C was chosen based on a study by Hilbert and Logan (1983), because postembryonic development thresholds were unavailable for the species. However, grasshoppers are known to increase body temperature by basking (Begon 1983, Chappell 1983). Hence, the latter index must be viewed as a relatively crude, but still informative, estimate of the thermally effective season length. Field phenologies were studied at the same sites and in the same year as the temperature records. Each site was visited in regular intervals of six to eleven days over the growing season. We censused by direct observation along transects, noting the stage (larval instars 1 /4, adult) of each grasshopper. (The insertion of an additional larval instar reported from other gomphocerine grasshoppers (references in Ingrisch and KoÂ¨hler 1998), occurred neither in the field nor laboratory.) Although males reach adulthood slightly earlier than females in this species, the sexes were pooled post hoc for simplicity. This did not influence the results substantially. As 2002 was a rather cool and cloudy year and 2003 was particularly sunny, we also checked the stage composition at site 1 in late June and site 10 in mid July 2003. These snapshots during the second year allowed a comparison of phenologies between climatically rather different seasons.

Breeding techniques and laboratory experiments To establish breeding populations, ca 14 individuals of each sex were caught at the beginning of the reproductive period at each of the eleven sites. The populations 2, 5, 7 and 10 were sampled in 2001, all others in 2002. The animals were kept in groups in cages in a greenhouse under natural photoperiods until death. Field-cut grass (largely of Dactylis glomerata and Agropyron repens ) was provided as food. Egg pods were collected twice a week as they were laid, put in plastic vials containing moist vermiculite, and incubated at 258C for 35 d which allows the embryos to reach the diapause stage (Wingerden et al. 1991). After this, the clutches were stored at 58C. Postdiapause embryonic development time was studied in a climate chamber set to a photoperiod of 14 h at 278C. Night temperature was 88C. All eggs had spent at least three month at 58C, which is enough to break diapause (Ingrisch and KoÂ¨hler 1998). The vials were ECOGRAPHY 27:6 (2004)

inspected twice daily for newly hatched larvae, until no further hatch occurred. Individual hatch dates were noted and converted to degree hours with 118C as threshold. We tested clutch medians in a general linear model (GLM), with study year as a fixed factor and altitude of origin ( /populations) as a continuous covariate. Effective sample size varied between 18 and 79 clutches per population. To verify the robustness of our 278C results, a subset of the clutches laid in 2002 was incubated at 198C, but otherwise treated and analyzed in the same way. Here, sample size varied between 14 and 48. To investigate larval development time, hatchlings were immediately transferred to another climate chamber with a 16 h photoperiod at 328C and a night temperature of 108C. This experiment was conducted with the seven 2002 populations only. Larvae were kept clutchwise in plastic containers (19 cm high, 8 cm in diameter) in groups of six at most. Small pots with a grass mixture provided food. Adult emergence was checked twice daily, and individual degree hours for larval development were determined using 148C as threshold. We analyzed clutch medians using GLM. Altitude was entered as a covariate, and sex as a fixed factor, since the sexes differed in development time. To assess diapause stages, ten random clutches from each of six populations (1, 2, 4, 7, 8, 9) were removed from the cold. The outer layer (chorion) of every single egg in the clutch was scraped off with a fine blade under a stereomicroscope so that the embryo could be seen and assigned a developmental stage. We used the classification system of Cherrill (1987), which divides the continuous process of differentiation to the fully developed embryo into twenty discrete morphological steps. Based on individual eggs, the clutch median stage was determined and treated as one data point. Differences between populations were tested using one-way ANOVA and a distribution-free Kruskal-Wallis test. All statistics were performed with SPSS 11.1.

Results Temperature regimes Daily mean temperatures of the top soil layer display a sharp decline with altitude (Fig. 2). Over the summer months, mean temperatures at 2215 m remain ca 78C below those at 410 m. Moreover, snow cover maintains spring temperatures around zero at the high elevation sites, most dramatically at 2215 m. Indeed, the very first hourly temperature record above the estimated embryonic threshold of 118C occurs as late as on the 31 May at 2215 m (Table 1). At the low elevation sites this threshold is exceeded almost three month earlier. Season length estimated as degree hours above 148C shows a more than tenfold reduction from 410 m to 2215 m 735

Fig. 2. Daily mean temperatures during the 2002 season at altitudes of 410 m (1), 1055 m (2), 1610 m (3) and 2215 m (4).

altitude (Table 1). Hardly any hourly records above 148C were made after the end of August at the highest site. Roughly speaking, postembryonic development was possible during seven months at low altitude, whereas only three months were available at the highest site in 2002.

population reaches adulthood by late June in both years. At 2215 m, however, the phenological difference between the two seasons is much larger. Clearly, the high altitude grasshoppers are delayed in both 2002 and 2003 relative to the lowland, but the phenological delay is more pronounced in the cooler year of 2002.

Field phenologies

Laboratory experiments

The phenology curves in Fig. 3 indicate a marked delay in the emergence of first instar larvae at the high elevation sites, where the first hatchlings appeared three (1610 m) and seven (2215 m) weeks later than at the lowest location. The delay carries over to the adults: the graphically estimated dates at which each population reaches an adult frequency of 75% are 26 June (410 m), 31 June (1055 m), 23 July (1610 m) and 19 August (2215 m). Consequently, adult emergence at the highest site is delayed by almost two months compared to the lowland site. During the 2002 season lowland grasshoppers had already started reproducing when first instar larvae just started hatching at high altitude. In accordance with the temperature regimes, the difference in phenology between the sites at 410 m and 1055 m is small. The comparison of 2002 (cool year) and 2003 (warm year) reveals a small difference in the low elevation phenology (Table 2). The greatest majority of the 410 m

There is a clear relationship between embryonic development time in the laboratory at 278C and a populationâ&#x20AC;&#x2122;s altitude of origin (Fig. 4): high altitude embryos complete development faster, resulting in earlier hatching of the first instar larvae (F1,431 /50.9, pB/0.001). However, the maximal difference between populations in development time amounts to some ten percent only. Expressed in real time, the population averages declined from 14.1 to 12.3 d. The year factor is also significant because temperature conditions differed slightly between the years (different climate chamber types; F1,431 /40.5, p B/0.001). Faster development of the high altitude embryos was also found at the lower experimental temperature of 198C. The correlation of population averages of embryonic development time at the two incubation temperatures yields coefficients of 0.85 (Pearsonâ&#x20AC;&#x2122;s r, p /0.015) and 0.93 (Spearmanâ&#x20AC;&#x2122;s rank, p /0.003). The duration of development through all larval instars to adults clearly gets shorter with altitudinal origin (Fig. 5; F1,223 /51.9, pB/0.001), similar to embryonic development. As in the field, males always reach adulthood earlier than females (F1,223 /16.1, pB/ 0.001), but the altitudinal response is similar in the sexes, as indicated by a nonsignificant interaction (F1,223 /1.3, p /0.26). Again, the difference between the fastest and slowest population is only about ten percent. In real time, the population averages for larval development ranged from 23.6 to 20.9 d in males and

Table 1. Indices of the 2002 season length at four altitudes, based on hourly temperature records at 1 cm soil depth. Altitude (m) 410 1055 1610 2215

Date of first record /118C 8 March 8 March 5 April 31 May

Degree hours /148C* 12536 10019 3182 904

* From the onset of larval hatch to the end of the year.

736

ECOGRAPHY 27:6 (2004)

Fig. 3. Field phenologies of O. viridulus at four altitudes during the season of 2002. The vertical line represents the first observation of hatchlings.

from 25.2 to 22.3 d in females for low and high altitude, respectively. The stage of embryonic diapause does not significantly differ between the populations, and no altitudinal trend is evident (Fig. 6; ANOVA F5,54 /1.34, p /0.26; Kruskal-Wallis x25 /8.09, p /0.15). In all O. viridulus populations studied, the vast majority of embryos diapauses at developmental stage nine, which corresponds to stage IVd of Cherrill (1987). The embryos are then arrested just before the onset of embryonic rotation. Most clutches contained some retarded eggs, but no single embryo developed further than stage nine.

Discussion Our laboratory study documents increasing rates of embryonic and juvenile development in O. viridulus with increasing altitude. As a consequence, high altitude grasshoppers attain adulthood in shorter time than their low altitude counterparts when grown in a common environment. In contrast, the diapause stage, another

key determinant of phenology, shows no difference among the populations. The field work indicates a time constraint on the life cycle of high altitude animals. The cooler high elevation temperature regimes substantially delay larval hatch and adulthood. In a cloudy year like 2002, the reproductive life span of alpine grasshoppers is thus severely truncated and reproductive success very poor. Moreover, a considerable fraction of the produced eggs may fail to reach the overwintering stage due to insufficient late season heat. This was shown to entail delayed hatching in Chorthippus brunneus (Cherrill and Begon 1991) and survival costs in Camnula pellucida (Pickford 1966). Certainly, cool and cloudy years are severe selection events at the speciesâ&#x20AC;&#x2122; upper range margin. Only during particularly sunny seasons like 2003 is the reproductive period sometimes terminated

Table 2. Frequency (%) of O. viridulus instars in the years 2002 and 2003 at low and high elevation. Note that the two populations were censused on different dates.

2nd instar 3rd instar 4th instar Adult

410 m, late June

2215 m, mid July

2002

2003

2002

2003

/ 3 18 79

/ / 3 97

21 64 15 /

/ 9 31 60

ECOGRAPHY 27:6 (2004)

Fig. 4. Physiological time required by O. viridulus populations from different altitudes for postdiapause embryonic development. Data from 2002 (k) and 2003 (m). Degree hours were calculated using 118C as threshold.

737

Fig. 5. Physiological time required by female (k) and male (m) grasshoppers from different altitudes for larval development. A threshold of 148C was used for calculation.

by intrinsic senescence at both low and high elevation. Thus the variance in the available season length increases with altitude. Under such a seasonal time constraint, annual organisms face the problem of optimally allocating time to development and reproduction. Life history models predict that decreased season length will favor faster development and hence decreased time to maturity (Cohen 1976, Roff 1980, 2002, Rowe and Ludwig 1991, Abrams et al. 1996). In Omocestus viridulus with its wide altitudinal distribution, therefore, we expected differences in traits determining postdiapause development time. The higher rates of development exhibited under laboratory conditions by the alpine populations thus conform well to the theoretical prediction. A genetic basis to the acceleration of development is strongly suggested because, firstly, maternal influence on offspring embryonic development appears negligible in the related Chorthippus parallelus (KoÂ¨hler 1983). Secondly, our field records indicate that the temperatures used in the laboratory may be experienced by all populations in the field. Absorption of solar radiation may allow even

Fig. 6. Embryonic developmental stage at diapause in O. viridulus populations from six altitudes. Plotted are clutch medians (N /10 per population).

738

high altitude larvae to rise body temperature well above 328C. Furthermore, embryonic development rates were found to increase with altitude at incubation temperatures of both 198C and 278C. Strong genotype by environment interactions are thus excluded. For these reasons we suggest that the observed developmental differences are robust and reflect an adaptive strategy. Increased embryonic and larval development rates allow high altitude animals to reach maturity relatively faster, and hence prolong reproductive life span when time is short. The hypothesis of within species life history differentiation on a small spatial scale is thus confirmed. Apparently the level of gene flow between the O. viridulus populations is too low to counteract local adaptation. Omocestus viridulus agrees well with some other ectotherms in which differentiation of development along gradients in season length has been documented (Masaki 1967, Berven and Gill 1983, Dingle et al. 1990, Ayres and Scriber 1994, Telfer and Hassall 1999). The shortening of development time proves a common adaptive response to seasonal time constraints. Intraspecific differentiation in traits related to phenology may be quite frequent in annual ectotherms with relatively long development times covering wide geographic ranges. However, our field surveys make it clear that the cool climates at high elevation retard grasshopper phenologies despite higher intrinsic rates of development in those populations. Thus, the high elevation grasshoppers are only partly able to compensate the delaying environmental influence on time to adulthood. This agrees with the relatively modest level of differentiation found in the laboratory. As the genetic response along the altitudinal gradient is opposed to the phenotypic response to the environmental conditions, O. viridulus provides an example of countergradient variation (Conover and Schultz 1995). A merely phenotypic comparison of development times within the species would have failed to demonstrate altitudinal differentiation. At the proximate level, increased development rates may be associated with metabolic temperature compensation (Danks 1987). Hadley and Massion (1985) for example report increased metabolic rates in high altitude populations of the grasshopper Aeropedellus clavatus. Likewise, latitudinal differences in metabolism were found in the butterfly Papilio canadensis (Ayres and Scriber 1994). However, physiological traits of O. viridulus populations have not been compared so far. A question arising from the observed patterns is why higher intrinsic rates of development did not evolve in the lowland populations. What could be the disadvantage of a similarly rapid development as at high elevation? On the one hand, adverse climatic conditions early in the season are likely to select against precocious ECOGRAPHY 27:6 (2004)

larval emergence. Carrie`re et al. (1996), for example, demonstrate a mortality cost associated with precocious larval hatch in Gryllus pennsylvanicus, due to unfavorable temperature conditions. Furthermore, trade-offs with other fitness components could maintain developmental rates below the physiological potential exhibited by the alpine populations (Schluter et al. 1991, Stearns 1992, Roff 2002). For instance, Tatar et al. (1997) found increased senescence in Melanoplus sanguinipes grasshoppers from high elevation sites compared to the slower developing low altitude animals. Likewise, given that juvenile development time, adult size, and fecundity are often correlated positively (Roff 1980, 2002, Rowe and Ludwig 1991, Honek 1993), a shortened juvenile development will negatively affect fecundity. According to Orr (1996) this is the case in M. sanguinipes. Most probably, elevated rates of development bear fitness costs and are selected against in the absence of a seasonal time constraint on the life cycle, as is the case at low elevation. However, low altitude seasons appear still too short for two generations, as the species exhibits an annual life cycle throughout its range. Besides embryonic and larval development rates, the stage of overwintering strongly determines time to adulthood. Central European grasshoppers of the gomphocerine subfamily are believed to show an obligatory diapause during embryonic development. According to some studies, the dormant stage is inserted shortly before embryonic rotation (KoÂ¨hler 1991, Ingrisch and KoÂ¨hler 1998). This stage has been designated IVd by Cherrill (1987). However, geographic variation in diapause stage within an insect species is possible in principle (references in Tauber et al. 1986), but has not been investigated to date in any European grasshopper. In O. viridulus we found no such altitudinal differentiation: in all populations, most clutches mainly contained embryos arrested at the aforementioned stage, and no embryo developed further. Thus, the species displays an obligatory diapause and is uniform with respect to the stage of dormancy, conforming to other members of the subfamily. This finding stands in striking contrast to other orthopteran studies, which document intraspecific variation in embryonic diapause stage and/or expression along gradients in season length (Mousseau and Roff 1989, Groeters and Shaw 1992, Tanaka 1994, Dingle and Mousseau 1994, Bradford and Roff 1995). The catantopine grasshopper Melanoplus sanguinipes, for example, occurs from sea level to above 3800 m (Chappell 1983) and displays enormous variation in embryonic diapause stage within its North American range. High elevation populations overwinter almost completely developed and attain the hatching state at low heat sums. This is interpreted as an effective means to decrease postdiapause development time under short seasons (Dingle et al. 1990, Dingle and Mousseau 1994). In M. sanguinipes, adult emergence at above 2600 m and at sea level ECOGRAPHY 27:6 (2004)

happens roughly at the same time! Not surprisingly, another catantopine, M. frigidus, is the highest reaching species in the Alps (Carron 1996). This grasshopper subfamily illustrates the importance of flexibility in the overwintering stage for altitudinal adaptation. In this light, the lack of variation in the stage of diapause within O. viridulus likely represents a phylogenetic constraint (Gould 1989, Stearns 1992, Schlichting and Pigliucci 1998) to altitudinal range expansion. The stage of dormancy as a conserved trait within the gomphocerine lineage precludes tuning of development in a way expected to be optimal under seasonal time constraints. However, this has to be confirmed by investigating other, closely related species. To summarize, O. viridulus exhibits altitudinal differentiation in development as an adaptive response to selection imposed by local climates. However, the potential for altitudinal adaptation is limited by the invariant stage of overwintering diapause, probably indicating a phylogenetic constraint within the gomphocerine grasshopper lineage. As a consequence, the degree of climatic compensation displayed by field populations along the altitudinal gradient is rather low as compared to other ectotherms. Acknowledgements / We thank T. Walter for organizing a parttime employment for D. Berner at the Swiss Federal Research Station for Agroecology and Agriculture (Agroscope FAL Reckenholz), which made this research project possible. J. Samietz, G. KoÂ¨hler and E.-F. Kiel gave valuable advice regarding the research design and/or grasshopper rearing techniques. S. Bosshart and M. Waldburger provided technical help and materials. P. Streckeisen kindly helped operate the climate chambers and made greenhouse space available. Distribution data of the species were obtained from the Swiss center of cartography of the fauna (CSCF). To all these people and institutions we are most grateful.

References Abrams, P. A. et al. 1996. The effect of flexible growth rates on optimal sizes and development times in a seasonal environment. / Am. Nat. 147: 381 /395. Ayres, M. P. and Scriber, J. M. 1994. Local adaptation to regional climates in Papilio canadensis (Lepidoptera: Papilionidae). / Ecol. Monogr. 64: 465 /482. Begon, M. 1983. Grasshopper populations and weather: the effects of insolation on Chorthippus brunneus. / Ecol. Entomol. 8: 361 /370. Berven, K. A. and Gill, D. E. 1983. Interpreting geographic variation in life-history traits. / Am. Zool. 23: 85 /97. Blanckenhorn, W. U. 1997. Altitudinal life history variation in the dung flies Scathophaga stercoraria and Sepsis cynipsea . / Oecologia 109: 342 /352. Blanckenhorn, W. U. and Fairbairn, D. J. 1995. Life history adaptation along a latitudinal cline in the water strider Aquarius remigis (Heteroptera: Gerridae). / J. Evol. Biol. 8: 21 /41. Bradford, M. J. and Roff, D. A. 1995. Genetic and phenotypic sources of life history variation along a cline in voltinism in the cricket Allonemobius socius. / Oecologia 103: 319 /326. Carrie`re, Y., Simons, A. M. and Roff, D. A. 1996. The effect of the timing of post-diapause egg development on survival,

739

growth and body size in Gryllus pennsylvanicus. / Oikos 75: 463 /470. Carron, G. 1996. Do alpine acridids have a shortened postembryonic development? / Articulata 11: 49 /72. Chappell, M. A. 1983. Metabolism and thermoregulation in desert and montane grasshoppers. / Oecologia 56: 126 /131. Cherrill, A. and Begon, M. 1991. Oviposition date and pattern of embryogenesis in the grasshopper Chorthippus brunneus (Orthoptera: Acrididae). / Holarct. Ecol. 14: 225 /233. Cherrill, A. J. 1987. The development and survival of the eggs and early instars of the grasshopper Chorthippus brunneus (Thunberg) in north west England. / Unpubl. Ph.D. thesis, Univ. of Liverpool. Cohen, D. 1976. The optimal timing of reproduction. / Am. Nat. 110: 801 /807. Conover, D. O. and Schultz, E. T. 1995. Phenotypic similarity and the evolutionary significance of countergradient variation. / Trends Ecol. Evol. 10: 248 /252. Danks, H. V. 1987. Insect dormancy: an ecological perspective. / Biological Survey of Canada. Danks, H. V. 1994. Insect life-cycle polymorphism: theory, evolution and ecological consequences for seasonality and diapause control. / Kluwer. Dingle, H. and Mousseau, T. A. 1994. Geographic variation in embryonic development time and stage of diapause in a grasshopper. / Oecologia 97: 179 /185. Dingle, H., Mousseau, A. and Scott, S. M. 1990. Altitudinal variation in life cycle syndromes of California populations of the grasshopper, Melanoplus sanguinipes (F.). / Oecologia 84: 199 /206. Gotthard, K. and Nylin, S. 1995. Adaptive plasticity and plasticity as an adaptation: a selective review of plasticity in animal morphology and life history. / Oikos 74: 3 /17. Gould, S. J. 1989. A developmental constraint in Cerion , with comments on the definition and interpretation of constraint in evolution. / Evolution 43: 516 /539. Groeters, F. R. and Shaw, D. D. 1992. Association between latitudinal variation for embryonic development time and chromosome structure in the grasshopper Caledia captiva (Orthoptera: Acrididae). / Evolution 46: 245 /257. Hadley, N. F. and Massion, D. D. 1985. Oxygen consumption, water loss and cuticular lipids of high and low elevation populations of the grasshopper Aeropedellus clavatus (Orthoptera: Acrididae). / Comp. Biochem. Physiol. 80A: 307 /311. Hilbert, D. W. and Logan, J. A. 1983. Empirical model of nymphal development for the migratory grasshopper, Melanoplus sanguinipes (Orthoptera: Acrididae). / Environ. Entomol. 12: 1 /5. Honek, A. 1993. Intraspecific variation in body size and fecundity in insects: a general relationship. / Oikos 66: 483 /492. Ingrisch, S. 1983. Zum Einfluss der Feuchte auf die Schlupfrate und Entwicklungsdauer der Eier mitteleuropa¨ischer Feldheuschrecken (Orthoptera: Acrididae). / Dtsch. Ent. Z. N. F. 30: 1 /15. Ingrisch, S. and Ko¨hler, G. 1998. Die Heuschrecken Mitteleuropas. / Westarp. Kingsolver, J. G. and Huey, R. B. 1998. Evolutionary analyses of morphological and physiological plasticity in thermally variable environments. / Am. Zool. 38: 545 /560. Ko¨hler, G. 1983. Investigations on the hatching polymorphism and its intrapopular consequences in Chorthippus parallelus (Zetterstedt) (Orthoptera: Acrididae). / Zool. Jb. Syst. 110: 31 /44. Ko¨hler, G. 1991. Investigations on diapause and non-diapause in central European grasshopper eggs (Acrididae: Gomphocerinae). / Zool. Jb. Syst. 118: 323 /344.

740

Lamb, R. J., MacKay, P. A. and Gerber, G. H. 1987. Are development and growth of pea aphids, Acyrtosiphon pisum , in North America adapted to local temperatures? / Oecologia 72: 170 /177. Leather, S. R., Waters, K. F. A. and Bale, J. S. 1993. The ecology of insect overwintering. / Cambridge Univ. Press. Masaki, S. 1967. Geographic variation and climatic adaptation in a field cricket (Orthoptera: Gryllidae). / Evolution 21: 725 /741. Merila¨, J. et al. 2000. Plasticity in age and size at metamorphosis in Rana temporaria / comparison of high and low latitude populations. / Ecography 23: 457 /465. Mousseau, T. A. and Roff, D. A. 1989. Adaptation to seasonality in a cricket: patterns of phenotypic and genotypic variation in body size and diapause expression along a cline in season length. / Evolution 43: 1483 /1496. Nylin, S. 1994. Seasonal plasticity and life-cycle adaptation in butterflies. / In: Danks, H. V. (ed.), Insect lifecycle polymorphism: theory, evolution and ecological consequences for seasonality and diapause control. Kluwer, pp. 41 /67. Orr, M. 1996. Life-history adaptation and reproductive isolation in a grasshopper hybrid zone. / Evolution 50: 704 /716. Pickford, R. 1966. The influence of date of oviposition and climatic conditions on hatching of Camnula pellucida (Acridiae). / Can. Entomol. 98: 1145 /1159. Roff, D. 1980. Optimizing development time in a seasonal environment: the ‘ups and downs’ of clinal variation. / Oecologia 45: 202 /208. Roff, D. A. 2002. Life history evolution. / Sinauer. Rowe, L. and Ludwig, D. 1991. Size and timing of metamorphosis in complex life cycles: time constraints and variation. / Ecology 72: 413 /427. Schlichting, C. D. and Pigliucci, M. 1998. Phenotypic evolution: a reaction norm perspective. / Sinauer. Schluter, D., Price, T. D. and Rowe, L. 1991. Conflicting selection pressures and life history trade-offs. / Proc. R. Soc. Lond. B 246: 11 /17. Slatkin, M. 1987. Gene flow and the geographic structure of natural populations. / Science 236: 787 /792. Stearns, S. C. 1992. The evolution of life histories. / Oxford Univ. Press. Tanaka, S. 1994. Diapause as a pivotal factor for latitudinal and seasonal adaptation in Locusta migratoria in Japan. / In: Danks, H. V. (ed.), Insect life-cycle polymorphism: theory, evolution and ecological consequences for seasonality and diapause control. Kluwer, pp. 173 /190. Tanaka, S. and Brookes, V. J. 1983. Altitudinal adaptation of the life cycle in Allonemobius fasciatus DeGeer (Orthoptera: Gryllidae). / Can. J. Zool. 61: 1986 /1990. Tatar, M., Gray, D. W. and Carey, J. R. 1997. Altitudinal variation for senescence in Melanoplus grasshoppers. / Oecologia 111: 357 /364. Tauber, M. J., Tauber, C. A. and Masaki, S. 1986. Seasonal adaptations of insects. / Oxford Univ. Press. Taylor, F. and Karban, R. 1986. The evolution of insect life cycles. / Springer. Telfer, M. G. and Hassall, M. 1999. Ecotypic differentiation in the grasshopper Chorthippus brunneus : life history varies in relation to climate. / Oecologia 121: 245 /254. Thorens, P. and Nadig, A. 1997. Atlas de distribution des Orthopte`res de Suisse. / Centre suisse de cartographie de la faune (CSCF), Documenta Faunistica Helvetiae 16. Wingerden, W. K. R. E. van, Musters, J. C. M. and Maaskamp, F. I. M. 1991. The influence of temperature on the duration of egg development in west European grasshoppers. / Oecologia 87: 417 /423.

ECOGRAPHY 27:6 (2004)

ECOGRAPHY 26: 456–466, 2003

Unique elevational diversity patterns of geometrid moths in an Andean montane rainforest Gunnar Brehm, Dirk Su¨ssenbach and Konrad Fiedler

Brehm, G., Su¨ssenbach, D. and Fiedler, K. 2003. Unique elevation diversity patterns of geometrid moths in an Andean montane rainforest. – Ecography 26: 456– 466. Alpha-diversity of geometrid moths was investigated along an elevational gradient in a tropical montane rainforest in southern Ecuador. Diversity was measured using 1) species number, 2) extrapolated species number (Chao 1 estimator), 3) rarefied species number, and 4) Fisher’s alpha. When applied to the empirical data set, 1 and 2 strongly depended on the sample size, whereas 3 and 4 were suitable and reliable measures of local diversity. At single sites, up to 292 species were observed, and extrapolation estimates range from 244 to 445 species. Values for Fisher’s alpha are among the highest ever measured for this moth family, and range from 69 to 131 per site. In contrast to theoretical assumptions and empirical studies in other regions of the world, the diversity of geometrid moths remained consistently high along the entire gradient studied. Diversity measures correlated with neither altitude nor ambient temperature. The large subfamily Ennominae has previously been assumed to be a group that occurs mainly at low and medium elevations. However, no decline in diversity was found in the study area. The diversity of the other large subfamily, Larentiinae, even increased from the lowest elevations and was highest at elevations above 1800 m. The roles of a decreasing diversity of potential host-plants, decreasing structural complexity of the vegetation, increasingly unfavourable climatic conditions and possible physiological adaptations in determining herbivore species richness are discussed. A relatively low predation pressure might be an advantage of high-altitude habitats. The physiognomy of the Andes (folded mountains, large areas at high altitudes) might also have allowed speciation events and the development of a species-rich high-altitude fauna. There is evidence that the species-richness of other groups of herbivorous insects in the same area declines as altitude increases. This emphasises difficulties that are associated with biodiversity indicator groups, and calls for caution when making generalisations from case studies. G. Brehm (gunnar – brehm@yahoo.com), D. Su¨ssenbach and K. Fiedler, Dept of Animal Ecology I, Uni6. of Bayreuth, D-95440 Bayreuth, Germany.

Biodiversity research in tropical regions has largely focussed on lowland rainforests, whereas montane rainforests have so far largely been neglected. This is particularly true with regard to studies on arthropods. For example, none of the 89 studies on canopy arthropods reviewed by Basset (2001) was conducted in a tropical montane habitat. The northern Andes have been recognised as a ‘‘hyper hotspot’’ on earth for vascular plants and vertebrates (Henderson et al. 1991, Myers et al. 2000). However, apart from a few exceptions, little is

known about patterns of insect diversity in Andean montane forest habitats. The dominant environmental gradient in mountainous habitats is altitude, which is directly related to decreasing temperature and various other abiotic and biotic environmental factors. Significant changes in the vegetation along altitudinal gradients in the Andes were first documented by Alexander von Humboldt (von Humboldt and Bonpland 1807). Later studies yielded increasingly refined pictures (e.g. Grubb et al. 1963,

456

ECOGRAPHY 26:4 (2003)

Gentry 1988). Elevational changes in diversity have also been documented for a variety of animals in the Neotropical region (references in Brehm 2002). In a review on species richness patterns along elevational gradients, Rahbeck (1995) found that approximately half of the studies showed a continuous decline of species numbers with increasing altitude, whereas the other half detected a peak at medium elevations. Although some authors suggested that the occurrence of mid-elevational diversity peaks might be a sampling artefact (e.g. Wolda 1987), there is now substantial evidence for the existence of such peaks in a broad range of organisms (McCoy 1990, Grytnes and Vetaas 2002 with references therein). Among insects, several families of Lepidoptera in south east Asia have been reported to exhibit their greatest diversity at medium elevations, between 600 and 1000 m a.s.l. (Holloway 1987, Holloway and Nielsen 1999). However, only a few exceptions to an overall declining diversity at altitudes higher than 1000 m have been documented for arthropods. Examples include tropical larentiine moths (a subfamily of Geometridae), which are most speciesrich at high altitudes, and also increase in relation to other subfamilies with latitude in temperate regions (Brehm and Fiedler 2003). Schulze (2000) reported that diversity of the moth families Geometridae and Arctiidae in Borneo attained richness maxima at altitudes between 1200 and 2000 m. Above this altitude, all available studies indicate that diversity of insect assemblages decreases. This suggests that environmental conditions within and above the cloud-forest zone of tropical mountains become so harsh that the number of ectothermic insect species that are able to cope with such circumstances decreases steadily above this level. The moth family Geometridae is one of the three largest clades of Lepidoptera and currently includes \21000 described species (Scoble 1999), with some 6400 (30%) occurring in the Neotropical region. Compared to most arthropod taxa, the taxonomy of geometrid moths is relatively advanced, and they are a suitable group for biodiversity studies in tropical forests (Intachat and Woiwod 1999). However, prior to the present study, no detailed ecological studies of geometrids in the Neotropical region have ever been conducted. Samples taken from tropical arthropod communities provide a methodological challenge for diversity measures. They are almost always incomplete and the numbers of specimens available for analysis often diverge considerably between sites (e.g. Schulze and Fiedler 2003). Moreover, tropical arthropod communities are characterised by a high proportion of rare species that cannot be excluded as artefact or a group of marginal importance (Price et al. 1995, Novotny and Basset 2000). From the plethora of available measures of alpha-diversity (e.g. Hayek and Buzas 1997, Southwood and Henderson 2000), four were selected and their ECOGRAPHY 26:4 (2003)

sample size dependence was tested. Suitable diversity measures should be able to discriminate between samples of different diversity, but at the same time be independent of sample size, in order to avoid misleading bias in the results. In this first study of its kind in the Neotropical region, we aimed to investigate diversity patterns of Geometridae as model organisms of a ‘‘mega-diverse’’ group of herbivorous insects in a montane Andean rainforest. An analysis of changes of faunal composition and beta diversity of geometrid moths was provided by Brehm and Fiedler (2003) and Brehm et al. (2003). In this paper, we tested the expectation, derived from all comparable studies, that species diversity should decline in the upper part of the elevational gradient.

Methods Study area, sampling, identification, temperature measurements The study area in southern Ecuador is situated within the East Cordillera of the Andes and belongs to the province Zamora-Chinchipe (Reserva Biolo´ gica San Francisco, 3°58%S, 79°5%W, and adjacent fractions of the Podocarpus National Park). It is covered with undisturbed to slightly disturbed montane rainforest. The vegetation in the upper part of the study area was described by Bussmann (2001) and Paulsch (2002). Moths were sampled manually using portable weak light-traps (2 ×15 W). Traps consisted of a white gauze cylinder (height 1.60 m, diameter 0.60 m) and were placed at the forest floor. We selected 22 sampling sites at 11 elevational levels between 1040 and 2677 m a.s.l. Sites were numbered from 1 to 11 (a + b) according to their elevational order. A detailed description of the sites and a discussion of the sampling methods was provided by Brehm (2002). Sampling occurred during three field periods (April –May 1999, October 1999 – January 2000, and October –December 2000). Lighttraps were run between 18:30 and 21:30 local time. Between two and four nightly catches from each site were pooled and analysed. Specimens were first sorted to morphospecies level and later determined in the Zoologische Staatssammlung, Munich and the Natural History Museum, London. Fifty-two percent of the taxa and 67% of the specimens were determined to species level (Brehm 2002). In parallel to the light trapping, the temperature during three to eleven nights per site was measured every 30 min from 18:30 to 21:30 using a Casio alti-thermo twin sensor at 1.60 m above ground level. Braun (2002) and Brehm (2002) showed that temperature in the study area declines linearly with increasing elevation. Thus, for each site the average of all available nightly measures taken at 20:00 was used as standard. 457

Alpha-diversity measures The analysis was restricted to four selected measures and was performed separately for Geometridae and the two largest subfamilies Ennominae and Larentiinae. The remaining subfamilies (Desmobathrinae, Geometrinae, Oenochrominae, and Sterrhinae) were not analysed separately because of their low representation in the samples.

1) Species number Measurement of species richness by complete census is only feasible for a few organisms. For most organisms, measurement means sampling (Colwell and Coddington 1994). However, species numbers are still widely used as a measure of diversity. Misleading results and biases must be expected with incomplete samples that differ in size (Gotelli and Colwell 2001). 2) Extrapolation According to Colwell and Coddington (1994), if certain assumptions are not violated, the ‘‘true’’ number can be estimated using extrapolation methods. These authors recommended the use of non-parametric estimators, such as Chao 1, as promising quantitative techniques. Analyses were performed using the computer program EstimateS 6.0b1 (Colwell 2000), with the bias-corrected formula. Since all samples contained at least 380 specimens, the use of ‘‘Chao 1’’ appeared to be justified. However, a certain dependence on sample size was expected, because the recorded number of species is an integral part of the formula of the estimator.

mens is usually recommended for calculating Fisher’s alpha (Hayek and Buzas 1997). This number was not reached in only one sample of the subfamily Larentiinae (site 1a: 65 specimens). All diversity measures were tested for correlation with specimen numbers (i.e. sample size dependence), altitude, and temperature. Since relationships between data in this study are non-linear, the Spearman rank correlation coefficient was used. Multiple tests of significance were Bonferroni-corrected according to Hochberg (1988). All standard statistical analyses were performed using the software package Statistica 5.5 (Anon. 1999).

Results A total of 14348 specimens were collected from 22 sites. Four hundred and ten specimens (2.9% of the total catch) could not be reliably sorted and had to be discarded. Figure 1A shows the fluctuating number of specimens per catch and site. Specimen numbers cannot be directly compared because of the variable number of nightly catches analysed. The minimum and maximum numbers of specimens per site are 384 and 1200, respectively. Figure 1B shows the total number of species per site, partitioned into the subfamilies. The number of observed species per site ranged from 134 (site 1a) to 292 (site 7a), but actual numbers were expected to be distinctly higher (see next section). Raw species counts showed no relationship with temperature or altitude for Geometridae and Ennominae, but strongly increased at higher elevations for Larentiinae.

3) Rarefaction This method is particularly useful if assemblages are sampled with different intensity or success (Hurlbert 1971, Gotelli and Colwell 2001). The geometrid samples were rarefied to a shared number of specimens using a program developed by Kenney and Krebs (2000). The program also provided standard deviations. Rarefied expected species numbers were calculated at the level of 350 specimens for Geometridae, of 100 specimens for Ennominae, and of 50 specimens for Larentiinae. This measure is expected to be independent of sample size since samples are standardised to an equal level. 4) Fisher’s alpha of the log-series In contrast to other diversity indices, Fisher’s alpha has been shown to be efficient in discriminating between sites, and is mainly influenced by the frequencies of species of medium abundance (Kempton and Taylor 1974). The fit of the log-series distribution was tested using a program developed by Henderson and Seaby (1998). Fisher’s alpha and standard deviations (according to Fisher’s original formula, Colwell pers. comm.) were calculated with the program EstimateS 6.0b1 (Colwell 2000). A minimum number of at least 100 speci458

Fig. 1. A) Number of specimens collected at 22 sites. Sites are sorted by altitude and are partitioned according to the nightly catch and the sampling period. Sp 1999 – Spring 1999 (April – May), Au 1999 – Autumn 1999 (October 1999 – January 2000), Au 2000 – Autumn 2000 (October – November 2000). B) Number of species at 22 sites (sorted by altitude and partitioned across subfamilies). Ster – Sterrhinae, Geom – Geometrinae, Lare – Larentiinae, Enno – Ennominae. Due to the chosen scale, Desmobathrinae and Oenochrominae are not visible. Desmobathrinae: one species at sites 1a and 1b. Oenochrominae: two species at sites 1b and 7b, one species at sites 1a, 2a, 5a, 7a, 7b, 8a, 10a, and 11b. ECOGRAPHY 26:4 (2003)

Table 1. Spearman rank correlations between four measures of alpha-diversity (species number, extrapolated species number (Chao 1), rarefied species number, and Fisher’s alpha) and 1) specimen number, 2) altitude, and 3) mean ambient temperature (at 20:00 local time). 1 Rarefied sample size for which the expected species number was calculated. Species numbers and extrapolated species numbers are highly significantly correlated with specimen numbers and are thus unreliable measures of local diversity. Only two (unreliable) measures in the subfamily Larentiinae are significantly correlated with altitude and temperature. ns not significant, ** pB0.01, *** pB0.005, **** pB0.001. Printed in bold are results that remain significant after sequential Bonferroni correction according to Hochberg (1988).

1) Correlation between specimen number and species number extrapolated species number (Chao 1) rarefied species number Fisher’s alpha 2) Correlation between altitude and species number extrapolated species number (Chao 1) rarefied species number Fisher’s alpha (3) Correlation between temperature and species number extrapolated species number (Chao 1) rarefied species number Fisher’s alpha Rarefaction level 1

Geometridae

Ennominae

0.91 0.58 −0.07 0.15

**** *** ns ns

0.93 0.54 −0.09 0.29

**** ** ns ns

0.94 0.83 0.32 0.11

**** **** ns ns

0.13 0.18 0.01 0.08

ns ns ns ns

−0.39 −0.30 0.57 0.11

ns ns ** ns

0.78 0.65 0.31 0.16

**** *** ns ns

−0.08 −0.14 0.09 0.11

ns ns ns ns

0.42 0.29 −0.54 −0.07

ns ns ** ns

−0.75 −0.60 −0.29 −0.17

**** *** ns ns

350

Extrapolation Extrapolated species numbers are on average 75% larger than the observed numbers in all three taxa. The estimate ranges from 29% (Larentiinae, site 11a) to 218% (Ennominae, site 8b) larger than the recorded species number. Figure 2A shows extrapolation results for Geometridae, Ennominae and Larentiinae for all 22 sites. For Geometridae, estimates range between 244 (site 2b) and 445 (site 7a) expected species per site. There is no relationship between the extrapolated number of species and either altitude or temperature for geometrid moths as a whole and Ennominae, while a distinct increase in species numbers for Larentiinae with altitude is notable (Table 1). However, these estimates have to be regarded with caution since the Chao 1 estimator is not a fully reliable measure of diversity in this particular data set because of its sample size dependence (see below).

Rarefaction Figure 3 shows rarefaction curves of Geometridae for all 22 sites. All curves lie within a relatively narrow band and no clear altitudinal pattern is visible. Figure 2B shows the expected species numbers for geometrid samples rarefied to a standard size of 350 specimens plotted against altitude. While one site (1a at 1040 m) has a significantly lower rarefied species number (122) than all other sites, these sites again range in a continuous band between 135 and 168 expected species and show no tendency along the altitudinal and temperature gradient (Table 1). ECOGRAPHY 26:4 (2003)

100

Larentiinae

Changes do occur at the subfamily level. Figure 2B shows rarefied species numbers for the subfamilies Ennominae and Larentiinae at rarefied sample sizes of 100 and 50 specimens, respectively. The patterns resemble each other. Surprisingly, rarefied species numbers of Ennominae even increase with altitude (Table 1), although this is not significant after sequential Bonferroni correction. A conspicuous difference between both subfamilies is visible at sites 3a, 3b, 4a, and 4b (at 1800 – 1875 m). While rarefied species numbers in Ennominae tended to be lower than in the neighbouring sites, those of Larentiinae tended to be higher.

Fisher’s alpha of the log-series A significant deviation from the log-series occurred in six ensembles of Geometridae, and in four ensembles of Ennominae, but not in Larentiinae (Table 2). However, after performing the sequential Bonferroni procedure suggested by Hochberg (1988), the log-series distribution was rejected only in Geometridae and Ennominae at the same single site (number 1b, Table 2). In all cases of deviation, the observed number of species in the first abundance class was larger than expected, i.e. there were ‘‘too many’’ rare species in the samples. Figure 2C shows Fisher’s alpha values for Geometridae at all 22 sites. They range from 69.3 95.4 to 130.6 9 10.4 and are among the highest ever measured for local geometrid ensembles. There is no significant consistent change in Fisher’s alpha along the altitudinal gradient. Alpha values for Ennominae range from 38.2 9 3.4 to 66.79 6.5, and for Larentiinae from 20.1 9 4.1 to 64.29 6.2. In both subfamilies, values of Fisher’s alpha 459

Fig. 2.

460

ECOGRAPHY 26:4 (2003)

Fisher’s alpha are all independent of sample size. Despite their differences with regard to sample size dependence, all four measures are highly associated (Kendall’s coefficient of concordance: Geometridae rK = 0.95, Ennominae rK = 0.91, Larentiinae rK = 0.98, all p B 0.001).

Discussion Diversity within a world context

Fig. 3. Rarefaction curves for Geometridae at all 22 sites. For clarity, standard deviation curves are omitted. The vertical line indicates the standardised sample size (350 specimens), see Fig. 2B. The lowest rarefied species number was calculated for site 1a (1040 m).

are neither correlated with altitude nor with temperature (Table 1). However, in Larentiinae, Fisher’s alpha is lowest at the lowest elevations (Fig. 2C).

Sample size dependence of diversity measures and concordance Table 1 shows correlations between specimen numbers and different measures of alpha-diversity for Geometridae, Ennominae, and Larentiinae across the 22 study sites. In all three taxa, the recorded species number is strongly correlated with the number of specimens collected. Extrapolated species numbers also show significant correlations in all taxa, although the relationship is less pronounced than with recorded specimen numbers. In contrast, values of rarefied species numbers and Table 2. Nominally significant deviations of samples from the log-series distribution. Provided are p-values from x2-tests after arranging species-abundances in octaves (between 3 and 4 degrees of freedom) (Henderson and Seaby 1998). Printed in bold: significant after sequential Bonferroni correction (Hochberg 1988). Samples from all other sites (levels 2, and 5–11) and all Larentiinae samples did not deviate from the log-series. A complete list of sampling sites and their geographical coordinates was provided by Brehm (2002) and Brehm and Fiedler (2003). Site number

Altitude (m)

Geometridae

Ennominae

1a 1b 3a 3b 4a 4b

1040 1040 1800 1800 1850 1875

0.038 B0.001 0.027 0.015 0.010 0.015

– 0.003 0.048 – 0.006 0.035

Values of Fisher’s alpha of up to 131 per site for the species-rich family Geometridae are among the highest ever measured in the world. While most geometrid ensembles in temperate regions reach values between 10 and 35 (Barlow and Woiwod 1989, Thomas and Thomas 1994, Thomas 2002), the highest values to date have been found in rainforests of Peninsular Malaysia and Borneo at elevations between 200 and 700 m, with alpha scores of 127 and 128, respectively (Barlow and Woiwod 1989, Beck et al. 2002). An all-year-round sampling would probably lead to even higher values of Fisher’s alpha because the sites in Ecuador were only sampled between two and four times. Since species numbers at single sites have been shown to be highly dependent on sample size, and samples are incomplete, raw species counts cannot be compared directly with other data. However, the total number of 1010 nocturnal geometrid moth species collected in the study area is the highest ever counted in a small geographic area (Brehm 2002).

Altitudinal patterns of different geometrid taxa All diversity measures reveal similar results along the altitudinal gradient (all Kendall’s rK \ 0.9). In all three taxa considered here (Geometridae, subfamilies Ennominae and Larentiinae) diversity either remained more or less constantly high or increased over an elevational range from 1040 to 2677 m. The patterns are very remarkable, because a decline of insect diversity towards higher altitudes was expected (Wolda 1987, a review by McCoy 1990, Hanski and Niemela¨ 1990, Bru¨ hl et al. 1999). Insect diversity often does not decrease steadily along elevational gradients, but peaks at medium elevations (Janzen et al. 1976, Holloway 1987, Olson 1994, Chey 2000). Our study is the first that revealed such an extended elevational diversity maximum of a very species-rich insect group at very high altitudes. So far, a similar pattern has only been observed in the geometrid subfamily Larentiinae in south

Fig. 2. Diversity of Geometridae (left), and the largest subfamilies Ennominae and Larentiinae (right) along an altitudinal gradient ranging from 1040 to 2677 m in a montane rainforest in southern Ecuador. Diversity measured A) by extrapolation with the estimator Chao 1, B) by rarefaction (the level of specimens to which samples have been rarefied is indicated), and C) with Fisher’s alpha. Error bars indicate 9 1 SD. ECOGRAPHY 26:4 (2003)

461

east Asia (Holloway et al. 1990, Schulze 2000). However, above 2000 m, even in Larentiinae only declining diversity has been detected so far. The high diversity of the subfamily Ennominae at high elevations was unexpected because this taxon was previously thought to be most diverse at lower montane zones (Holloway et al. 1990). We assume that the diversity of geometrid moths in the Andes declines only at very high altitudes in the transition zone between montane cloud rainforest and the paÂ´ ramo vegetation. Indeed, sporadic sampling in the summit region of the study area (ca 3100 m) revealed minimal numbers of geometrid moths. However, effects of isolation and small habitat area might have been responsible for this (see below). Further studies with an extended elevational coverage in neighbouring or other areas could reveal more detailed patterns.

Diversity patterns and environmental factors The exceptional, unique distributions of geometrid moths including their two largest subfamilies Ennominae and Larentiinae, call for explanations. Lawton et al. (1987) listed four hypotheses for declining diversity of herbivorous insects at higher altitudes: 1) reduction of habitat area, 2) reduction of resource diversity, 3) reduction of primary productivity, 4) increasingly unfavourable environments. Despite these constraints, diversity of geometrids remained constant or was even lower at low altitudes. How can this be explained? The points listed by Lawton et al. seem to be either not applicable in the study area (1 and 2), or might be compensated for by other factors (3 and 4).

Habitat area Habitat area is not expected to be a limiting factor in this study, because the highest site is situated ca 500 m below the mountain summit and large areas of similar forested habitat exist next to the study area. However, area might also come into play as a limiting factor at higher altitudes in the Andes. Rahbeck (1997) and KoÂ¨ rner (2000) emphasised the importance of decreasing area at high altitudes, and Holloway (1987) discussed a higher diversity of moths in montane Papua New Guinea compared to Borneo, as a consequence of greater land area situated above 2000 m. The Andes are folded mountains and provide a habitat area at high altitudes that is by far larger than on relatively isolated mountains such as Mount Kinabalu in Borneo, where the high altitude fauna is less species-rich. Furthermore, the biogeographical conditions in the Andes support the isolation of local populations (e.g. during glacial periods), subsequent speciation events, and a later co-existence of species. These fea462

tures might explain a considerable part of the exceptional diversity of geometrid moths in Ecuador.

Resource diversity The possible reduction of resource diversity, including spatial niches, is difficult to assess because of widely lacking information describing which resources are actually used by Neotropical geometrid moths. However, some conclusions can be drawn from available information on host-plant use (Brehm 2002), and from vegetation data covering the study area. The structural complexity of the forest clearly declines along the altitudinal gradient (Paulsch 2002). Upper montane forests in the study area provide a far poorer offer of structural niches, e.g. because of the lower height and the absence of lianas. As a consequence, structural niches do not seem to be a limiting factor in the diversity of geometrids. The level of floristic diversity is more difficult to interpret, but there is evidence that the diversity of potential host-plants of geometrid moths generally decreases along altitudinal gradients in the Neotropical region (e.g. Gentry 1988, Lieberman et al. 1996). In a few cases there are indications of specialism towards certain host plant groups, and these geometrid groups indeed decline towards high altitudes (Brehm 2002). For example, the ennomine tribes Cassymini and Macariini are specialised Fabaceae feeders and are not present at the highest sites in the study area. Only very few Fabaceae species have been found in the study area (Homeier pers. comm.). Lianas also decrease along the altitudinal gradient in the study area, whereas in herbaceous vines this trend is less pronounced (Matezki pers. comm.). The number of tree species ( \5 cm diameter at breast height, 400 m2 plots) in ridge forests in the study area decreases from ca 30 species at 1850 m to ca 20 species at 2450 m (Homeier et al. 2002). In contrast, the diversity of shrubs might be constant or even increase with increasing altitude. This latter resource (e.g. shrub-like Asteraceae) might play an important role for the very species-rich larentiine genus Eupithecia. There are also indications of a constant and high diversity of epiphytic plants (Werner 2002), which is generally known to be very high in Neotropical montane forests (Gentry and Dodson 1987, Nieder et al. 2001). However, most epiphytic vascular plants are monocotyledons and ferns (Rauer and Rudolph 2001). Both these plant groups are hardly exploited by Neotropical Geometridae (Brehm 2002). Overall, the total diversity of hosts that are actually used can be expected to decrease with altitude, but the extent of this decrease remains unknown. Irrespective of these uncertainties, the diversity of geometrids does not appear to be limited by a reduction of potential resource diversity. ECOGRAPHY 26:4 (2003)

Primary productivity Primary productivity usually decreases along altitudinal gradients (Brujinzeel and Veneklaas 1998, Waide et al. 1998, but see Singh et al. 1994). According to Tanner et al. (1998), nutrient limitation is widespread in montane soils and foliar nitrogen decreases with increasing altitude. Significant changes in soil properties have also been documented along the elevational gradient in the study area (Schrumpf et al. 2001). They reported decreasing pH values and nitrogen availability with rising altitude. Given these constraints, diversity of herbivorous geometrid moths remained unaffected.

Climate and physiology Obviously, the moths are able to resist the cool and humid climate at high altitudes. The monthly average temperature decreases linearly by ca 10 K along the gradient, and precipitation more than doubles from ca 2000 mm to ca 5500 mm per annum (Hagedorn 2001, Emck pers. comm.). Geometrid moths appear to be physiologically predisposed towards such cool conditions. Heinrich (1993) and Rydell and Lancaster (2000) reported that many geometrid moth species were able to fly with lower thoracic temperatures than most other Lepidoptera species do. This could be a major energetic advantage and might allow geometrid moths to colonise habitats that are unsuitable for most other insects. However, the knowledge about the flight physiology of the vast majority of species is still unknown. Further investigations could test this hypothesis. We expect that members of the subfamily Larentiinae are particularly cold-adapted because of their dominance at high altitudes and latitudes.

Are high-altitude habitats an enemy-free space? Predation can have impacts on the diversity of herbivorous insects. For example, Williams et al. (2001) pointed out that resources may often be less important than natural enemies in determining the distribution of herbivores. On the one hand, predators might regulate prey populations and prevent the dominance of single species. On the other hand, low predation pressure might allow an unconstrained radiation of herbivorous insects in nearly enemy-free space. Since predation pressure probably strongly declines with increasing altitude (see below), we find little support for the first hypothesis because neither diversity nor dominance values significantly change along the elevational gradient (Brehm 2002). The diversity of insectivorous species of bats, birds and ants in the Andes markedly decreases with altitude. Up to 38 species of insectivoECOGRAPHY 26:4 (2003)

rous bat species co-occur in lowland rainforests in Panama (Kalko 1997), whereas only eight occur above 1800 m and four above 2800 m in the study area in Ecuador (Matt 2001). Mixed species flocks of birds that forage in rainforests are expected to have a large impact on leaf-chewing insects (Braun 2002) and occur more prominently in lowland rather than montane forests (Rahbeck 1997, Thiollay 1999). Ants are the most prevalent invertebrate predators in many tropical forests (Ho¨ lldobler and Wilson 1990). However, they strongly decrease in diversity as altitude increases (Stork and Brendell 1990, Bru¨ hl et al. 1999). At higher altitudes in the Ecuadorian study area (above ca 2000 m), only very few ant species occur (unpubl.). Therefore, habitats are indeed a nearly enemy-free space with regard to this otherwise very important group of potential predators (Novotny et al. 1999, Floren et al. 2002). The knowledge on parasitoids and their role for Neotropical geometrid moths is still extremely scarce. However, parasitoids appear not to be a dominant factor since Brehm (2003) found only a low percentage of parasitised larvae in the study area.

Are the results representative of other groups? This study has demonstrated exceptional altitudinal patterns of one major group of herbivorous insects. Further sampling would be required to confirm whether the results of this study are also applicable for other groups. According to Holloway (1987) the relative contribution of geometrids to local moth assemblages increases with altitude, suggesting that diversity patterns should be discordant even among moths as a guild. Indeed moth taxa in the Ecuadorian study area such as Pyraloidea and Arctiidae exhibit completely different altitudinal diversity patterns (Su¨ ßenbach 2003). Beccaloni and Gaston (1995) found a relatively constant ratio of species of the nymphalid subfamily Ithomiinae among all butterflies, and Longino (1994) reported a number of tropical invertebrate ‘‘focal taxa’’ that might represent suitable ‘‘survey taxa’’. The transfer of results from one group to others is part of the controversial debate about the usefulness of biodiversity indicators. Although several studies have established parallels between diversity patterns of different groups of organisms (Wolda 1996, Kerr et al. 2000), others found that there were none (Lawton et al. 1998, Ricketts et al. 2002). Simberloff (1998) criticised the concept of biodiversity indicators because of a lacking consensus as to what indicators should indicate and which organisms are the most versatile. If various taxa exhibit fundamentally different diversity patterns even among the herbivorous Lepidoptera, there is no reason to assume that patterns of, for example, detrivorous or predatory insects are better reflected. 463

Choice of measurement and sample size dependence Our results confirm that unless it is possible to achieve complete inventories, the recorded species number is an unreliable measure of diversity because of its extreme dependence on the number of specimens collected (correlation coefficients all \0.9, p B0.001). As expected, it has to be rejected as a meaningful measure of diversity for the purpose of discriminating between samples that are incompletely sampled (Gotelli and Colwell 2001). The estimator Chao 1 has also been shown to be significantly sample size dependent, though not to the same extent as species number. It is very probable that the true local richness is still substantially underestimated at most sites. This is illustrated by the very high ratios of singletons at single sites, i.e. species that were collected only once (41 –60% of the species). A very high ratio of rare species is typical for samples of tropical arthropods. For example, Novotny and Basset (2000) found very similar singleton rates of 45% in samples of herbivorous insects in New Guinea. Underestimation occurs if samples are too sparse (Colwell and Coddington 1994). This study shows that even samples of at least 134 species and 384 individuals can be ‘‘too sparse’’ for extrapolation in extremely rich moth ensembles. According to Colwell and Coddington (1994), estimators correlate with sample size until about half the total fauna is observed and thereafter become gradually independent of sample size. Obviously, this level has not been reached at many sites because they could not be sampled more than two to four times. Rarefied species numbers were independent of sample size. The measure can overestimate diversity if species have clumped distributions (Achtziger et al. 1992), but this is of relatively little importance in large samples and does not affect the results presented in this study. Fisher’s alpha values did not correlate with specimen numbers and were thus independent of sample size. However, it is possible that values of Fisher’s alpha would further increase with an increasing number of samples (Wolda 1987, Intachat and Holloway 2000, Schulze and Fiedler 2003). Since goodness-of-fit of the log-series model was not always satisfactory, it does not seem to be appropriate to rely solely on this measure. The use of several different measures can be recommended since they complement each other in different aspects of diversity as well as in the mathematical assumptions underlying their usage. Acknowledgements – We are indebted to the taxonomists who were kind enough to allow access to collections under their care, and provided advice and literature: Linda M. Pitkin, Malcolm J. Scoble, and David J. Carter at the Natural History Museum, London, and Axel Hausmann, Manfred Sommerer and Robert Trusch at the Zoologische Staatssammlung, Munich. Paul Emck, Ju¨ rgen Homeier and Steffen Matezki provided unpublished results of their investigations on climate and vegetation of the study area. Jennifer Kay, Adrienne Hogg, Rita Schneider and Teresa Baethmann helped to pre-

464

pare and database the moths. Giovanni Onore and Christoph L. Ha¨ user offered administrative support. The Ministerio del Medio Ambiente del Ecuador granted research permits, NCI (Loja, Ecuador) allowed access to parts of the study area, and the Deutsche Forschungsgemeinschaft financed the project (Fi 547/5-1/3, FOR 402/1-1).

References Achtziger, R., Nigmann, U. and Zwo¨ lfer, H. 1992. Rarefaction-Methoden und ihre Einsatzmo¨ glichkeiten bei der zooo¨ kologischen Zustandsanalyse und Bewertung von Biotopen. – Z. O8 kol. Natursch. 1: 89 – 105. Anon. 1999. Statistica 5.5. for Windows. – StatSoft, Tulsa. Barlow, H. S. and Woiwod, I. P. 1989. Moth diversity of a tropical forest in Peninsular Malaysia. – J. Trop. Ecol. 5: 37 – 50. Basset, Y. 2001. Invertebrates in the canopy of tropical rain forests – how much do we know? – Plant Ecol. 153: 87 – 107. Beccaloni, G. W. and Gaston, K. J. 1995. Predicting the species richness of Neotropical forest butterflies: Ithomiinae (Lepidoptera: Nymphalidae) as indicators. – Biol. Conserv. 71: 77 –86. Beck, J. et al. 2002. From forest to farmland: diversity of geometrid moths along two habitat gradients on Borneo. – J. Trop. Ecol. 17: 33 – 51. Braun, H. 2002. Die Laubheuschrecken (Orthoptera, Tettigoniidae) eines Bergregenwaldes in Su¨ d-Ecuador – faunistische, bioakustische und o¨ kologische Untersuchungen. – Ph.D. thesis, Univ. of Erlangen-Nu¨ rnberg. Brehm, G. 2002. Diversity of geometrid moths in a montane rainforest in Ecuador. – Ph.D. thesis, Univ. of Bayreuth. Brehm, G. 2003. Host-plant records and illustrations of the larvae of 19 geometrid moths from a montane rainforest in Ecuador. – Nachr. Entomol. Ver. Apollo. 24, in press. Brehm, G. and Fiedler, K. 2003. Faunal composition of geometrid moths changes with altitude in an Andean montane rainforest. – J. Biogeogr. 30: 431 – 440. Brehm, G., Homeier, J. and Fiedler, K. 2003. Beta diversity of geometrid moths (Lepidoptera: Geometridae) in an Andean montane rainforest. – Divers. Distrib. 9, in press. Bru¨ hl, C. A., Mohamed, M. and Linsenmair, K. E. 1999. Altitudinal distribution of leaf litter ants along a transect in primary forest on Mount Kinabalu, Sabah, Malaysia. – J. Trop. Ecol. 15: 265 – 277. Brujinzeel, L. A. and Veneklaas, E. J. 1998. Climatic conditions and tropical montane forest productivity: the fog has not lifted. – Ecology 79: 3 – 9. Bussmann, R. W. 2001. The montane forests of Reserva Biolo´ gica San Francisco (Zamora-Chinchipe, Ecuador). – Erde 132: 9 – 25. Chey, V. K. 2000. Moth diversity in the tropical rain forest of Lanjak-Entimau, Sarawak, Malaysia. – Malayan Nat. J. 54: 305 – 318. Colwell, R. K. 2000. EstimateS: statistical estimation of species richness and shared species from samples, ver. 5.1. – Bhttp://viceroy.eeb.uconn.edu/estimates\. Colwell, R. K. and Coddington, J. A. 1994. Estimating terrestrial biodiversity through extrapolation. – Philos. Trans. R. Soc. Lond. B 345: 101 – 118. Floren, A., Biun, A. and Linsenmair, K. E. 2002. Arboreal ants as key predators in tropical lowland rainforest trees. – Oecologia 131: 137 – 144. Gentry, A. H. 1988. Changes in plant community diversity and floristic composition on environmental and geographical gradients. – Ann. Mo. Bot. Gard. 75: 1 – 34. Gentry, A. H. and Dodson, C. 1987. Contribution of nontrees to species richness of a tropical rainforest. – Biotropica 19: 149 – 156. ECOGRAPHY 26:4 (2003)

Gotelli, N. J. and Colwell, R. K. 2001. Quantifying biodiversity: procedures and pitfalls in the measurement and comparison of species richness. – Ecol. Lett. 4: 379 – 391. Grubb, P. J. et al. 1963. A comparison of montane and lowland rain forest in Ecuador I. The forest structure, physiognomy, and floristics. – J. Ecol. 51: 567 – 601. Grytnes, J. A. and Vetaas, O. R. 2002. Species richness and altitude: a comparison between null models and interpolated plant species richness along the Himalayan altitudinal gradient, Nepal. – Am. Nat. 159: 294 – 304. Hagedorn, A. 2001. Extent and significance of soil erosion in southern Ecuador. – Erde 132: 75 –92. Hanski, I. and Niemela¨ , J. 1990. Elevational distributions of dung and carrion beetles in northern Sulawesi. – In: Knight, W. J. and Holloway, J. D. (eds), Insects and the rain-forests of south east Asia (Wallacea). The Roy. Entomol. Soc., pp. 145 –152. Hayek, L.-A. and Buzas, M. A. 1997. Surveying natural populations. – Columbia Univ. Press. Heinrich, B. 1993. The hot-blooded insects. – Springer. Henderson, A., Churchill, S. P. and Luteyn, J. L. 1991. Neotropical plant diversity. – Nature 351: 21 –22. Henderson, P. A. and Seaby, R. M. H. 1998. Species diversity and richness, ver. 2.65. – Pisces Conserv., Pennington. Hochberg, Y. 1988. A sharper Bonferroni procedure for multiple tests of significance. – Biometrika 75: 800 – 802. Ho¨ lldobler and Wilson, 1990. The ants. – Springer. Holloway, J. D. 1987. Macrolepidoptera diversity in the IndoAustralian tropics: geographic, biotopic and taxonomic variations. – Biol. J. Linn. Soc. 30: 325 –341. Holloway, J. D. and Nielsen, E. S. 1999. Biogeography of the Lepidoptera. – In: Kristensen, N. P. (ed.), Lepidoptera, moths and butterflies, Vol. 1, part 35. De Gruyter. Holloway, J. D., Robinson, G. S. and Tuck, K. R. 1990. Zonation of the Lepidoptera of northern Sulawesi. – In: Knight, W. J. and Holloway, J. D. (eds), Insects and the rain-forests of south east Asia (Wallacea). The Roy. Entomol. Soc., pp. 153 –166. Homeier, J., Dalitz, H. and Breckle, S.-W. 2002. Waldstruktur und Baumartendiversita¨ t im montanen Regenwald der Estacio´ n Cientı´fica San Francisco in Su¨ decuador. – Ber. Reinh.-Tu¨ xen-Ges. 14: 109 –118. Hurlbert, S. H. 1971. The nonconcept of species diversity: a critique and alternative parameters. – Ecology 52: 577 – 586. Intachat, J. and Woiwod, I. P. 1999. Trap design for monitoring moth biodiversity in tropical rainforests. – Bull. Entomol. Res. 89: 153 –163. Intachat, J. and Holloway, J. D. 2000. Is there stratification in diversity or preferred flight height of geometroid moths in Malaysian lowland tropical forest? – Biodiv. Conserv. 9: 1417 – 1439. Janzen, D. H. et al. 1976. Changes in arthropod community along an elevational transect in the Venezuelan Andes. – Biotropica 8: 193 –203. Kalko, E. V. K. 1997. Diversity in tropical bats. – In: Ulrich, H. (ed.), Tropical biodiversity and systematics. Proc. of the Int. Symp. on Biodiv. and Syst. in tropical ecosystems. Zool. Forschungsinst. und Mus. Alexander Koenig Bonn, pp. 13– 43. Kempton, R. A. and Taylor, L. R. 1974. Log-series and log-normal parameters as diversity discriminants for the Lepidoptera. – J. Anim. Ecol. 43: 381 –399. Kenney, A. J. and Krebs, C. J. 2000. Programs for ecological methodology, ver. 5.2. – \ http://www.zoology.ubc.ca/ krebs \ . Kerr, J. T., Sugar, A. and Packer, L. 2000. Indicator taxa, rapid biodiversity assessment, and nestedness in an endangered ecosystem. – Conserv. Biol. 14: 1726 –1734. Ko¨ rner, C. 2000. Why are there global gradients in species richness? Mountains might hold the answer. – Trends Ecol. Evol. 15: 513 –514. ECOGRAPHY 26:4 (2003)

Lawton, J. H., MacGarvin, M. and Heads, P. A. 1987. Effects of altitude on the abundance and species richness of insect herbivores on bracken. – J. Anim. Ecol. 56: 147 – 160. Lawton, J. H. et al. 1998. Biodiversity inventories, indicator taxa and effects of habitat modification in tropical forest. – Nature 391: 72 – 75. Lieberman, D. et al. 1996. Tropical forest structure and composition on a large-scale altitudinal gradient in Costa Rica. – J. Ecol. 84: 137 – 152. Longino, J. T. 1994. How to measure arthropod diversity in a tropical rainforest. – Biol. Internat. 28: 3 – 13. Matt, F. 2001. Pflanzenbesuchende Flederma¨ use im tropischen Bergregenwald: Diversita¨ t, Einnischung und Gildenstruktur – Eine Untersuchung der Fledermausgemeinschaften in drei Ho¨ henstufen der Andenostabdachung des Podocarpus Nationalparks in Su¨ decuador. – Ph.D. thesis, Univ. of Erlangen-Nu¨ rnberg. McCoy, E. D. 1990. The distribution of insects along elevational gradients. – Oikos 58: 313 – 322. Myers, N. et al. 2000. Biodiversity hotspots for conservation priorities. – Nature 403: 853 – 858. Nieder, J., Prosperı´, J. and Michaloud, G. 2001. Epiphytes and their contribution to canopy diversity. – Plant Ecol. 153: 51 – 63. Novotny, V. and Basset, Y. 2000. Rare species in communities of tropical insect herbivores: pondering the mystery of singletons. – Oikos 89: 564 – 572. Novotny, V. et al. 1999. Predation risk for herbivorous insects on tropical vegetation: a search for enemy-free space and time. – Aust. J. Ecol. 24: 477 – 483. Olson, D. M. 1994. The distribution of leaf litter invertebrates along a Neotropical altitudinal gradient. – J. Trop. Ecol. 10: 129 – 150. Paulsch, A. 2002. Development and application of a classification system for undisturbed and disturbed tropical montane forests based on vegetation structure. – Ph.D. thesis, Univ. of Bayreuth. Price, P. W. et al. 1995. The abundance of insect herbivore species in the tropics: the high local richness of rare species. – Biotropica 27: 468 – 478. Rahbeck, C. 1995. The elevational gradient of species richness: a uniform pattern? – Ecography 18: 200 – 205. Rahbeck, C. 1997. The relationship among area, elevation, and regional species richness in Neotropical birds. – Am. Nat. 149: 875 – 902. Rauer, G. and Rudolph, D. 2001. Vaskula¨ re Epiphyten eines westandinen Bergregenwaldes in Ecuador. – In: Nieder, J. and Barthlott, W. (eds), The flora of the Rio Guajalito mountain rain forest. Books on Demand, pp. 323 – 469. Ricketts, T. H., Daily, G. C. and Ehrlich, P. R. 2002. Does butterfly diversity predict moth diversity? Testing a popular indicator taxon at local scales. – Biol. Conserv. 103: 361 – 370. Rydell, J. and Lancaster, W. C. 2000. Flight and thermoregulation in moths were shaped by predation from bats. – Oikos 88: 13 – 18. Schrumpf, M. et al. 2001. Tropical montane rain forest soils. – Erde 132: 43 – 59. Schulze, C. H. 2000. Auswirkungen anthropogener Sto¨ rungen auf die Diversita¨ t von Herbivoren – Analyse von Nachtfalterzo¨ nosen entlang von Habitatgradienten in Ost-Malaysia. – Ph.D. thesis, Univ. of Bayreuth. Schulze, C. H. and Fiedler, K. 2003. Vertical and temporal diversity of a species-rich moth taxon in Borneo. – In: Basset, Y. et al. (eds), Arthropods of tropical forests – spatio-temporal dynamics and resource use in the canopy. Cambridge Univ. Press, pp. 69 – 85. Scoble, M. J. (ed.) 1999. Geometrid moths of the world – a catalogue (Lepidoptera: Geometridae). – CSIRO Publ., Collingwood. Simberloff, D. 1998. Flagships, umbrellas, and keystones: is a single-species management passe´ in the landscape era? – Biol. Conserv. 83: 247 – 257.

465

Singh, S. P., Adhikari, B. S. and Zobel, D. B. 1994. Biomass, productivity, leaf longevity, and forest structure in the central Himalaya. – Ecol. Monogr. 64: 401 –421. Southwood, T. R. and Henderson, P. A. 2000. Ecological methods. – Blackwell. Stork, N. E. and Brendell, M. J. D. 1990. Variation in the insect fauna of Sulawesi trees with season, altitude, and forest type. – In: Knight, W. J. and Holloway, J. D. (eds), Insects and the rain-forests of south east Asia (Wallacea). The Roy. Entomol. Soc., pp. 173 – 190 Su¨ ßenbach, D. 2003. Diversita¨ t von Nachtfaltergemeinschaften entlang eines Ho¨ hengradienten in Su¨ decuador (Lepidoptera: Pyraloidea, Arctiidae). – Ph.D. thesis, Univ. of Bayreuth. Tanner, E. V. J., Vitousek, P. M. and Cuevas, E. 1998. Experimental investigation of nutrient limitations of forest growth on wet tropical mountains. – Ecology 79: 10 – 22. Thiollay, J. M. 1999. Frequency of mixed species flocking in tropical forest birds and correlates of predation risk: an intertropical comparison. – J. Avian Biol. 30: 282 – 294. Thomas, A. W. 2002. Moth diversity in a northeastern North America, red spruce forest, II. The effect of silvicultural practices on geometrid diversity (Lepidoptera: Geometridae). – Can. For. Serv., Fredericton. Thomas, A. W. and Thomas, G. M. 1994. Sampling strategies

466

for estimating moth species diversity using a light trap in a northeastern softwood forest. – J. Lepidop. Soc. 48: 85 – 10. von Humboldt, A. and Bonpland, A. 1807. Essai sur la geographie des plantes – accompagne´ d’un tableau physique des re´ gions e´ quinoxiales, fonde´ sur des mesures exe´ cute´ es, depuis le dixie`me degre´ de latitude bore´ ale jusqu’au dixie`me degre´ de latitude australe, pendant les anne´ es 1799, 1800, 1801, 1802 et 1803. – Levrault et Schoell, Paris. Waide, R. B., Zimmerman, J. K. and Scatena, F. N. 1998. Controls of primary productivity: lessons from the Luquilo mountains in Puerto Rico. – Ecology 79: 31 – 37. Werner, F. 2002. Ecology of vascular epiphytes in a montane forest and on remnant trees of adjacent pastures. – Diploma thesis, Univ. of Bonn. Williams, I. S., Jones, H. and Hartley, S. 2001. The role of resources and natural enemies in determining the distribution of an insect herbivore population. – Ecol. Entomol. 26: 204 – 211. Wolda, H. 1987. Altitude, habitat and tropical insect diversity. – Biol. J. Linn. Soc. 30: 313 – 323. Wolda, H. 1996. Between-site similarity in species composition of a number of Panamanian insect groups. – Miscella`nia Zool. 19: 39 – 50.

ECOGRAPHY 26:4 (2003)

Ecography 33: 425 434, 2010 doi: 10.1111/j.1600-0587.2009.06016.x # 2010 The Authors. Journal compilation # 2010 Ecography Subject Editor: Jens-Christian Svenning. Accepted 5 June 2009

Change within and among forest communities: the influence of historic disturbance, environmental gradients, and community attributes Windy A. Bunn, Michael A. Jenkins, Claire B. Brown and Nathan J. Sanders W. A. Bunn (Windy_Bunn@nps.gov), C. B. Brown and N. J. Sanders, Dept of Ecology and Evolutionary Biology, Univ. of Tennessee, 569 Dabney Hall, Knoxville, TN 37996, USA. (Present address of W. A. B.: USDOI, National Park Service, Grand Canyon National Park, P. O. Box 129, Grand Canyon, AZ 86023, USA.) M. A. Jenkins, USDOI, National Park Service, Great Smoky Mountains National Park, 1314 Cherokee Orchard Road, Gatlinburg, TN 37738, USA. (Present address of M. A. J.: Purdue Univ., Dept of Forestry and Natural Resources, 715 West State Street, West Lafayette, IN 47907, USA.)

Understanding how ecological communities change over time is critical for biodiversity conservation, but few long-term studies directly address decadal-scale changes in both the within- and among-community components of diversity. In this study, we use a network of permanent forest vegetation plots, established in Great Smoky Mountains National Park (USA) in 1978, to examine the factors that influence change in community composition within and among communities. In 2007, we resampled 15 plots that were logged in the late 1920s and 15 plots that had no documented history of intensive human disturbance. We found that understory species richness decreased by an average of 4.3 species over the 30-yr study period in the logged plots, but remained relatively unchanged in the unlogged plots. In addition, tree density decreased by an average of 145 stems ha 1 in the logged plots, but was relatively stable in the unlogged plots. However, we found that historic logging had no effect on within-community understory or tree compositional turnover during this time period. Instead, sites at lower elevations and sites with lower understory biomass in 1978 had higher understory compositional turnover than did sites at higher elevations and sites with higher understory biomass. In addition, sites with lower soil cation exchange capacity (CEC) and with lower tree basal area in 1978 had higher tree compositional turnover than did sites with higher soil CEC and higher tree basal area. Among-community similarity was unchanged from 1978 to 2007 for both the logged and unlogged plots. Overall, our results indicate that human disturbance can affect plant communities for decades, but the extent of temporal change in community composition may nevertheless depend more on environmental gradients and community attributes.

Both within- and among-community attributes can change over time, and understanding these changes often requires long-term empirical data (Magnuson 1990, Wardle et al. 2008). Within a community, the total number of species present as well as the abundance of particular species can change with time. These within-community changes can lead to temporal differences in similarity among communities within a region (Loreau 2000). Despite the potential for within-community changes to influence similarity among communities, few studies directly measure longterm changes in both the within- and among-community components of diversity (but see Chalcraft et al. 2004). In this study, we use long-term monitoring data from Great Smoky Mountains National Park (GSMNP) in eastern Tennessee to examine changes in forest understory plant communities and tree communities across 30 yr. Specifically, we examine the factors that influence compositional change within communities and change in compositional similarity among communities.

Human-caused disturbance can lead to large and persistent differences in understory communities in disturbed forests compared with undisturbed forests (Meier et al. 1995, Flinn and Vellend 2005, Harrelson and Matlack 2006). Furthermore, disturbance may also influence the magnitude of compositional change through time in disturbed versus undisturbed communities (Collins and Smith 2006). While general models of forest development (Oliver and Larson 1996) describe forest communities as undergoing rapid changes in the short term after disturbance, the long-term influence of disturbance on withincommunity change is less clear. In addition to disturbance, environmental gradients can influence variation in within-community compositional turnover. For example, the extent of compositional turnover through time in low-elevation sites may be higher than that of high elevation sites (Aplet and Vitousek 1994, Selmants and Knight 2003, Taverna et al. 2005), and a number of factors that are often associated with 425

elevation precipitation, soil fertility, species richness, and primary productivity-have been shown or hypothesized to affect temporal change within communities (Peet and Christensen 1988, Chase and Leibold 2002, Verheyen et al. 2003, Yurkonis and Meiners 2004, Taverna et al. 2005, Smart et al. 2006, White et al. 2006, Anderson 2008). However, the relative influence of these factors on compositional turnover within plant communities is poorly understood and likely varies with community composition and structure. Over time, communities within a region can become either more similar or less similar to one another depending on the extent of change within individual communities of the region. Recent studies of decadal-scale change in forest communities in the eastern U.S. indicate that declines in species richness (Rooney et al. 2004, Taverna et al. 2005), shifts in plant community composition (Taverna et al. 2005), and changes in regional community similarity (Rooney et al. 2004) over time may be common. Landuse history can have large effects on forest communities (Foster et al. 1998, Vellend et al. 2007), and the legacy of human disturbance may therefore be important for understanding patterns in among-community similarity in these long-lived communities. In this study, we investigated changes in understory plant communities as well as tree communities in forests of GSMNP that were logged in the late 1920s and forests that were not logged. Specifically, we examined plant community data collected in 1978 (50 yr after logging) and in 2007 (80 yr after logging) to test four explicit hypotheses: 1) community composition of logged plots differs from that of unlogged plots both 50 and 80 yr after the logging event. 2) Historically logged plots have greater within-community compositional turnover than unlogged plots. 3) Elevation and associated edaphic and community attributes influence within-community compositional turnover. 4) The temporal change in similarity of logged communities to one another differs from the temporal change in similarity of unlogged communities to one another.

Methods Great Smoky Mountains National Park (GSMNP) is a 211 000 ha protected area that straddles the Tennessee North Carolina stateline. Elevations in GSMNP range from 271 to 2025 m, and climate and vegetation types vary

considerably along the elevational gradient. Mean annual rainfall in low elevation sites is 1400 mm with mean temperatures 128C, while annual rainfall is 2000 mm and temperature averages 68C at high-elevation peaks. GSMNP contains over 70 vegetation associations, varying from low- to mid-elevation mixed hardwood forests and xeric Pinus and Quercus forests to high-elevation Picea Abies forests and heath balds. Prior to its establishment, ca 80% of the area that became GSMNP was subject to anthropogenic disturbance (Pyle 1988). Despite its history of disturbance, GSMNP is considered a center for diversity in North America. Plot selection and field methods In 2007, we resampled thirty 20 50 m forest plots originally established in 1978. The plots were randomly distributed using a stratified design that divided watersheds into units based upon elevation, slope position, and aspect. Fifteen of the resampled plots were in historically logged forests and fifteen plots were in unlogged forests. Historically logged forests were defined using the ‘‘corporate logging’’ category of Pyle (1988) and included only those areas in which the use of railroads, mechanized skidding, non-selective cutting practices, and highly extensive cutting on slopes occured. We defined unlogged forests using Pyle’s ‘‘high in virgin forest attributes’’ and ‘‘big trees with diffuse disturbance’’ categories. We chose the thirty resampled plots from a pool of over 100 permanent plots. Since our primary goal was to evaluate the effects of historic disturbance and elevation on community dynamics, we used 1978 field data and 2007 pre-sampling surveys to exclude plots with high levels of recent disturbance. Toward this end, we resampled only plots dominated by hardwood species, sites without recent or frequent fires, and sites that were not (or have not been) influenced by Dendroctonus frontalis (southern pine beetle), Adelges piceae (balsam woolly adelgid) or Adelges tsugae (hemlock woolly adelgid). Because of this rigorous selection process, we consider the thirty resampled plots to be relatively free of disturbance in the thirty years between sampling events. The logged and unlogged plots were topographically similar to one another: mean elevation and percent slope of logged plots did not differ from the mean elevation and percent slope of unlogged plots (Table 1). The 15 logged plots ranged in elevation from 727 to 1402 m and

Table 1. Comparison of topographic and edaphic variables in historically logged and unlogged plots, using t-tests or Wilcoxon rank-sum tests to test for mean differences. Variable

Elevation (m) Slope (%) Soil pH Soil cation exchange capacity (meq 100 g 1) Soil K (ppm) Soil Ca (ppm) Soil Mg (ppm) Soil P (ppm) Soil organic matter (%)

426

Mean9SE

Logged vs unlogged

logged

unlogged

p-value

1104961.6 19.093.26 4.290.11 9.490.40 68.599.04 326971.9 44.996.63 14.391.70 5.190.28

1034967.3 19.892.42 4.590.12 7.190.16 64.894.93 165927.1 32.393.45 15.492.46 3.190.33

0.41 0.95 0.02 B0.0001 0.63 0.12 0.13 0.77 0.0002

occurred on north- (n 8), east- (n 4), and west- (n 3) facing slopes. The 15 unlogged plots ranged in elevation from 664 to 1400 m and occurred on north- (n 11), east(n 1), and west- (n 3) facing slopes. Both the logged and unlogged plots are characterized by well-drained loamy soils classified as either humic or typic dystrudepts (A. R. Khiel, NRCS, unpubl. report). Tree cores collected and analyzed by the GSMNP Vegetation Monitoring Program were available for a subset of the plots (Jenkins unpubl.). Dominant trees in unlogged plots were 150 228 yr old (complete cores) or a minimum 132 147 yr old (cores without pith) in 2007. As expected from logging history records, dominant trees in logged plots were 75 80 yr old in 2007. In 1978, the 20 50 m forest plots were permanently marked with rebar and witness tree tags, which allowed us to reestablish the plots in 2007. In the 2007, we used the same sampling design used in 1978. Within each 20 50 m plot, we recorded understory shrub and tree seedling species B1 m tall in 25 4-m2 subplots and understory herbaceous species in 25 1-m2 subplots nested within the shrub and seedling subplots. We defined the understory community as the shrub, seedling, and herbaceous species recorded in all 25 subplots within the 20 50 m plot. We also recorded all individual trees (]10 cm dbh) by species in the 20 50 m plot and used these data to characterize the tree community. In 2007, we sampled plots at roughly the same time of year that they were sampled in 1978. Sampling was conducted between 19 June and 26 August in 1978 and between 9 July and 26 August in 2007. Sampling of plots was paired, as best as possible, within seasons. That is, if a plot was sampled late in the field season in 1978, we attempted to sample that plot late in the field season of 2007. To characterize the sampling plots, we estimated a suite of topographic and edaphic parameters (Table 1). We estimated elevation using topographic maps and calculated percent slope by averaging three slope measurements taken at the two 20-m end lines and at the center of each plot facing downslope. Between 2002 and 2007, soil samples were collected from the top 10 cm of soil at five locations throughout each of the 30 plots with a hand spade. The five subsamples were combined into one composite sample per plot, dried at 438C for at least 8 h, and sieved through a 2 mm mesh. The samples were analyzed for pH, cation exchange capacity (CEC), total phosphorus (P), potassium (K), calcium (Ca), and magnesium (Mg), and percent organic matter by A&L Analytical Laboratories, Memphis, TN (see Jenkins et al. 2007 for details of cation extraction procedures). Analysis of the influence of historic logging on community composition We used nonparametric, permutational multivariate analysis of variance (PERMANOVA; Anderson 2001) based on Bray-Curtis similarity values of species abundances to compare understory and tree community composition between logged and unlogged plots in both 1978 and 2007. PERMANOVA compares the variability in species similarity between plots within a treatment to the

variability in species similarity between plots from different treatments and is performed using the FORTRAN program PERMANOVA (Anderson 2005). The test statistic for PERMANOVA is the pseudo F-ratio, where a large pseudo F-ratio indicates that logged plots are closer to one another in multivariate space than they are to unlogged plots and that the logged and unlogged plots differ in community composition. The significance of the pseudo F-ratio is tested using a permutation test that randomly shuffles the sample labels within and among treatment groups and calculates the pseudo F-ratio for 9999 arbitrary reassignments of the data. The pseudo F-ratios of these randomly assigned communities are then compared to the pseudo F-ratio of the observed community to calculate the significance level of the test (Anderson 2001). For the understory communities, we calculated species abundance as the percentage of the 25 subplots in which the species occurred. For the tree communities, we used the number of individual trees of a species as the abundance value. We chose to use understory frequency and tree density as the abundance measures rather than understory cover and tree basal area because these measures are more repeatable between observers and across years. The choice to use these frequency-based abundance measures did not change the results. Since the scale of abundance values in our study was small (ranging from 0 to 100 for understory species and 0 to 52 for tree species), we did not transform the data to reduce the influence of abundant species. A significant pseudo F-ratio from the PERMANOVA can indicate a difference in community composition between treatments due either to differences in the location of the treatment communities in multivariate space or to differences in dispersion of communities in multivariate space within the treatments (Anderson 2001). To confirm that compositional differences between logged and unlogged communities were due to location differences rather than to dispersion differences, we used permutational analysis of multivariate dispersions (PERMDISP; Anderson et al. 2006) performed in the FORTRAN program PERMDISP (Anderson 2004). PERMDISP calculates the centroid of each treatment (logged or unlogged) in multivariate space based on the chosen similarity measure (in this case, Bray-Curtis), and then calculates the distance of each plot within the treatment from the treatment centroid. To compare average dispersion values between treatments (logged understory communities vs unlogged understory communities; logged tree communities vs unlogged tree communities), PERMDISP performs a permutational ANOVA and calculates a pseudo F-statistic and p-value in the same manner as the PERMANOVA described above. A significant pseudo F-ratio from the PERMANOVA and a non-significant difference in dispersion between logged and unlogged plots from the PERMDISP analysis would suggest that logged and unlogged communities differ in multivariate composition and do not differ in variation around the mean composition within logged and unlogged communities. We tested whether particular species accounted for the observed differences in community composition between logged and unlogged communities with indicator species analysis (DufreË&#x2020;ne and Legendre 1997) using PC-ORD 5.0 427

(MjM Software Design, Gleneden Beach, OR). The indicator analysis uses the relative abundance of each species (for example, percent cover of the species in the logged plots divided by percent cover of the species in all plots) and the relative frequency of the species within each group (for example, the number of logged plots in which the species occurs out of the 15 total logged plots) to calculate an indicator value that ranges from 0 to 100. An indicator value of 100 indicates that the species was observed in only one group (in this case, logged or unlogged plots) and that each plot within that group contained at least one individual of that species. In other words, a species with an indicator value of 100 for logged plots occurs in every logged plot and no unlogged plots, and is thus, a good indicator of plots that have been logged. A Monte Carlo test based on Bray-Curtis distance was used to test the significance of the indicator value (Dufreˆne and Legendre 1997). Analysis of change within communities We analyzed within-plot change in species richness and tree density with paired t-tests, where the species richness and density values for a plot in 1978 were compared with the species richness and density values for the same plot in 2007. For the analyses of tree species richness, we used both the observed number of species present in a plot and an estimate of species richness generated by individual-based rarefaction (PRIMER, ver. 6, PRIMER-E, Plymouth Marine Laboratory, Plymouth, UK). Rarefaction allowed us to correct species richness values for differences in the number of individuals sampled in each plot by using a resampling procedure to generate estimated species richness values based on the number of individuals sampled in the plot with the fewest trees (Gotelli and Colwell 2001). To estimate compositional turnover within communities (i.e. change within a plot over the 30-yr period), we calculated the similarity of each plot in 1978 to itself in 2007 using the Chao Sørensen incidence-based index and the Bray Curtis index in EstimateS (Colwell 2005). The Chao Sørensen incidence-based index (Linc) is a modified form of the traditional Sørensen similarity index that accounts for both the frequency of individual species in the community and for the effects of ‘‘unseen shared species’’ (species that are missing from the sample data but are likely present in the community) on community similarity (Chao et al. 2005). The Chao Sørensen index is useful for assessing similarity between diverse communities that contain many rare species, such as the forest understory plant community. The Bray Curtis index (also referred to as the Sørensen quantitative index or the Czekanowski coefficient; CN) is widely used to assess similarity between two communities (Magurran 2004). The Chao Sørensen and Bray Curtis indices produced qualitatively similar results for the understory community, so we chose to focus on only the Chao Sørensen index in the results and discussion. Since we did not expect ‘‘unseen shared species’’ in the tree community, we used the Bray Curtis index to assess tree compositional turnover. For both the Bray Curtis and the Chao Sørensen indices, values near 1 indicate nearly identical community composition between time periods and values near 0 428

indicate that communities have very little compositional overlap between time periods. We defined turnover as the degree of compositional dissimilarity between 1978 and 2007 within an individual plot. Therefore, we calculated turnover as 1-Linc for the understory community and 1-CN for the tree community. To evaluate whether historic logging influenced compositional turnover, we performed t-tests to compare mean logged and unlogged community turnover values. We then used mixed stepwise multiple regression (a combination of forward and backward steps; a 0.1) to determine whether elevation, edaphic, or community attributes influenced variation in compositional turnover. We used JMP 6.0 (SAS Inst., Cary, NC, USA) for all analyses of within-plot compositional turnover. Analysis of change among communities To examine whether similarity among communities (i.e. how similar plots were to one another within a sampling period) changed over time in logged and unlogged plots, we used a test for homogeneity of multivariate dispersions (Anderson et al. 2006) based on Bray Curtis dissimilarity. Among-community similarity is the average distance among plots within a group to the group centroid in multivariate space (i.e. multivariate dispersion as in Anderson et al. 2006) and is statistically tested for differences in amongcommunity similarity between years with a permutational ANOVA (described above) in the PERMDISP program. Using this approach, a significant p-value indicates that plots within a treatment (logged or unlogged) became either more homogeneous (had lower multivariate dispersion in 2007 than in 1978) or more dissimilar to one another (had higher multivariate dispersion in 2007 than in 1978) over time. For our study plots, average distance of individual plots to the group centroid is directly comparable to traditional measures that calculate mean similarity of each plot to all other plots within the group. For understory communities, average Bray Curtis similarity was highly correlated with average distance to the group centroid in 1978 (r 0.99, p B0.001) and in 2007 (r 0.99, pB0.0001). For tree communities, average Bray Curtis similarity was highly correlated with average distance to the group centroid in 1978 (r 0.99, pB0.001) and in 2007 (r 0.87, p B0.0001).

Results In 1978 (50 yr after logging), historically logged plots contained a total of 132 understory species and 29 tree species while unlogged plots contained 157 understory species and 25 tree species. In 2007 (80 yr after logging), historically logged plots contained a total of 110 understory species and 24 tree species while unlogged plots contained 134 understory species and 26 tree species. Across both sampling periods, historically logged plots contained 25 understory species and 5 tree species that were not found in unlogged plots, and unlogged plots contained 48 understory species and 2 tree species that were unique to unlogged plots (Supplementary material Table S1). Overall, 39 species recorded in 1978 were not

seen in 2007, and 9 new species were encountered in 2007 that were not recorded in 1978 (Supplementary material Table S2).

F1, 28 2.11, p 0.03) and 2007 (PERMANOVA: F1, 28 2.37, p 0.02). These differences were due to differences in the location of the logged and unlogged plots in multivariate space rather than to differences in the relative dispersion of plots within the logged and unlogged groups (1978 PERMDISP: F 0.02, p 0.89; 2007 PERMDISP: p 0.36, P 0.59). Betula lenta and Prunus serotina had significantly higher indicator values in logged plots than in unlogged plots in both years. In addition, Prunus pensylvanica had a significantly higher indicator value in logged plots than in unlogged plots in 1978 and Magnolia fraseri had a significantly higher indicator value in logged plots than in unlogged plots in 2007. Acer saccharum Marsh. was the only tree species with a significantly higher indicator value in unlogged plots than in logged plots and was an indicator of unlogged plots in only 2007 (Table 2).

Influence of historic logging on community composition We found slight differences in understory community composition between logged and unlogged plots in both 1978 (PERMANOVA: F1, 28 1.84, p 0.05) and 2007 (PERMANOVA: F1, 28 1.80, p 0.05). These differences were due to differences in the location of the logged and unlogged plots in multivariate space rather than to differences in the relative dispersion of plots within the logged and unlogged groups (1978 PERMDISP: F 0.54, p 0.52; 2007 PERMDISP: F 0.40, p 0.59). Seven understory species had significantly higher indicator values (a combination of relative abundance and relative frequency) in unlogged plots than in logged plots in 1978 and eight understory species were significant indicators of unlogged plots in 2007 (Table 2). Five understory species had significantly higher indicator values in logged plots in 1978, but only one understory species was still an indicator of logged plots in 2007 (Table 2). Indicators of unlogged plots included tree seedlings, small shrubs, and slowdispersing forest interior herbs, such as Trillium spp., Viola hastata, Arisaema triphyllum, and Eurybia divaricata. Four of the five significant indicators of logged plots were woody seedlings or shrubs. Composition of tree communities differed between the logged and unlogged plots in both 1978 (PERMANOVA:

Change within communities Understory species richness in individual logged plots was, on average, 13% lower in 2007 than in 1978 (t 2.35, DF 14, p 0.03). However, richness did not change in the unlogged plots over the 30-yr study period (t 1.06, DF 14, p 0.31). The decrease in overall understory species richness in the logged plots resulted from decreased richness of herbaceous species (Supplementary material Fig. S1). Shrub, seedling, and tree species richness did not change over time in either logged or unlogged plots (p 0.08 in all cases; Supplementary material Fig. S1). Stem density of trees decreased by an average of 21% in the historically logged plots (t 6.14, DF 14, p B0.0001)

Table 2. Indicator species analysis for compositional differences between logged and unlogged plots. Indicator values (IV) represent the degree to which a species is an indicator of the listed group, with 100 representing perfect indication. Understory includes herbs, shrubs, and seedlingsB1 m tall and tree includes trees ]10 cm dbh. Species with significant IV in at least one year are listed alphabetically. Species name

1978

2007

Group

Understory Acer saccharum Amphicarpaea bracteata Arisaema triphyllum Athyrium filix-femina Betula alleghaniensis Betula lenta Calycanthus floridus Collinsonia canadensis Dioscorea villosa Eurybia divaricata Liriodendron tulipifera Osmunda claytoniana Prunus serotina Quercus rubra Rhododendron maximum Rubus spp. Thalictrum thalictroides Trillium spp. Viola hastata

unlogged unlogged unlogged unlogged logged unlogged unlogged unlogged unlogged unlogged unlogged logged logged unlogged logged logged unlogged unlogged unlogged

44.6 31.7 63.6 12.6 44.9 10.0 53.3 60.0 48.0 51.2 51.2 40.0 72.5 58.9 42.9 55.7 22.2 29.2 49.6

0.16 0.04 0.001 0.91 0.03 0.87 0.002 0.002 0.01 0.16 0.02 0.02 0.001 0.01 0.02 0.02 0.23 0.15 0.07

unlogged unlogged unlogged unlogged logged unlogged unlogged unlogged unlogged unlogged unlogged logged unlogged logged unlogged unlogged unlogged unlogged

55.0 18.7 45.1 44.6 20.0 40.0 37.0 6.7 20.0 59.3 37.2 51.4 31.7 49.9 33.7 36.7 72.3 58.4

0.03 0.45 0.02 0.02 0.25 0.02 0.06 1 0.25 0.04 0.29 0.15 0.63 0.01 0.97 0.05 0.002 0.003

Tree Acer saccharum Betula lenta Magnolia fraseri Prunus pensylvanica Prunus serotina

unlogged logged logged logged logged

39.2 68.3 40.0 33.3 56.9

0.24 0.005 0.05 0.03 0.02

unlogged logged logged logged

62.2 60.4 43.9 50.0

0.03 0.02 0.05 0.02

429

430

0.0005 10.06

0.003

2, 27

7.22

B0.0001 0.001 0.03

0.23 0.13

0.32 0.11

0.43

2, 26

0 0.57 0.33

B0.0001 0.02 0.03 0 0.40 0.36

0.65 0.03 0.003

0.32 0.0001 0.001

Intercept CEC 1978 tree basal area

0.36

p F DF

Model

Partial R2 p Standardized estimate Parameter estimate

Intercept Elevation 1978 understory cover Tree compositional turnover

Logging history affected change in understory species richness and tree density within communities. However, logging history did not affect the extent of withincommunity compositional turnover for either the understory or tree community. Instead, factors such as elevation, soil properties, and community biomass explained withincommunity compositional change. Among-community similarity was unchanged from 1978 to 2007 in the understory communities and in the tree communities of both the logged and unlogged plots.

Understory compositional turnover

Discussion

Predictor(s)

Among-community similarity was unchanged from 1978 to 2007 in both the logged and unlogged plots (Table 4). We found no differences in the multivariate dispersion of understory communities in 2007 compared with 1978 in either the logged (PERMDISP: F 0.18, p 0.74) or unlogged (PERMDISP: F 0.19, p 0.69) plots. Similarly, multivariate dispersion of tree communities did not change between 1978 and 2007 in logged (PERMDISP: F 0.06, p 0.82) or unlogged (PERMDISP: F 1.17, p 0.35) plots.

Dependent variable

Change among communities

Table 3. Results of stepwise multiple regressions for change in species richness and community composition over 30 yr. Understory includes herbs, shrubs, and seedlings B1 m tall and trees includes trees]10 cm dbh.

but did not change in unlogged plots (t 0.30, DF 14, p 0.77) (Supplementary material Fig. S2). Tree basal area did not change over 30 yr in either the historically logged plots (t 2.0, DF 14, p 0.07) or the unlogged plots (t 1.07, DF 14, p 0.30). Historic logging had no effect on within-plot understory compositional turnover (t 0.08, DF 28, p 0.94) or on within-plot tree compositional turnover (t 1.4, DF 27, p 0.17). A model containing elevation and 1978 understory biomass (estimated using percent cover values) accounted for 36% of the variation in within-plot understory compositional turnover (i.e. how similar a plot was to itself over the 30-yr period; p 0.003; Table 3). Overall, plots at higher elevations had lower understory compositional turnover over the 30-yr period than did plots at lower elevations (Fig. 1a). With the exception of one statistical outlier (determined using Cookâ&#x20AC;&#x2122;s D and hat matrix analyses) that contained ca 55% cover of two fern species (Phegopteris hexagonoptera and Dennstaedtia punctilobula), plots with high understory biomass in 1978 had lower understory turnover than did plots with low understory biomass in 1978 (Fig. 1b). Change in understory community composition was not related to tree compositional turnover (r 0.19, p 0.32). Variation in tree compositional turnover was best explained by a model containing soil cation exchange capacity (CEC) and 1978 tree biomass (R2 0.43, p 0.0005; Table 3). Plots with high CEC had lower compositional turnover in the tree community than did plots with low CEC (Fig. 2a). In addition, plots with high tree biomass (estimated using stand basal area) in 1978 had lower tree turnover than did plots with low tree biomass in 1978 (Fig. 2b).

Figure 1. Correlation between understory compositional turnover across 30 yr and elevation (a) and 1978 understory percent cover (b). Filled circles represent logged plots and unfilled circles represent unlogged plots. Arrow in panel b points to an outlying data point that was excluded from the correlation.

Influence of historic logging on community composition In 1978 (50 yr after logging), understory community composition differed only slightly between historically logged and unlogged plots. Historically logged and unlogged plots also differed only slightly in understory composition in 2007 (80 yr after logging). These results are similar to some chronosequence studies comparing understory communities in recently logged forests with understory communities of older forests (Gilliam et al. 1995, Ford et al. 2000). However, other studies have found larger and more persistent differences in the understory communities of anthropogenically disturbed

Figure 2. Correlation between tree compositional turnover and soil cation exchange capacity (a) and 1978 tree basal area (b). Filled circles represent logged plots and unfilled circles represent unlogged plots.

and undisturbed forests (Meier et al. 1995, Flinn and Vellend 2005, Harrelson and Matlack 2006). In our study site, the relatively small differences we observed in community composition between logged and unlogged plots could be due to the short duration of logging activities or to our focus on late-season species. Logging activities lasted for four years in our study area (Schmidt and Hooks 1994), and understory species may have persisted in the soil seedbank and subsequently recolonized logged areas or survived as scattered mature individuals in the logged areas. In addition, our 1978 and 2007 sampling data were collected between mid-June and late-August and included primarily late-season understory species, which may be more resistant to logging effects than vernal species. 431

Table 4. Among-community similarity in logged and unlogged plots in 1978 and 2007. Mean distance of plots within a group from the group centroid (multivariate dispersion) in multivariate space is defined by Bray Curtis dissimilarity. The pseudo F-statistics and p-values were generated from permutational ANOVA tests of differences in among-community similarity between 1978 and 2007. Abundance variable

Understory frequency1 2

Tree density

Group

logged unlogged logged unlogged

Mean9SE distance to centroid 1978

2007

53.0592.14 50.9292.09 51.7092.22 51.1692.58

51.6892.39 49.5992.21 50.8792.41 47.7291.87

p-value

0.18 0.19 0.06 1.17

0.74 0.69 0.82 0.35

Understory frequency is an abundance measure based on the percentage of the 25 subplots within a 20 50 m forest plot in which each species occurred. Tree density is an abundance measure based on the number of individual trees of a species within a 20 50 m forest plot.

Overall, six understory species that were indicators of unlogged plots in 1978 were no longer indicators in 2007 as these species began to recolonize logged plots. Likewise, four understory species that were indicators of logged plots in 1978 were no longer indicators in 2007 as earlycolonizing species became less abundant in the logged plots. Seven understory species became indicators of unlogged plots between 1978 and 2007. Three of the seven species are spring ephemerals that generally increase in abundance as forests mature and their inclusion as indicator species of unlogged plots in 2007 but not in 1978 could be a result of variation in phenology between sample years. The remaining species that became indicators of unlogged plots in 2007 either became newly established in unlogged plots or remained unchanged in unlogged plots while occurring in fewer logged plots over time. This indicates that after 150 to 200 yr, the unlogged plots are still undergoing measureable changes in community composition. In addition, it suggests that some species associated with older forests may become established and subsequently lost then regained as disturbed forests mature. We note that although spatial autocorrelation may have contributed to differences between logged and unlogged plots, the stratified random sampling design, wide spatial distribution of plots within disturbance types, and physical similarity of logged and unlogged sample areas likely reduced its effects. Factors influencing change within communities Even though change in understory species richness and tree density differed between logged and unlogged plots, the occurrence of historic logging did not affect the extent of within-community compositional turnover in the understory community or the extent of within-community tree compositional turnover during our 30-yr study period. We had expected greater compositional change in the understory of logged plots than in unlogged plots during this period due to the changing understory light environment as the logged plots moved through the stem exclusion and understory reinitiation stages of development (Oliver and Larson 1996) and tree density decreased. It may be that we found no difference in the extent of compositional change in the understory of logged and unlogged plots because logging events that occurred 50 yr ago no longer affect understory dynamics in these forests. Also contrary to our expectations, the extent of tree compositional turnover in 432

unlogged plots was similar to tree turnover in logged plots. Since tree density in the logged plots was almost twice as high as tree density in the unlogged plots in 1978, the loss of individual trees likely had a smaller effect on compositional turnover in the logged plots than in the unlogged plots, which could explain the similar extent of compositional turnover we observed in logged and unlogged plots. Elevation and community biomass best explained compositional turnover within forest understory communities. Consistent with other studies (Aplet and Vitousek 1994, Selmants and Knight 2003), we found that the extent of turnover in the understory community decreased along the elevational gradient. Mean annual temperature decreases ca 48C and soils become more acidic over the 740 m elevational gradient in our study site (Garten and Hanson 2006). The lower temperatures at higher elevations could slow decomposition, decrease nutrient availability, and reduce overall plant growth (Vitousek et al. 1992, Aplet and Vitousek 1994), which might result in lower compositional turnover at high elevations. In Great Smoky Mountains National Park, decomposition rates generally decrease with elevation, but nitrogen availability increases due to low soil C-to-N ratios at high elevations (Garten 2004). We did not measure decomposition rates or nitrogen availability in our study plots; therefore, it is unclear whether slower ecosystem processes at high elevations are responsible for the smaller changes in understory composition over time that we observed. In addition to temperature, soil pH also varies with elevation in our study sites, with more acidic soils at higher elevation sites. In the acidic soils of southeastern U.S. forests, higher soil pH can indicate greater nutrient availability to plants. In the understory, greater nutrient availability could lead to increased compositional turnover either by increasing the likelihood that newly arriving species will establish in a community (Peet and Christensen 1988) or by increasing the growth of dominant species that could out-compete other species in the community. In addition, the large regional pool of species that favor high pH sites (Peet et al. 2003) could increase the chance that new species would colonize these sites over time or that more species would be present in the initial community. A larger pool of potential colonizers could increase understory compositional change in high pH communities compared to low pH communities. Unlike understory compositional turnover, tree compositional turnover was not related to elevation, but was related to soil CEC. Because turnover of individual tree

stems can be a function of elevation, latitude, and productivity in some forest systems (Phillips et al. 2004, Stephenson and van Mantgem 2005), we expected higher compositional turnover in the tree community in low elevation plots than in high elevation plots. However, elevation was not correlated with within-site change in tree community composition, stem density, or tree species richness over time in our study. We did not directly measure turnover of individual trees and cannot say whether the rate of stem recruitment or mortality changed with elevation. For trees, it may be that fertile sites allow faster tree growth regardless of elevation. Faster tree growth could increase recruitment into the overstory or increase mortality through competitive exclusion. Either increased recruitment or mortality could lead to greater tree compositional turnover in more fertile plots. Understory biomass in 1978 was correlated with understory compositional turnover, and tree biomass in 1978 was correlated to tree compositional turnover. Plots with high understory percent cover (an estimate of understory biomass; Gilliam and Turrill 1993) in 1978 had lower compositional turnover than did plots with low understory cover in 1978. Similarly, plots with high tree basal area (a surrogate for tree biomass; Wardle et al. 2008) in 1978 had lower tree compositional turnover than did plots with low tree basal area in 1978. Understory percent cover and tree basal area in 1978 were not correlated with any measured topographic or edaphic factors (Supplementary material Table S3). Sites with higher biomass may have low within-site turnover because in higher biomass sites, a larger proportion of the plot is occupied by established species. The proportion of a plot initially occupied might affect community change in these forests over time due to resident species excluding new species from establishing or limiting the population growth of other resident species. Influence of historic logging on change among communities We documented a wide range of changes in species composition within communities, but these withincommunity changes did not translate into a change in among-community similarity. We found no change in among-community similarity from 1978 to 2007 in either the logged and unlogged plots. Evidence that amongcommunity similarity is lower in forests formerly disturbed by agriculture compared with older forests (Christensen and Peet 1984, Vellend et al. 2007) suggests that younger forests may become more similar to one another over time. However, we found no change in among-community similarity in either understory or tree communities of the logged plots. We also found no change in amongcommunity similarity in plant communities of unlogged plots over this same 30-yr time period. The theoretical expectation for temporal change in among-community similarity in undisturbed forests is unclear. It may be that the 30-yr study period here was a relatively stable period within larger cycles of amongcommunity heterogeneity in forest development. Additionally, among-community heterogeneity may have reached a

static point where it will remain unchanged in the absence of further disturbance (RejmaÂ´nek and RoseÂ´n 1992).

Conclusions The extent of change in community composition was not related to historic disturbance, and among-community similarity did not change over time in either historically disturbed or undisturbed plots. Our results indicate that the extent of change in community composition over time may depend more on environmental gradients and community attributes than on the legacy of large-scale, but short-lived historic disturbances such as logging. In addition, variation in turnover within communities may not necessarily translate into changes in compositional similarity among communities over time. Additional long-term studies that directly measure temporal change both within and among communities are needed in order to increase our understanding of the factors that control multi-scale diversity across time and space.

Acknowledgements We are grateful to A. Classen, D. Simberloff, H. H. Bruun, and two anonymous reviewers for helpful comments on the manuscript. L. Souza assisted with the 2007 data collection and J. Rock assisted with plant identification. Funding for the resampling portion of this project was provided to W.A.B. by the Dept of Ecology and Evolutionary Biology at the Univ. of Tennessee.

References Anderson, M. J. 2001. A new method for non-parametric multivariate analysis of variance. Austral Ecol. 26: 32 46. Anderson, M. J. 2004. PERMDISP: a FORTRAN computer program for permutational analysis of multivariate dispersions (for any two-factor ANOVA design) using permutation tests. Dept of Statistics, Univ. of Auckland, New Zealand. Anderson, M. J. 2005. PERMANOVA: a FORTRAN computer program for permutational multivariate analysis of variance. Dept of Statistics, Univ. of Auckland, New Zealand. Anderson, M. J. et al. 2006. Multivariate dispersion as a measure of beta diversity. Ecol. Lett. 9: 683 693. Anderson, T. M. 2008. Plant compositional change over time increases with rainfall in Serengeti grasslands. Oikos 117: 675 682. Aplet, G. H. and Vitousek, P. M. 1994. An age-altitude matrix analysis of Hawaiian rain forest succession. J. Ecol. 82: 137 147. Chalcraft, D. R. et al. 2004. Scale dependence in the speciesrichness-productivity relationship: the role of species turnover. Ecology 85: 2701 2708. Chao, A. et al. 2005. A new statistical approach for assessing similarity of species composition with incidence and abundance data. Ecol. Lett. 8: 148 159. Chase, J. M. and Leibold, M. A. 2002. Spatial scale dictates the productivity-biodiversity relationship. Nature 416: 427 430. Christensen, N. L. and Peet, R. K. 1984. Convergence during secondary forest succession. J. Ecol. 72: 25 36. Collins, S. L. and Smith, M. D. 2006. Scale-dependent interaction of fire and grazing on community heterogeneity in tallgrass prairie. Ecology 87: 2058 2067.

433

Colwell, R. K. 2005. EstimateS: statistical estimation of species richness and shared species from samples. /<purl.oclc.org/ estimates/>. DufreË&#x2020;ne, M. and Legendre, P. 1997. Species assemblages and indicator species: the need for a flexible asymmetrical approach. Ecol. Monogr. 67: 345 366. Flinn, K. M. and Vellend, M. 2005. Recovery of forest plant communities in post-agricultural landscapes. Front. Ecol. Environ. 3: 243 250. Ford, W. M. et al. 2000. Stand-age, stand characteristics, and landform effects on understory herbaceous communities in southern Appalachian cove-hardwoods. Biol. Conserv. 93: 237 246. Foster, D. R. et al. 1998. Land-use history as long-term broadscale disturbance: regional forest dynamics in central New England. Ecosystems 1: 96 119. Garten, C. T. 2004. Potential net soil N mineralization and decomposition of glycine-C-13 in forest soils along an elevation gradient. Soil Biol. Biochem. 36: 1491 1496. Garten, C. T. and Hanson, P. J. 2006. Measured forest soil C stocks and estimated turnover times along an elevation gradient. Geoderma 136: 342 352. Gilliam, F. S. and Turrill, N. L. 1993. Herbaceous layer cover and biomass in a young versus a mature stand of a central Appalachian hardwood forest. Bull. Torrey Bot. Club 120: 445 450. Gilliam, F. S. et al. 1995. Herbaceous-layer and overstory species in clear-cut and mature central Appalachian hardwood forests. Ecol. Appl. 5: 947 955. Gotelli, N. J. and Colwell, R. K. 2001. Quantifying biodiversity: procedures and pitfalls in the measurement and comparison of species richness. Ecol. Lett. 4: 379 391. Harrelson, S. M. and Matlack, G. R. 2006. Influence of stand age and physical environment on the herb composition of secondgrowth forest, Strouds Run, Ohio, USA. J. Biogeogr. 33: 1139 1149. Jenkins, M. A. et al. 2007. Impacts of an exotic disease and vegetation change on foliar calcium cycling in Appalachian forests. Ecol. Appl. 17: 869 881. Loreau, M. 2000. Are communities saturated? On the relationship between alpha, beta and gamma diversity. Ecol. Lett. 3: 73 76. Magnuson, J. J. 1990. Long-term ecological research and the invisible present. Bioscience 40: 495 501. Magurran, A. E. 2004. Measuring biological diversity. Blackwell. Meier, A. J. et al. 1995. Possible ecological mechanisms for loss of vernal-herb diversity in logged eastern deciduous forests. Ecol. Appl. 5: 935 946. Oliver, C. D. and Larson, B. C. 1996. Forest stand dynamics. Wiley. Peet, R. K. and Christensen, N. L. 1988. Changes in species diversity during secondary forest succession on the North

Download the Supplementary material as file E6016 from / <www.oikos.ekol.lu.se/appendix/>.

434

Carolina Piedmont. In: During, H. J. et al. (eds), Diversity and pattern in plant communities. SPB Academic Publ., pp. 233 245. Peet, R. K. et al. 2003. Variation in species richness and species pool size across a pH gradient in forests of the southern Blue Ridge Mountains. Folia Geobot. 38: 391 401. Phillips, O. L. et al. 2004. Pattern and process in Amazon tree turnover, 1976 2001. Phil. Trans. R. Soc. B 359: 381 407. Pyle, C. 1988. The type and extent of anthropogenic vegetation disturbance in the Great Smoky Mountains before National Park Service acquisition. Castanea 53: 183 196. RejmaÂ´nek, M. and RoseÂ´n, E. 1992. Cycles of heterogeneity during succession a premature generalization. Ecology 73: 2329 2331. Rooney, T. P. et al. 2004. Biotic impoverishment and homogenization in unfragmented forest understory communities. Conserv. Biol. 18: 787 798. Schmidt, R. G. and Hooks, W. S. 1994. Whistle over the mountain: timber, track & trails in the Tennessee Smokies: an historical and field guide to the Little River Lumber Company and the Little River Railroad in the Great Smoky Mountains National Park in Tennessee. Graphicom Press. Selmants, P. C. and Knight, D. H. 2003. Understory plant species composition 30 50 years after clearcutting in southeastern Wyoming coniferous forests. For. Ecol. Manage. 185: 275 289. Smart, S. M. et al. 2006. Biotic homogenization and changes in species diversity across human-modified ecosystems. Proc. R. Soc. B 273: 2659 2665. Stephenson, N. L. and van Mantgem, P. J. 2005. Forest turnover rates follow global and regional patterns of productivity. Ecol. Lett. 8: 524 531. Taverna, K. et al. 2005. Long-term change in ground-layer vegetation of deciduous forests of the North Carolina Piedmont, USA. J. Ecol. 93: 202 213. Vellend, M. et al. 2007. Homogenization of forest plant communities and weakening of species-environment relationships via agricultural land use. J. Ecol. 95: 565 573. Verheyen, K. et al. 2003. Herbaceous plant community structure of ancient and recent forests in two contrasting forest types. Basic Appl. Ecol. 4: 537 546. Vitousek, P. M. et al. 1992. The Mauna-Loa environmental matrix foliar and soil nutrients. Oecologia 89: 372 382. Wardle, D. A. et al. 2008. The response of plant diversity to ecosystem regression: evidence from contrasting long-term chronosequences. Oikos 117: 93 103. White, E. P. et al. 2006. A comparison of the species-time relationship across ecosystems and taxonomic groups. Oikos 112: 185 195. Yurkonis, K. A. and Meiners, S. J. 2004. Invasion impacts local species turnover in a successional system. Ecol. Lett. 7: 764 769.

ECOGRAPHY 29: 111 /119, 2006

Impact of warming and timing of snow melt on soil microarthropod assemblages associated with Dryas-dominated plant communities on Svalbard Rebecca Dollery, Ian D. Hodkinson and Ingibjo¨rg S. Jo´nsdo´ttir

Dollery, R., Hodkinson, I. D. and Jo´nsdo´ttir, I. S. 2006. Impact of warming and timing of snow melt on soil microarthropod assemblages associated with Dryas- dominated plant communities on Svalbard. / Ecography 29: 111 /119. Open Top Chambers (OTCs) were used to measure impacts of predicted global warming on the structure of the invertebrate community of a Dryas octopetala heath in West Spitsbergen. Results from the OTC experiment were compared with natural variation in invertebrate community structure along a snowmelt transect through similar vegetation up the adjacent hillside. Changes along this transect represent the natural response of the invertebrate community to progressively longer and potentially warmer and drier growing seasons. Using MANOVA, ANOVA, Linear Discriminant Analysis and x2 tests, significant differences in community composition were found between OTCs and controls and among stations along the transect. Numbers of cryptostigmatic and predatory mites tended to be higher in the warmer OTC treatment but numbers of the aphid Acyrthosiphon svalbardicum , hymenopterous parasitoids, Symphyta larvae, and weevils were higher in control plots. Most Collembola, including Hypogastrura tullbergi , Lepidocyrtus lignorum and Isotoma anglicana , followed a similar trend to the aphid, but Folsomia bisetosa was more abundant in the OTC treatment. Trends along the transect showed clear parallels with the OTC experiment. However, mite species, particularly Diapterobates notatus, tended to increase in numbers under warming, with several species collectively increasing at the earlier exposed transect stations. Overall, the results suggest that the composition and structure of Arctic invertebrate communities associated with Dryas will change significantly under global warming. R. Dollery and I. D.Hodkinson (correspondence: i.d.hodkinson@ljmu.ac.uk), School of Biological & Earth Sciences, Liverpool John Moores Univ., Byrom St., Liverpool, L3 3AF, UK. / I. S. Jo´nsdo´ttir, Univ. Centre in Svalbard (UNIS), Post box 156, Longyearbyen, N-9171 Svalbard, Norway.

Many field studies have been conducted on the potential effects of global warming on the growth and community structure of vegetation in polar/alpine ecosystems. These experiments have frequently used some form of passive warming such as a cloche, miniature glasshouse or opentop chamber (OTC) to raise the ambient temperature (Strathdee and Bale 1993, Kennedy 1995, Chapin et al. 1995, Marion et al. 1997, Day et al. 1999, Hollister and Webber 2000, Convey et al. 2002, Sjursen et al. 2005).

Such studies are often standardised in design and form part of a broader geographical network of sites (e.g. The International Tundra Experiment, ITEX) that allow comparison of responses on a circumpolar scale (Molau and Mølgaard 1996). By contrast, equivalent experimental studies on invertebrates are sparse and comparative studies across sites, despite relatively good baseline data derived from the efforts of the International Biological Programme Tundra Biome (Ryan 1981), are

Accepted 19 September 2005 Copyright # ECOGRAPHY 2006 ISSN 0906-7590 ECOGRAPHY 29:1 (2006)

111

few. Detailed sampling programmes for invertebrates, particularly within the soil, have rarely been incorporated into the initial design of these experiments. Consequently, excellent opportunities have been lost and our understanding of how animal communities might respond to changing climate and their impact on soil nutrient dynamics is based on limited data for relatively few sites. Most studies on the effects of climate warming on polar/alpine invertebrate communities have focused on communities associated with particular broad types of vegetation, such as polar semi-desert, tundra heath, montane mixed herb or moss communities (Ryan 1981, Kennedy 1994, Hodkinson et al. 1996, 1998, Rae 2003) or on particular above-ground herbivores associated with selected plant species (Strathdee et al. 1993, 1995, Roy et al. 2004). However, invertebrate communities living in the soil beneath different plant species in tundra, including Dryas octopetala , despite their similar low species richness, may differ in their structure, particularly the relative abundance of species (Coulson et al. 2003). This is probably explained by subtle differences in the physico-chemical environment, including soil temperature and moisture content, below different plant species. Despite this clear link between the soil fauna, vegetation and microclimate, few attempts have been made to exploit the numerous temperature manipulation experiments on plants within polar or montane habitats to measure simultaneously parallel changes in the soil fauna. Where it has been done soil invertebrates have usually been sampled retrospectively (Ruess et al. 1999a,b, Convey and Wynn-Williams 2002, Convey et al. 2002, Sjursen et al. 2005). Experiments to date suggest that different groups of invertebrates respond differently to climate warming. Soft-bodied arthropods, with permeable cuticles, such as soil-dwelling Collembola and some prostigmatic mites, tend to be desiccation susceptible and their populations are apt to decline within experimental chambers as temperatures rise (Hodkinson et al. 1996, Convey and Wynn-Williams 2002, Convey et al. 2002, Sjursen et al. 2005). Cryptostigmatic mites, by contrast, tend to be desiccation resistant and appear less affected, although their slow rate of population turnover makes them potentially slow responders to rising temperature (Hodkinson et al. 1996, Webb et al. 1998). Some above ground herbivores, including the Dryas -feeding aphid Acyrthosiphon svalbardicum responded rapidly within one season, producing a 20-fold increase in the number of overwintering eggs (Strathdee et al. 1993, 1995). Snowmelt gradients reflect in many ways natural conditions that are analogous to those changes that are likely to occur over time in an artificial warming experiment. These include a lengthening of the growing season, warmer mean temperatures and changes in soil moisture levels. This in turn is reflected in differences in 112

plant phenology, growth, flowering and reproduction and distribution, which, together with changes in the rate of soil microbial processes, has consequences for both the above-ground insect herbivores and the soil invertebrates (Galen and Stanton 1995, Walker et al. 1995, Stanton et al. 1997, Sinclair and Sjursen 2001, Heegaard 2002, Totland and Alatalo 2002, Inouye et al. 2003, Van Wijk et al. 2003, Gerber et al. 2004, StenstroÂ¨m and JoÂ´nsdoÂ´ttir 2005). This paper reports the results of an Open Top Chamber (OTC) experiment designed to test and measure the impacts of predicted global warming on the structure and species composition of the invertebrate communities of a Dryas octopetala heath in West Spitsbergen, Svalbard. Results and trends from the OTC experiment are compared with variation in invertebrate community structure measured at four stations along a natural snowmelt transect through similar Dryas vegetation up the adjacent hillside. Changes along this transect represents the natural response of the invertebrate community to progressively longer and potentially warmer average growing seasons. The two complementary experiments thus allow comparison of spatial variation in the invertebrate community with respect to local microclimate and the long-term community response to externally imposed climatic variation. Dryas octopetala dominated plant communities are widespread throughout the eastern Arctic where they tend to occupy drier, better-drained sites, often on limestone or other basic substrates (Bliss and Matveyeva 1992, Anon. 2003). They are common and broadly distributed on Svalbard (Elvebakk 1997).

Materials and methods Experimental site The study was conducted on a southeast-facing hillside in Endalen, West Spitsbergen Svalbard (15845?E, 78811?N) where Dryas octopetala heath formed the dominant vegetation type. In addition to the dominant D. octopetala , Salix polaris, Saxifraga oppositifolia , Bistorta vivipara and Carex rupestris also occurred with high frequency. Sampling was conducted along and adjacent to a 135 m snowmelt belt transect (width 20 m) extending directly up the hillside, with transect station 1 situated at the top and station 4 at the bottom. Stations 2 and 3 were spaced equally between. The direction of spring snowmelt was from the bottom to the top of the transect and represented a gradient of decreasing length of growing season and cooler spring temperatures. Late snowmelt, however, ensures that soil moisture remains higher for longer during the early growing season. The OTC experiment straddled the mid point stations (2 and 3) of the snowmelt transect. During 2003 snow clearance dates at transect stations 4, 3, 2, ECOGRAPHY 29:1 (2006)

and 1 was 31 May, 3 June, 7 June and after 10 June, respectively. Snow melt on the main treatment plots was ca 10 d earlier than on the equivalent control plots.

Animal sampling Five randomly located experimental plots were covered with standard hexagonal ITEX design OTCs with a base diameter of 1.5 m (Molau and Mølgaard 1996) at the beginning of the 2001 growing season. A further 8 plots of similar dimensions were identified as controls. The original purpose of the experiment was to examine the response of the plant community to warming and to this end each OTC enclosed a small central inner 0.75 /0.75 m plot, which was sampled continuously but non-destructively for vegetation. Samples for animals, comprising both soil core and pitfall trap samples, were taken randomly in summer 2003 within each chamber but avoiding the central plot. They were thus taken from the broad outer band between the central plot and the edge of the larger enclosing chamber. Five soil core samples per plot (each 4 cm diameter, depth 4 /5 cm) were taken between 3 and 8 July. Equivalent samples were taken from each control plot, where the same spatial pattern of sampling was replicated as in the treatment plots. Core samples were extracted for soil invertebrates using Macfadyen high gradient extractors (Leinaas 1978), collected into benzoic acid solution and then stored in 70% alcohol for subsequent identification. Within each treatment and control plot five 4-cm diameter pitfall traps containing water with a drop of surfactant were layed out on 8 July 2003 and emptied on 11 and 26 July. A further set of soil cores, as above, were taken and extracted from the snowmelt transect between 8 and 21 July. Extractor availability prevented all cores being taken simultaneously and sampling was equally distributed between sites on each date. At each of the four stations samples were taken at 1 m intervals (n /20) along a line extending horizontally 10 m either side of central spot defining the position of that station. Nomenclature of animals follows Fjellberg (1994), Coulson et al. (2003) and Coulson and Refseth (2004).

Environmental measurements Temperature was measured in OTC treatment and control plots using Tinytag Plus data loggers with standard microthermistor probes. Temperatures were recorded in the air immediately above the soil surface, within the top 1 cm of soil and at a height of 1.5 m above the soil. Soil moisture was measured on treatment and control plots and at the various stations along the snowmelt transect, using a HH Soil Moisture Meter with a ML2x/d Theta probe. ECOGRAPHY 29:1 (2006)

Statistical analyses General Linear Model MANOVA, on Log (n /1) transformed invertebrate population data, was used to test for overall differences between treatments, sample plots and replicates in both the OTC and transect experiments. Wilks’ lambda was used as the test statistic. General Linear Model ANOVA, using adjusted sums of squares, was then used to analyse differences in individual species population densities between treatments in the OTC experiment. One-way ANOVA, with post-hoc Tukey and Fisher pairwise comparisons, were applied to compare invertebrate community data for stations along the transect. In these comparisons, involving a large number of tests on individual species, it is recognised that there is a random chance of the occasional spurious significant result. The distinctness of the invertebrate community samples in the OTC treatment and controls and among transect stations was further tested using Linear Discriminant Analysis of individual samples and x2 analyses testing for differences in relative species abundances between treatments in the OTC experiment and among communities at different stations along the transect. In these tests less common species were grouped under the collective heading ‘‘minor taxa’’. For the x2 tests, the null hypothesis was that the numbers of individuals in each species as a proportion of the number of individuals in the whole community did not differ between treatments. The overall x2 value was thus used to indicate significant differences between the treatments/transect stations while the contributing x2 values for individual species was used to identify species that made a significant contribution to the overall x2 value.

Results Microclimate effects Over the summer period 18 June /5 August hourly mean air temperature inside the OTC was 1.28C higher than in the controls and 1.68C higher than the temperature at 1.5 m above the ground (Fig. 1). Soil temperature within the OTC was enhanced by 1.48C compared with the control plot. Soil moisture was variable over short distances and this masked possible trends along the transect and between OTC treatment and control plots.

Invertebrate community composition The full list of species captured and their distribution across sampling sites and stations (Appendix 1) shows a few rare species with highly restricted distributions. These included the Collembolans Micranurida pygmaea and Arrhopalites principalis found occasionally 113

Fig. 1. Mean air and soil temperatures within OTC and control plots in Endalen during the summer growing season 2003.

at transect station 1 or on control plots. By contrast the weevil Isochnus flagellum appeared only towards the bottom of the transect. Most species, however, were broadly distributed, albeit at varying densities, across the sites. MANOVA across all taxa showed significant treatment effects in the OTC core (Wilks, F21,31 /3.82, pB/ 0.001), the OTC pitfall (Wilks, F30,22 /12.53, pB/0.001) and in the transect core samples (Wilks, F75,99 /1.59, pB/0.001). There was an additional significant between sample plot effect for OTC core samples (Wilks, F147,217 /1.47, pB/0.01), but not for OTC pitfall or transect station core samples. The Linear Discriminant Analysis model for the relative abundance of species in soil core samples in the OTC chamber experiment correctly allocated samples to the OTC treatment or control groups with 92 and 87% accuracy respectively. This suggests a high level of predictability with respect to the distinctness of the invertebrate communities in the two treatments. Similarly, a x2 analysis, used to test the hypothesis that the relative species abundance within the invertebrate community of the OTC treatment soil core samples (based on total numbers caught) differed from that in the control samples, was highly significant (x2 /118, pB/0.001, DF /18) (Table 1). The equivalent analysis for pitfall samples was correspondingly highly significant (x2 /1146, pB/0.001, DF /19) (Table 1). The five species making the highest contributions to the overall x2 value in each of these tests are highlighted. In the OTC experiment, the collembolan Folsomia bisetosa was more abundant than expected (treatment vs control) in soil core samples whereas the collembolans Hypogastrura tullbergi and Folsomia quadrioculata were underrepresented. By contrast, the mites Diapterobatus notatus and Prostigmata were overrepresented in pitfall treatment samples. Numbers of the aphid Acyrthosiphon svalbardicum were consistently lower than expected in both soil core and pitfall treatment samples. Results of the GLM analysis for individual species within the OTC experiment showed a number of 114

significant effects, with differences being more accentuated for species in the pitfall than the core samples (Table 2). For pitfall samples, the cryptostigmatic mites D. notatus and Hermannia reticulata , the prostigmatic mites and the gamasid mites were significantly more abundant in OTC samples than in controls. The aphid A. svalbardicum , the collembolans Isotoma anglicana , H. tullbergi, the hymenopteran parasitoid Stenomacrus groenlandicus and Symphyta larvae were most abundant in the controls. The coleopteran I. flagellum and total aphid parasitoids data were similarly close to significance, with greater numbers of individuals caught in the control treatment (see Fig. 2). For core samples, total cryptostigmatic mites, but not individual species alone, were significantly more abundant in the OTCs whereas A. svalbardicum , H. tullbergi , Tetracanthella arctica and Symphyta larvae were more abundant in controls. In addition to the main treatment effect there were also significant among-plot effects for Lepidocyrtus lignorum , I. anglicana , H. tullbergi , Oligophura ursi , H. reticulata and Symphyta larvae and among-sample effects in L. lignorum and H. reticulata . The Linear Discriminant Analysis model of the invertebrate relative species abundance data for the transect allocated samples from the upper and lower limits of the gradient to their correct station with a high degree of certainty (80%), indicating that the communities at the gradient extremes were distinct (Table 3). Communities at the intermediate stations again showed relatively high distinctness, with a majority of samples (60 or 75%) being correctly classified but with higher overlap with other stations. Misclassified samples in most cases were placed in the group for an adjacent station. These differences were again reflected in the highly significant results of a x2 analysis (Table 1) (x2 /118, pB/0.001, DF /18) used to test the hypothesis that the relative species abundances of invertebrates (based on total individuals captured) was identical in samples from different stations along the transect. Taxa displaying a tendency to be more abundant than expected towards the upper end of the transect (Table 1) included Enchrytraeidae, A. svalbardicum , F. quadrioculata and L. lignorum . Species appearing more abundantly towards the bottom of the transect included the mites D. notatus and H. reticulata and the collembolan Tetracanthella arctica . The mite Camisia anomia uniquely lacked a recognizeable trend in abundance. Taxa showing a general trend of decrease from the top (station 1) to the bottom (station 4) in the ANOVA of snowmelt transect data included L. lignorum , H. tullbergi , A. principalis, Areneae and Enchytraeidae (Table 4). Other species, such as F. quadrioculata and Folsomia bisetosa , displayed maximum density at the mid point of the transect, peaking at stations 2 or 3, with the former showing lowest population density at the bottom of the transect. Populations of other species, ECOGRAPHY 29:1 (2006)

F. quadrioculata *** ( /) C. anomia ** ( /) D. notatus ( /) T. arctica ( /) H. reticulata ( /)

F. bisetosa ( /) A. svalbardicum *** ( /) Enchytraeidae ( /) Other minor taxa ( /) D. notatus ( /) F. quadrioculata ( /) C. anomia ( /) H. tullbergi ( /) H. reticulata ( /) Cryptostigmata indet. ( /) Enchytraeidae ( /) F. bisetosa ( /) D. notatus ( /) A. svalbardicum ( /) L. lignorum ( /) A. svalbardicum *** ( /) L. lignorum *** ( /) Adult Chironomidae*** ( /) Prostigmatic mites** ( /) D. notatus ** ( /) H. tullbergi * ( /) A. svalbardicum ( /) F. bisetosa ( /) F. quadrioculata ( /) Chironomidae larvae ( /) 1 2 3 4 5

OTC pitfalls

Transect cores

Discussion

OTC soil cores

Table 1. Summary of the five species that made the greatest contribution, in descending order of importance, to the significance of each of the x2 tests outlined in the text. ( /) indicates greater than expected numbers, ( /) indicates lower than expected numbers. In the case of the OTC experiment this applies to numbers in the OTC treatment versus the control. For the transect it represents deviation from a uniform distribution of species abundances along the gradient. Species that on their own exceed the critical probability value for the overall x2 total in a particular test are indicated by asterisks * p B/0.05, ** pB/0.01, *** p B/0.001. ‘‘Other minor taxa’’ is the grouped total number of individuals in the rarer species and includes some insects, collembolans and mites. ECOGRAPHY 29:1 (2006)

notably Camisia anomia , rose abruptly at the bottom site. Overall these results suggest significant changes in invertebrate community structure along the snowmelt gradient that can be identified in samples taken at successive stations.

One of the main criticisms of retrospective sampling of manipulation experiments, as applied in this and several previous studies, is that equivalent sampling was not carried out at the time that the experiment was established, i.e. there is no absolute baseline at time zero for comparison. Thus, there is a small probability that differences observed among treatments, after a period of elapsed time, existed by chance at the time the experiment was initiated. Furthermore, it is possible that exogenous factors, acting over the duration of the experiment, have produced parallel changes in both treatment and control plots. The parallel trends observed in the transect and OTC experiments suggest, however, that this is not the case. Soil core samples and pitfall traps provide different but complementary information on the invertebrate community. Core samples give a quantitative estimate of invertebrates living in a unit area of soil. Pitfall traps catch many more animals, primarily those moving around on the soil surface, but they also catch small insects that are flying among the low vegetation such as several parasitoid Hymenoptera species. The difference between methods is strongly reflected in the Collembola where pitfalls tend to catch the large surface active species such as L. lignorum and I. anglicana that are more rarely seen in core samples. Conversely, true soil dwelling species such as F. quadrioculata and P. ursi are rarely found in pitfalls. Aphids appear to be efficiently extracted from the base of Dryas in the heated core samples but they also fall into pitfalls in large numbers. Both the OTC and transect experiments revealed significant changes or differences in the structure of the invertebrate community in response to rising ambient temperatures or extended growing season, clearly indicating the potential response of these communities to global warming. Thus, despite broad similarities in the species present, significant differences in the invertebrate community composition between treatments were established for both core and pitfall samples in the OTC experiment and among stations along the snowmelt transect using MANOVA. Differences were sufficient for individual samples in both experiments to be classified as to their origin with a high degree of probability (60% /) using a Linear Discriminant Analysis model and for the significant distinctiveness of the communities between treatments or sites to be demonstrated using x2 analyses. 115

Table 2. Results of General Linear Model ANOVA summarising significant effects for different taxa in the OTC chamber experiment. Probability values for soil core and pitfall trap effects are listed separately. * Signifies close to significant. ( /) indicates significantly more in the OTC treatment than in the control, ( /) indicates significantly fewer than in the control. Pitfall traps source of variation Taxon A. svalbardicum Total aphid parasitoids L. lignorum I. anglicana H. tullbergi O. ursi T. arctica F. quadrioculata Total Collembola

Between treatments

Among plots

B/0.001 ( /) 0.08* B/0.05 ( /) B/0.01 ( /)

Between treatments

Among plots

Among samples

B/0.05 ( /) B/0.01 B/0.01

B/0.01 B/0.001 ( /)

B/0.01 B/0.05

B/0.05 ( /) B/0.05 B/0.01

Prostigmastic mites Gamasid mites Total predatory mites

B/0.01 ( /) B/0.001 ( /) B/0.01 ( /)

D. notatus H. reticulata Total cryptostigmatic mites

B/0.05 ( /) B/0.05 ( /) B/0.05 ( /)

B/0.05 B/0.05

S. groenlandicus Symphyta larvae I. flagellum

B/0.05 ( /) B/0.001 ( /) 0.06*

B/0.05

The main trends in the OTC experiment tended to be higher numbers of both cryptostigmatic and predatory mites in the warmed treatment plots and higher numbers of aphids, hymenopterous parasitoids, Symphyta larvae,

Fig. 2. Comparison of the captures of the aphid A. svalbardicum (numbers per trap9/SE), its mite predators and hymenopterous parasitoids in pitfall traps placed within OTC and control plots. Data for the two sampling periods are presented separately to illustrate the consistency of the trends. Significant differences in ANOVA tests between control and treatment plots are indicated by ***pB/0.001, **pB/0.01, *pB/0.05.

116

Among samples

Soil cores source of variation

B/0.05 B/0.001 ( /)

B/0.05

B/0.001 ( /)

B/0.001

and weevils in the control plots. Most Collembola, including H. tullbergi , L. lignorum and I. anglicana followed a similar trend to the aphid but F. bisetosa was commoner in the OTC treatment, probably reflecting its broader thermal and humidity tolerances (Fjellberg 1994). These results suggest that future warming is likely to produce a significant shift in the composition and structure of the invertebrate communities associated with Dryas. The OTC chambers, as expected, appeared artificially to produce effects that mimicked those that occurred naturally along the snow melt transect. These included an extended growing season, warmer soil temperatures and possibly lower soil moisture, to which the soil invertebrate fauna is collectively responding. Soil animal species responded in analogous ways across experiments, with trends along the transect showing several parallels with the OTC experiment. Aphids and the moistureloving Enchytraeidae, taxa that decreased in the OTC treatment, were more abundant at the later exposed transect stations. Collembola species similarly tended to follow the same negative response to warming as in the OTC experiment, although F. bisetosa increased in abundance in the OTCs and was underrepresented in the site 1 community. Mite species, particularly D. notatus, by contrast, tended to increase in numbers under warming with several species collectively tending to increase in abundance at the earlier exposed transect stations. Our results show similar trends to climatic temperature manipulation experiments elsewhere and conform ECOGRAPHY 29:1 (2006)

Table 3. Results of Linear Discriminant Analysis showing the allocation of soil core samples from the four stations (n /20 per station) along the snowmelt transect to their correct or incorrect group. True group Allocated group Station 1 Station 2 Station 3 Station 4 % correct

Station 1

Station 2

Station 3

Station 4

16 2 2 0 80

2 12 6 0 60

1 3 15 1 75

1 1 2 16 80

Localised variation in some extraneous factor such as soil moisture or distribution of a particular plant species may provide an explanation. Similar spatial aggregation, independent of treatment, occurred in soil arthropod communities in chamber experiments on Anvers Island, West Antarctic Peninsula (Convey et al. 2002). A major contrast in response to climate warming is seen in the aphid A. svalbardicum in the Endalen experiment compared with previous studies at ˚ lesund where enclosed cloches produced a 20-fold Ny-A increase in population over one year compared with controls (Strathdee et al. 1993, 1995). Initial nonquantitative casual observations during year 1 suggested that this was happening also in Endalen. However, the populations at Endalen within the OTCs were significantly lower after the second season, as measured by both core and pitfall sampling, than in the controls. Furthermore, populations of aphids were higher towards the upper cooler end of the transect. There are several possible explanations. Endalen is significantly warmer ˚ lesund. It is known that enhanced temperathan Ny-A tures produced a stimulatory effect on growth and reproduction at low temperatures. However, it may be that once temperatures exceed an optimum then

to predictions by Hodkinson et al. (1998) that under conditions of climate warming, without increased moisture, desiccation-susceptible Collembola numbers would tend to decrease as habitats became warmer and drier but that numbers of desiccation resistant Acarina would increase. Coulson et al. (1996) found declining numbers of Collembola in cloches after 3 yr, a trend repeated in similar OTC experiments in Antarctica (Convey et al. 2002) and subarctic Sweden (Sjursen et al. 2005). Kennedy (1994), by contrast, found higher numbers of the dominant collembolan Cryptopygus antarcticus and other arthropods in cloche treatments in similar experiments on Signy Island, Antarctica, a fact that might be explained by increased moistureretaining ground vegetation cover in a longer term (8 yr) experiment. Both the overall MANOVA for the OTC core samples and the corresponding GLM ANOVAs for core and pitfall trap samples for individual species showed evidence for a significant among-plots effect as well as the main treatment effect. This suggests that for those species involved, primarily Collembola such as L. lignorum and H. tullbergi , there was an element of patchiness in their abundance operating at the plot scale.

Table 4. Summary of significant differences for individual species among stations along the transect derived from a one-way ANOVA of core sample data using Fisher’s (F) and Tukey’s (T) pair wise comparisons. Overall significant differences along the transect are indicated by critical probability values. Some species are included where pair wise comparisons indicate significance between at least two transect stations but the overall ANOVA is only approaching significance. In these cases the exact probability value is stated. Transect station 1 Species Collembola total L. lignorum H. tullbergi F. quadrioculata A. principalis F. bisetosa T. arctica C. anomia H. reticulata Araneae Enchytraeidae

ECOGRAPHY 29:1 (2006)

2 T

4 3,4 3 2,3,4 4 3,4

3,4

4 1 4 4 4 4 3

4 1 1 4 1

4 4 2 1

4 T

4 1

1,2, 3 1

2,3 1 2 2 1,2,3 2,3

2,3

Overall p B/0.01 0.07 0.15 B/0.001 0.10 0.07 0.10 B/0.01 0.07 0.13 B/0.01

117

warming begins to produce deleterious effects. This could partially explain the numbers both in the plots and along the transect. A second explanation might lie in the increased numbers of predators, particularly prostigmatic and gamasid mites, and host-specific aphid parasitoids within the OTC when compared with controls. Numbers of predatory mites were significantly higher within the OTCs (Fig. 2) and numbers of parasitoids were similarly approaching significance. This points to a delayed build up of predator/parasitoid numbers within the OTCs that was subsequently impacting significantly on the aphid population. It is also notable that treat˚ lesund, were enclosed, thereby ment cloches at Ny-A excluding flying parasitoids. The OTC’s in Endalen were also less efficient at raising the temperature /1.2 degrees compared with 2.8 degrees for the cloches at Ny˚ lesund. A It appears that desiccation susceptible predatory mites may be able to sustain body moisture by feeding on plant sap-sucking aphids within the OTCs, as commonly observed. Altenatively, these active mites may enter the OTC under the plastic wall and replenish declining populations. Physical exclusion may also explain why Symphyta larvae were common outside the OTCs but rare inside. Adult female sawflies are strongly dispersive and tend to fly immediately above the ground surface: the walls of the OTCs may thus form a significant barrier to colonisation. While parasitoids and predators may be exerting a top down effect on the population of some prey species, bottom up effects, induced by warmer conditions, may also be important, either directly or indirectly increasing food availability by changing the rate of microbial release of nutrients within the soil (Hedlund and Ohrn 2000). Several experiments have demonstrated effects of altered nutrient availability on soil invertebrates, including Enchytraeidae, nematodes and Collembola, under conditions of warming (Ruess et al. 1999a, b, Convey and Wynn-Williams 2002, Cole et al. 2002, Sjursen et al. 2005). The effect may also be physical, with improved or reduced vascular plant growth altering the available inter-plant habitat space occupied by surface-active species, such as several Collembola, or restricting the effect of incident UV-B (Convey et al. 2002). Our experiments indicate several significant changes in existing invertebrate community structure that are likely to occur in the short to medium term under conditions of climate warming. It fails inevitably, however, to account for the probability of immigration by more thermophilous species and the extinction of cold adapted species that is likely to occur over the longer term. Nevertheless, despite the obvious limitations, retrospective sampling on an ad hoc basis can provide valuable insights into changes that are occurring within invertebrate communities in response to environmental manipulation. 118

Acknowledgements / We thank Arne Fjellberg for checking the identity of the Collembola species. UNIS provided financial support and logistics for field work.

References Anon. 2003. Circumpolar Arctic vegetation map. Scale 1:7,500,000. Conservation of Arctic Flora and Fauna (CAFF)-Map No. 1. / U.S. Fish and Wildlife Service, Anchorage, AK. Bliss, L. C. and Matveyeva, N. V. 1992. Circumpolar arctic vegetation. / In: Chapin, F. S. et al. (eds), Arctic ecosystems in a changing climate. An ecophysiological perspective. Academic Press, pp. 59 /89. Chapin, F. S. et al. 1995. Responses of arctic tundra to experimental and observed changes in climate. / Ecology 76: 694 /711. Cole, L. et al. 2002. Enchytraeid worm (Oligochaeta) influences on microbial community structure, nutrient dynamics and plant growth in blanket peat subjected to warming. / Soil Biol. Biochem. 34: 83 /92. Convey, P. and Wynn-Williams, D. D. 2002. Antarctic soil nematode response to artificial climate amelioration. / Eur. J. Soil Biol. 38: 255 /259. Convey, P. et al. 2002. Response of antarctic terrestrial microarthropods to long-term climate manipulations. / Ecology 83: 3130 /3140. Couslon, S. J. and Refseth, D. 2004. The terrestrial and freshwater invertebrate fauna of Svalbard (and Jan Mayen). / In: Prestrud, P., Strøm, H. and Goldman, H. V. (eds), A catalogue of the Svalbard terrestrial and marine animals: invertebrates, fishes, birds and mammals. Skrifter 201, Norwegian Polar Inst., Tromsø, pp. 57 /122. Coulson, S. J. et al. 1996. Effects of experimental temperature elevation on high-arctic soil microarthropod populations. / Polar Biol. 16: 147 /153. Coulson, S. J., Hodkinson, I. D. and Webb, N. R. 2003. Microscale distribution patterns in high Arctic soil microarthropod communities: the influence of plant species within the vegetation mosaic. / Ecography 26: 801 /809. Day, T. A. et al. 1999. Growth and reproduction of Antarctic vascular plants in response to warming and UV radiation reductions in the field. / Oecologia 119: 24 /35. Elvebakk, A. 1997. Tundra diversity and ecological characteristics of Svalbard. / In: Wielgolaski, F. E. (ed.), Polar and Alpine tundra. Ecosystems of the world 3. Elsevier, pp. 347 / 359. Fjellberg, A. 1994. The Collembola of the Norwegian arctic islands. / Medd. Norskpolarinst. 133: 1 /57. Galen, C. and Stanton, M. L. 1995. Responses of snowbed plant-species to changes in growing-season length. / Ecology 76: 1546 /1557. Gerber, J. D., Baltisberger, M. and Leuchtmann, A. 2004. Effects of a snowmelt gradient on the population structure of Ranunculus alpestris (Ranunculaceae). / Bot. Helv. 114: 67 /78. Hedlund, K. and Ohrn, M. S. 2000. Tritrophic interactions in a soil community enhance decomposition rates. / Oikos 88: 585 /591. Heegaard, E. 2002. A model of alpine species distribution in relation to snowmelt time and altitude. / J. Veg. Sci. 13: 493 /504. Hodkinson, I. D. et al. 1996. Can high Arctic soil microarthropods survive elevated summer temperatures? / Funct. Ecol. 10: 314 /321. Hodkinson, I. D. et al. 1998. Global change and Arctic ecosystems: conclusions and predictions from experiments with terrestrial invertebrates on Spitsbergen. / Arct. Alp. Res. 30: 306 /313. ECOGRAPHY 29:1 (2006)

Hollister, R. D. and Webber, P. J. 2000. Biotic validation of small open-top chambers in a tundra ecosystem. / Global Change Biol. 6: 835 /842. Inouye, D. W., Saavedra, F. and Lee-Yang, W. 2003. Environmental influences on the phenology and abundance of flowering by Androsace septentrionalis (Primulaceae). / Am. J. Bot. 90: 905 /910. Kennedy, A. D. 1994. Simulated climate-change / a field manipulation study of polar microarthropod community response to global warming. / Ecography 17: 131 /140. Kennedy, A. D. 1995. Temperature effects of passive greenhouse apparatus in high-latitude climate-change experiments. / Funct. Ecol. 9: 340 /350. Leinaas, H. P. 1978. Sampling of soil microarthropods from coniferous forest podzol. / Norw. J. Entomol 25: 57 /62. Marion, G. M. et al. 1997. Open-top designs for manipulating field temperature in high-latitude ecosystems. / Global Change Biol. 3 (Suppl. 1): 20 /32. Molau, U. and Mølgaard, P. 1996. ITEX manual. / Danish Polar Centre, Copenhagen. Rae, D. A. 2003. Plant and invertebrate community responses to species interaction and microclimatic gradients in alpine and Arctic environments. / Ph.D. thesis, Dept of Biology, Norwegian Univ. of Science and Technology. Roy, B. A., Gusewell, S. and Harte, J. 2004. Response of plant pathogens and herbivores to a warming experiment. / Ecology 85: 2570 /2581. Ruess, L., Michelsen, A. and Jonasson, S. 1999a. Simulated climate change in subarctic soils: responses in nematode species composition and dominance structure. / Nematology 1: 513 /526. Ruess, L. et al. 1999b. Simulated climate change affecting microorganisms, nematode density and biodiversity in subarctic soils. / Plant Soil 212: 63 /73. Ryan, J. K. 1981. Invertebrate faunas at IBP tundra sites. / In: Bliss, L. C., Heal, O. W. and Moore, J. J. (eds), Tundra ecosystems a comparative analysis. Cambridge Univ. Press, pp. 517 /539.

Download the appendix as file E 4366 from B/www.oikos.ekol.lu.se/appendix /.

ECOGRAPHY 29:1 (2006)

Sinclair, B. J. and Sjursen, H. 2001. Terrestrial invertebrate abundance across a habitat transect in Keble Valley, Ross Island, Antarctica. / Pedobiologia 45: 134 /145. Sjursen, H., Michelsen, A. and Jonasson, S. 2005. Effects of long-term soil warming and fertilisation on microarthropod abundances in three sub-arctic ecosystems. / Appl. Soil Ecol. 30: 148 /161. Stanton, M. L., Galen, C. and Shore, J. 1997. Population structure along a steep environmental gradient: consequences of flowering time and habitat variation in the snow buttercup, Ranunculus adoneus. / Evolution 51: 79 / 94. Stenstro¨m, A. and Jo´nsdo´ttir, I. S. 2005. Effects of simulated climate change on phenology and life history traits in Carex bigelowii . / Nord. J. Bot., in press. Strathdee, A. T. and Bale, J. S. 1993. A new cloche design for elevating temperature in polar terrestrial ecosystems. / Polar Biol. 13: 577 /580. Strathdee, A. T. et al. 1993. Effects of temperature elevation on a field population of Acyrthosiphon svalbardicum (Hemiptera, Aphididae) on Spitsbergen. / Oecologia 96: 457 /465. Strathdee, A. T. et al. 1995. Climatic severity and the response to temperature elevation of Arctic aphids. / Global Change Biol. 1: 23 /28. Totland, O. and Alatalo, J. M. 2002. Effects of temperature and date of snowmelt on growth, reproduction, and flowering phenology in the arctic/alpine herb, Ranunculus glacialis. / Oecologia 133: 168 /175. Van Wijk, M. T. et al. 2003. Interannual variability of plant phenology in tussock tundra: modelling interactions of plant productivity, plant phenology, snowmelt and soil thaw. / Global Change Biol. 9: 743 /758. Walker, M. D., Ingersoll, R. C. and Webber, P. J. 1995. Effects of interannual climate variation on phenology and growth of two alpine forbs. / Ecology 76: 1067 /1083. Webb, N. R. et al. 1998. The effects of experimental temperature elevation on populations of cryptostigmatic mites in high Arctic soils. / Pedobiologia 42: 298 /308.

Subject Editor: John Spence.

119

Ecography 30: 193 208, 2007 doi: 10.1111/j.2007.0906-7590.04818.x Copyright # Ecography 2007, ISSN 0906-7590 Subject Editor: John Spence. Accepted 5 February 2007

From forest to pasture: an evaluation of the influence of environment and biogeography on the structure of dung beetle (Scarabaeinae) assemblages along three altitudinal gradients in the Neotropical region Federico Escobar, Gonzalo Halffter and Lucrecia Arellano F. Escobar (federico.escobarf@gmail.com), G. Halffter and L. Arellano, Dept de Biodiversidad y Comportamiento Animal, Inst. De Ecolog覺織a A.C., Apartado Postal 63, 91000 Xalapa, Veracruz, Mexico. (Present address of F. E.: Dept of Zoology and Entomology, Univ. of Pretoria, 0002 Pretoria, South Africa.)

The objective of this study is to evaluate the effect of environmental (associated with the expansion of cattle ranching) and biogeographical factors on the diversity of dung beetle (Scarabaeinae) assemblages along three altitudinal gradients in the Neotropical region. One gradient is located in the Mexican Transition Zone, on the Cofre de Perote mountain, the other two are in the northern Andes (the Chiles Volcano and the R覺織o Cusiana Basin). For the three gradients, the number of species decreased as altitude increased. On the Cofre de Perote, regardless of altitude, the number of species and of individuals was similar in both forest and pasture, while species composition was different between habitats. On this mountain, species turnover in pastures was characterized by the addition of new species as altitude increased. In the northern Andes, species diversity was always greater in the forest than in the pasture, and species turnover between habitats was notably influenced by species loss with increasing altitude. As such the pasture fauna of the northern Andes was an impoverished derivative of the fauna present in the forests at the same altitude characterized by species of Neotropical affinity with a limited capacity for colonizing open, sunnier habitats. The opposite occurs in the areas used by cattle on the Cofre de Perote. This habitat has its own fauna, which is mainly comprised of Holarctic and Afrotropical species adapted to the prevailing environmental conditions of areas lacking arboreal vegetation. These results suggest that the impact on beetle communities caused by human activities can differ depending on the geographic position of each mountain and, particularly, the biogeographical history of the species assemblage that lives there.

A decrease in species richness and changes in the composition of the flora and fauna with increasing altitude have frequently been described in the literature (Huston 1994). However, it is not possible to describe a general relationship between changes in diversity and increasing altitude (Rahbek 1995), because the ecological factors (current environmental conditions, e.g. land use) and biogeographical factors differ in their relative effect depending on the mountain system being studied (Brown 2001). This is emphasized by the fact that we still do not understand how these factors vary and interact as altitude changes. For this reason comparative studies using the same taxonomic group

along altitudinal gradients on different mountains (such as this one), and those that use several taxonomic groups along the same altitudinal gradient of the same mountain are very useful for explaining this co-variation (Lomolino 2001). Studying the altitudinal variation in dung beetle assemblages on different mountains around the world, Lobo and Halffter (2000) proposed two different biological and interrelated processes to explain the conformation of the biota on mountains, the patterns of species richness and variations in composition: horizontal colonization by elements originating from lineages inhabiting higher altitudes, and vertical

193

colonization by lineages from surrounding lower lands at the same latitude. The relative effect of both processes depends on the orientation and location of the mountains, and on their degree of isolation and biogeographical history, as these characteristics greatly influence the refuge and ‘‘corridor’’ capacity of mountain areas (Lobo and Halffter 2000). Two hypotheses for explaining the changes in diversity with altitude emerge from the relative influence of these processes: 1) the mountain fauna is composed of a lesser number of phylogenetically related species relative to the fauna of lower altitudes (colonization vertical model), and 2) the mountain fauna is composed of elements with different evolutionary histories and origins compared to the lowlands fauna (horizontal colonization model). According to the first hypothesis, we would expect to find mountains that, owing to their geographic isolation, limited extension or recent geological formation, have yet to accumulate their own species assemblages. Therefore we would expect species substitution to be slow and species richness to notably decrease with increasing altitude. This would be the consequence of the environmental restrictions imposed by high altitudes on the fauna from warmer altitudes, especially in tropical regions (Janzen 1967). This appears to be the case for the northern Andes (Escobar et al. 2005) and southeastern Asia (Hanski 1983). In contrast, in the horizontal colonization model (second hypothesis), geographical and historical factors become especially relevant. This situation has been described for the mountains located in the central Iberian Peninsula (Martin-Piera et al. 1992), those of southern France (Errouissi et al. 2004) and of the Mexican Transition Zone (MTZ, Halffter 1987). Mountains that owing to their geographic location and orientation have served as refuge and speciation areas for lineages from more northern regions during the climatic changes of the Plio-Pleistocene. They consequently host a fauna adapted to cold environment and annual fluctuations in climate (Halffter 1987, Martin-Piera et al. 1992). When horizontal colonization is the main process governing the establishment of mountain fauna, one would expect to find greater phylogenetic diversity, fast species substitution and a less pronounced decrease in species richness with increasing altitude (Lobo and Halffter 2000). Our ability to understand contemporary biogeographical patterns also relies on our understanding of human impact, and specifically on how human impact affects natural ecosystems, modifying the spatial distribution of species, community and population structure throughout large geographic areas (Lomolino and Perault 2004). One such activity is cattle ranching in the mountains of the Neotropical region, which has resulted in a continuous increase in the area covered by pasture and is responsible for the homogenization of the

194

mountain landscape (Kappelle and Brown 2001). During this process, sunnier open areas are created and the environmental conditions there become much more severe. Additionally, the quantity of dung, mainly from cattle, is greater, which has been shown to modify the structure of the dung beetle assemblages at the local and landscape levels in the different tropical and subtropical regions of the world (Halffter 1991, Nichols et al. in press). From these studies, the general conclusion is that vegetation cover determines species abundance and richness of this group of beetles. The association of the species with a given type of habitat appears to be related to its micro-climatic requirements (i.e. temperature, relative humidity, light intensity) and to its close dependence on mammalian dung for feeding and reproduction (Halffter and Matthews 1966). It has also been observed, however, that the preferences of species for certain habitats, both natural and anthropogenic, varies with altitude in different ways depending on the geographic position of the mountain and, therefore, the biogeographical affinity of the fauna that live there (Halffter et al. 1995, Davis et al. 1999, Romero-Alcaraz and Avila 2000, Errouissi et al. 2004). From the above, it might be expected that the effects of expanding cattle ranching on dung beetle assemblage diversity would be different in each mountain region. To evaluate this prediction we studied the diversity patterns of dung beetles along three altitudinal gradients that are ecologically similar but have different biogeographical histories. One is in the MTZ and two are in Colombia, located on opposite slopes of the northern Andes. In both mountain systems the main change in the landscape is a notable conversion of the original forest into cattle pasture and, to a lesser degree, agricultural crop fields. Given that the pastures are a relatively new type of vegetation, and the mountains vary with respect to the influence of the vertical vs horizontal colonization processes, we expect: a) the fauna of the pastures in the northern Andes to be a rarefied subset of the species found in the surrounding forested areas, comprised of species belonging to genera of Neotropical affinity with a limited capacity for colonizing open areas (Amat et al. 1997, Escobar 2004), and b) the fauna of the MTZ pastures to be a mixture of species of wide ecological tolearance, capable of leaving the forested sites, and species from nonNeotropical lineages adapted to the prevailing conditions of the areas without any arboreal cover (Halffter et al. 1995, Arellano and Halffter 2003). In this study we address the following questions: 1) how does dung beetle diversity changes with respect to altitude along three altitudinal gradients? 2) How do changes in the dung beetle assemblages occur in forests and pastures at each altitude, along the altitudinal gradient, and according to the geographical position of the mountain? 3) How do environmental and biogeographical factors

influence any differences detected (points 1 and 2) between the MTZ gradient and the two northern Andean gradients?

coffee plantations and at high altitudes, seasonal agriculture (wheat, oats and potato) and intensive dairy cattle ranching (Challenger 1998).

Materials and methods

Northern Andes, the Chiles Volcano and the Rı´o Cusiana

MTZ, Sierra Madre Oriental-Sistema Volcańico Transversal Halffter (1987) defined the MTZ as a complex and varied region extending from northern Mexico to southern Nicaragua in which Nearctic and Neotropical biota overlap. To the north, there are Nearctic elements that gradually decrease towards the south. The northern lineages have dispersed through the mountains, which in the MTZ have a generally N-S orientation, facilitating horizontal colonization. In contrast, the coastal plains and tropical lowlands are the penetration route for Neotropical elements. This double occupation of the territory that occurs latitudinally in the MTZ has an altitudinal equivalent in the mountains: higher altitudes are occupied by lineages of northern affinity, lower altitudes by lineages of neotropical affinity and intermediate altitudes are characterized by an overlap of these two lineages and strong in situ speciation, particularly of those lineages with a long evolutionary history in the zone. The altitudinal gradient is located on the eastern slope of the Cofre de Perote volcano in the state of Veracruz (19819? 19839?N, 96824? 97812?W; Fig. 1) and on the eastern end of the Sistema Volcańico Transversal where it meets the Sierra Madre Oriental. The Sierra Madre Oriental is comprised of a series of NNW-SSE folds, with an average width of 80 to 600 km each. It starts in the north on the Texas platform and is interrupted by the Sistema Volcańico Transversal. The latter is a complex mountain chain that runs W-E and is 950 km long, and 50 150 km wide. It is considered one of the youngest mountain systems in the country (Pliocene-Quaternary, 2 3 million years BP; Ferrusquıá-Villafranca 1993). This altitudinal gradient covers different types of vegetation: tropical deciduous forest in the lowlands (B1000 m a.s.l.; temperature: 228 248C; annual precipitation: 1500 2000 mm), cloud forest, oak forest and pine-oak forest at intermediate altitudes (1000 2000 m a.s.l.; 128 188C; 2000 3000 mm), pine forest and oyamel fir forests at 2000 3000/3500 m a.s.l. (58 128C; 1000 1800 mm). Above this altitude, there are high natural pastures ( 28 58C; B1200 mm). Land use varies with altitude. In the lowlands extensive cattle ranching is the main land use, but there are also irrigation agriculture, sugar cane and fruit crops (orange, mango and tamarind). At intermediate altitudes there are corn crops, dairy farming and especially

The northern Andes belong to a mountain system that extends from northern Peru (the Huancabamba Depression) to Venezuela. In Colombia, the Andes represent an enormous mass of mountains that occupies ca 30% of the country, and diverges into three branches or ranges: the Western, the Central and the Eastern, each extending in a general south-north direction. The Western Range is considered one of the oldest ranges of the Colombian Andes (Oligocene, 38 million years BP). This range is ca 650 km long and is the narrowest of the three at B50 km wide and altitudes no higher than 4800 m a.s.l. The Eastern Range, with a length of 1200 km, is 200 km at its widest and reaches altitudes 5500 m a.s.l. It is considered the main mountain chain of the northern Andes. It originated in the Miocene (18 million years BP) and its final formation occurred in the PliocenePleistocene (2.5 million years BP; van der Hammen and Hooghiemstra 2001). The slopes of the altitudinal gradients studied in the Colombian Andes have different aspects. The Chiles Volcano (0810? 1817?N, 78815? 77811?W; Narinõ Department) is on the western slope of the Eastern Range, facing the Pacific Plain. The other gradient is in the Rıó Cusiana Basin (5826? 5823?N, 72841? 728 42?W, Boyaca´ Department), on the eastern slope of the Eastern Range, facing the Amazonia-Orinoquia region. The following vegetation types are found in the northern Andes: tropical lowland forest (up to 1000/ 1250 m a.s.l.; 228 268C; 4000 8000 mm), tropical sub-Andean forest (1000/1250 2000/2300 m a.s.l.; 168 228C; 2000 4000 mm); high Andean forest (2000/2300 3200/3600 m a.s.l.; 68 128C; 1000 1500 mm) and pa´ramos (above 3200/3900 m a.s.l.; 38 68C; 500 1000 mm). The limits of the vegetation zones vary depending on the topography and local climate, as indicated by the values given in parentheses (van der Hammen 1995). On the Chiles Volcano, in addition to extensive cattle ranching, the lowlands are mostly used for cultivating African palm, bananas and corn. At intermediate altitudes there are coffee plantations, sugar cane and cattle ranching; and the highest altitudes (above 2500 m a.s.l.) are used for dairy farming and potato crops. In the Rıó Cusiana basin, human activites have reduced the forested areas to remnants of varying sizes along the entire altitudinal gradient. The lower part of the mountain is dominated by pastures and small parcels with corn crops, intermediate altitudes are used for banana and sugar cane

195

Forest

Pasture

(a)

8 4

30 m 900 m 1800 m

450 m 1360 m 2000 m

2340 m

3300 m

8 4

Cumulative number of species

0 20

200

400

600

800

100

200

300

(b)

12 8

8 4

50 m

520 m

1000 m

1350 m

1800 m

0 0 20

200

400

600

800

100

150

200

(c)

450 m

900 m

1250 m

1450 m

1750 m

2000 m

2250 m 0

0 0

100

200

300

400

100

Cumulative number of individuals Fig. 1. Smoothed species accumulation curves using the number of individuals collected as a substitute for the sampling effort applied at each forest and pasture site along each altitudinal gradient: (a) Cofre de Perote; (b) Chiles Volcano (except the pasture at 1000 m a.s.l.) and, (c) R覺織o Cusiana (except the pasture at 2000 m a.s.l.). For those sites where abundance values were 52 individuals it was not possible to estimate species richness, such as at 2600 and 3300 m a.s.l. for the Chiles Volcano.

crops, and extensive cattle ranching, and finally the highest altitudes are mostly used for dairy farming and potato, barley and wheat crops. In both the MTZ and the northern Andes mountain systems, the transformation of the forest into pastures for cattle began with the arrival of the Spaniards 500 yr ago, and has dramatically modified native ecosystem throughout tropical America (Murgueitio 2003). It is currently estimated that ca 33% (602 million ha) of this

196

region is covered by permanent cattle pastures (Anon. 2002). In the mountainous regions of the Americas, the most conspicuous changes to the landscape occurred at the beginning of the 20th century, particularly form the 1950s onwards and the transformation process continues at an alarming pace to this day (Challenger 1998, Kapelle and Brown 2001). In these mountains the cattle management systems vary widely and largely depend on the climatic and topographic conditions. They also vary

in size and range from 1 to 500 ha (Murgueitio 2003). In the lowlands there are 1 8 cows ha 1. At intermediate altitudes there is greater variation (1 15 ha 1) and at very high sites where natural pastures are used, cattle density is 5 8 cows ha 1 (Anon. 2002). Sampling Sampling was carried out in two contrasting habitat types occurring along each of the three altitude gradients: forested areas and induced pastures used for cattle. Beetles were caught with buried pitfall traps (top flush with the soil) with two types of bait (excrement and carrion). The bait was wrapped in muslin and suspended from a wire right above the trap. The volume of the traps was ca 1000 ml (13 mm deep and 11 mm in diameter) and a mixture of water and detergent was placed inside to prevent caught beetles from leaving. On the Cofre de Perote, 16 sites were sampled at eight altitudes between 50 and 3000 m a.s.l., from May to October 1994. At 450 m a.s.l., the sites were sampled in April and May of 1993. At each site, a line of traps was set with alternating bait of fresh excrement (human and cattle mixed) and decomposing squid. Eight to 17 traps were set per forest site (mean9SD: 12.893.0) and 15 traps were set in the pastures (12.591.9). Traps were placed 25 30 m apart and left in the field for one day and one night (24 h) before being collected. On the Chiles Volcano, 13 sites were sampled at seven altitudes between 50 and 3300 m a.s.l. during April and September 1993. At each site we placed a line of 12 traps with 25 30 m between traps. Similarly, in the Rı´o Cusiana basin, 13 sites were sampled at seven altitudes between 450 and 2500 m a.s.l. during May and June 1997. At each site we placed a line of 10 traps, with the traps 25 30 m apart. In both cases, traps were alternately baited with fresh human excrement and decomposing meat. Baited traps were left in place for two days and two nights (ca 48 h) at each site before being collected. For the gradients in the northern Andes it was not possible to collect from the pastures at 1000 m a.s.l. (on the Chiles Volcano) or at 2000 m a.s.l. (Rı´o Cusiana) owing to the lack of sites appropriate for sampling. Data analysis Given that sampling effort was different in each mountain region, we used accumulation curves with the number of individuals collected, rarefaction based on individuals, as a measure of sampling effort. For each site, the number of species observed was obtained995% confidence interval (Colwell et al. 2004). As an estimate of species richness, we used the Michaelis-Menten equation (MM), one of the curvi-

linear asymptotic functions most commonly used in the evaluation of diversity inventories and adequate for a small number of samples (Colwell and Coddington 1994). The smoothed accumulation curves were obtained by repeated random reordering (500 times) of the samples using v. 7.5.0 of EstimateS program (Colwell 2005). Analyses were carried out for three levels of comparison: a) total diversity (Gamma diversity, g) defined in this case as the cumulative number of species by habitat type along each altitudinal gradient, b) local diversity between habitats (Alpha diversity, a) along each gradient, using the total number of species recorded at each site (St) and the mean number of species per trap (Sm) and, c) species turnover (Beta diversity, b). To compare the process of g diversity accumulation in each habitat type along each gradient, we calculated the slope of the linear regression of altitude (independent variable) against the observed cumulative number of species (dependent variable). We used a Student’s t to test whether the slopes (forest vs pasture) were significantly different between habitats (Zar 1996). In order to determine the relationship between local species richness (quantified as St and Sm) and habitat along each gradient, we used an analysis of covariance (ANCOVA) with altitude as the covariable. For all cases, the model fit to the data Y m Habitat Altitude Habitat Altitude o. For St, we obtained the complete model assuming a Poisson distribution of errors (link function Log; Crawley 2002). For Sm, error distribution was assumed to be normal. In both cases, the model was verified by examining the standardized residuals vs the fit values, in addition to the graphical distribution of errors. Species turnover along each gradient was analyzed in two ways: a) between adjacent altitudes for each habitat type along each altitudinal gradient and, b) between habitat types (forest vs pasture) at each altitudinal level. Wilson and Shmida’s (1984) index (bt) was used: (bt) (a c)/(2a b c), where a is the number of species found at two sites and b and c are the number of species lost and gained in each comparison. Values of bt vary between 0 and 1, with 1 indicating the greatest degree of dissimilarity between sites. This index produces results similar to those of other indices of b diversity and is one of the most recommended since it provides a direct expression of species turnover when the samples are arranged along an environmental gradient and because it is independent of a diversity (Wilson and Shmida 1984). Since the indices of b diversity do not reveal whether turnover values are a product of the loss or gain of species, and in order to understand the relative influence of each process, we calculated the number of species lost and gained for each comparison.

197

The abundance distribution of species in each habitat was compared using range-abundance curves. These curves can also be used to describe the changes in community structure (Magurran 1988). In order to determine how different observed changes are from random differences in the structure of the beetle community when forest is replaced by pasture at each altitude, we used the test developed by Solow (1993, available in the program developed by Henderson and Seaby 2002). This randomization test can be used together with any other species abundance-based measure of community structure. We used Simpson’s index (D) (in its reciprocal form 1/D) to evaluate the change in diversity between sites: D ani (ni 1)=[N(N 1)]; where ni is the number of individuals of species i and N ani : Simpson’s index represents the probability that two individuals randomly selected from a sample belong to different species, and in its reciprocal expression is a measure of dominance (Magurran 1988). In Solow’s test (1993), the observed change (d) in 1/D is compared with the values obtained from 10000 random partitions of the total sample of individuals in a set of samples similar in size to the observed. The statistical significance of the observed value of d can be evaluated by its position relative to those of the ordered values of d obtained randomly. In this test, the value of probability for a two-tailed test is given by the proportion of partitions where the simulated value ½d½ is greater than the observed value ½d½.

Results On the Cofre de Perote, we captured a total of 3245 individuals belonging to 40 species. The number of species and individuals caught in the forest was similar to that of the pasture (Table 1). For this mountain the rate of accumulation of g diversity was not different between habitats (bforest 8.1 species/1000 m, bpasture 7.5 species/1000 m; t 1.62, DF 12, p 0.13). In contrast, the dung beetle diversity in the northern Andes was remarkably different between habitats. On the Chiles Volcano 1746 individuals belonging to 37 species were caught: 89% of the species (87% of individuals) were from the forest and 57% of the species (13% of individuals) were from the pasture (Table 1). Although there was no significant difference in the rate of accumulation of g diversity with increasing altitude, the value was higher in the forest than in the pasture (bforest 5.4 species/1000 m, bpasture 3.4 species/1000 m; t 1.79, DF 9, p 0.10). At Rı´o Cusiana, 1518 individuals belonging to 49 species were caught. Of these 88% (90% of individuals) were caught in the forest and 45% (10% of individuals) were caught in the pasture (Table 1)

198

while g diversity accumulated at a greater rate in the forest than in the pasture (bforest 17.2 species/1000 m, bpasture 9.3 species/1000 m; t 7.36, DF 9, p B 0.0001). In spite of the limited variation in the total number of tribes, genera and species between mountains, on the Cofre de Perote, 58% of species and 53% of the individuals caught were Neotropical. On this mountain the species belonging to genera with an Afrotropical affinity (Digitonthophagus ) were captured more often in pastures, while those of Holarctic affinity (Copris and Onthophagus ) were captured equally in the forest and in the pasture (Table 1). In contrast, in the northern Andes ca 92% of species and 85% of the individuals belonged to genera of Neotropical affinity and these were caught more frequently in forested areas (Table 2). In these mountains, no genus was clearly dominant in the pastures and only two species (Anisocanthon villosus and Canthon sp. 1) were exclusive to areas used for cattle located at low altitudes (Appendix 1). The complete list of species by habitat along each altitudinal gradient is given in Appendix 1. Species richness reliability Visual comparison of the species accumulation curves for the forest and the pasture at each altitude indicate that at the Cofre de Perote most of the sites reached an asymptote (Fig. 1). Estimated species richness values (MM) indicate that a large proportion of the species ( 80%) present at each site were captured (Table 2). In contrast, for the two northern Andes gradients, regardless of altitude the species accumulation curves for the pastures rarely reached the asymptotic phase (Fig. 1), due in part to the low number of individuals captured and to the high dominance of a few species. In this environment, the percentage of species captured was 70% in only a few cases and the estimated species richness values were highly variable, ranging from 37 to 88% (Table 2). On the other hand, at the forest sites the species accumulation curves were clearly asymptotic (Fig. 1b, c), and the proportion of species captured at each site was 76 95% (Table 2). The analysis of the entire gradient by habitat type indicates that at the Cofre de Perote 90% of the species present in each type of habitat were captured. For the northern Andes, and similar to the results found at the sites level, the proportion of species captured was lower in the pasture than in the forest (Table 2). The analysis for each mountain shows that between 88% (Rı´o Cusiana) and 95% (Cofre de Perote) of all the species present on each gradient were captured (Table 2).

Table 1. Composition at the levels of tribe and genus: number of species and individuals (in parentheses) captured in the forest and pasture along each altitudinal gradient. Biogeographical affinity of each genus: AFR Afrotropical, HOL Holarctic, NEO Neotropical. *Endemic to the Neotropical region. The observed species richness (Sobs9995% CI) was calculated using the Analytical formula proposed by Colwell et al. (2004); a indicates the proportion of species observed relative to Michaelis-Menten (MM) that was used as an estimate of expected richness. Tribe

Genus

Cofre de Perote Forest

Canthonini

Coprini Dichotomiini

Eurysternini Phanaeini

Onthophagini Sysiphini Total number of tribes Total number of genera Total number of species (Sobs995%CI) Michaelis-Menten (MM) Estimate(%)a Total number of individuals

Anisocanthon NEO* Canthon NEO Cryptocanthon NEO* Deltochilum NEO* Scybalocanthon NEO* Copris HOL Ateuchus NEO* Canthidium NEO* Dichotomius NEO* Ontherus NEO* Scatimus NEO* Uroxys NEO* Eurysternus NEO* Coprophanaeus NEO* Phanaeus NEO Oxysternon NEO* Sulcophanaeus NEO* Digitonthophagus AFR Onthophagus HOL Sisyphus AFR/HOL

Pasture

Chiles Volcano Both

Forest

Pasture

R覺織o Cusiana Both

7 (723)

6 (321)

7 (1044)

2 (3)

2 (120)

3 (123)

3 (287)

2 (36)

3 (323)

1 (8)

3 (89)

3 (97)

4 (285) 1 (164) 1 (4)

2 (5) 1 (23) 1 (2)

4 (290) 1 (187) 1 (6)

1 (7) 3 (77) 1 (27)

1 2 1 1

1 3 1 1 1 1 2 3

3 (29) 2 (6)

2 (4) 3 (202) 3 (259)

(31) (13) (54) (9)

1 (45) 2 (30) 2 (16)

11 (608) 1 (21) 6 11 3393.0 36.2 91.1 1849

1 (2) 1 (33) 2 (4) 1 (53) 10 (751) 6 12 3194.4 34.3 90.4 1396

(38) (90) (81) (9) (45) (2) (63) (20)

1 (53) 12 (1359) 1 (21) 7 14 4092.7 42.2 94.8 3245

2 (4) 3 (173) 3 (253) 6 4 1 1 2 2

(400) (29) (5) (20) (21) (86)

1 (4)

7 4 1 1 2 2

3 (72)

2 (12)

3 (84)

6 14 3592.8

6 11 2195.4

6 14 3792.7

38.2 91.6 1519

2 (5) 4 (19) 1 (2)

27.6 76.0 227

(405) (48) (7) (20) (21) (90)

40.5 91.3 1746

Forest 3 (197) 1 (46) 5 (100)

Pasture 1 (5) 1 (4) 1 (1)

1 4 1 5

(5) (201) (46) (101)

1 10 6 2

(1) (129) (210) (279)

5 5 1 3

(63) (179) (3) (30)

1 10 6 2

(1) (129) (193) (231)

3 (17) 2 (48)

5 4 1 2

(56) (153) (2) (26)

3 4 1 2

(7) (26) (1) (4)

1 (30) 2 (199) 5 13 4393.8 48.7 88.3 1363

Both

1 (30) 4 (42) 5 10 2293.9 27.9 78.8 155

4 (241) 5 14 4993.8 55.4 88.4 1519

199

Table 2. Observed and estimated species richness (forest/pasture) in each of the sites along each altitudinal gradient. *Denotes those sites where the estimates were lower than 80%. For the Chiles Volcano at 2600 m a.s.l., it was not possible to calculate the estimated value of species richness because abundance was52 individuals. Mountain/elevation (m a.s.l.)

Forest/pasture No. individuals

Cofre de Perote 50 450 900 1340 1860 2000 2340 3000 Chiles Volcano 50 520 1000 1350 1800 2600 3300 R覺織o Cusiana 450 900 1250 1500 1750 2000 2500

Estimate(%)

306/226 721/125 389/250 102/236 57/47 44/186 216/240 14/68

1093.3/1192.5 1593.5/1094.5 1092.4/1494.0 1090.9/692.3 591.2/593.0 491.3/892.4 391.2/390.0 290.0/392.2

11.6/13.4 16.9/11.7 11.1/18.2 11.3/6.6 5.7/5.9 4.2/8.3 3.5/3.6 2.8/3.2

86.2/89.1 88.7/85.4 90.1/76.2* 88.5/90.9 87.7/84.7 95.2/96.4 85.7/83.3 71.4*/93.7

216/181 85/21 357/ 684/14 175/6 2/1 0/4

1791.5/992.4 1295.1/693.1 1291.8/ 1390.0/592.3 891.3/492.9

19.5/10.2 14.5/8.8 12.3/ 13.7/8.5 8.8/10.7

87.2/88.2 82.7/68.1* 97.5/ 94.9/58.8* 90.9/37.3*

13/17 338/33 280/57 187/20 309/20 53/ 65/16

Variation in Alpha diversity The number of species decreased with increasing altitude on the three gradients (Fig. 2). However, the pattern of change in species richness at the local level (St and Sm) at each habitat type was different for each mountain system. On the Cofre de Perote values of St were similar for the forest and the pasture regardless of altitude (mean9SD: forest 7.6590.81, pasture 7.7590.80, p 0.92; Table 3), and there were even some altitudes for which St was greater in the pasture (Fig. 2a). On this mountain, altitude explained 74% of the variation and habitat type explained 1%. In contrast, on the Chiles Volcano St values were always higher for the forest than for the pasture. (forest 9.090.61, pasture 4.390.66, p 0.001; Fig. 2b). On this mountain, altitude explained 62% of the variation and habitat type accounted for 17% (Table 3). Similarly, at R覺織o Cusiana, St was also consistently greater in forest than in pasture (forest 10.5891.4, pasture 5.891.53, p 0.002; Fig. 2c). The model explained 48% of the total variation, much of which was associated with habitat type (40%), while altitude accounted for the remaining 8% (Table 3). The lack of fit of the model, particularly for the forest, is a result of the increase in St at intermediate altitudes. The analyses

200

Sobs.995% IC

0/190.0 1092.5/390.9 1691.3/692.4 1294.8/1195.1 1392.7/391.2 1492.7/391.2 591.2/ 591.3/490.9

/1.5 11.2/7.16 18.3/13.7 14.5/15.6 14.9/3.4 18.4/3.4 5.5/ 5.9/5.7

/66.6* 89.3/41.9* 87.4/43.8* 82.7/70.5* 87.2/88.2 76.1*/88.2 90.9/ 84.7/70.1*

did not detect significant differences in the three altitudinal gradients regarding the rate at which species are lost as altitude increases for either type of habitat (Table 3). The analysis using mean species richness (Sm) produced a pattern similar to that described for St (Table 3). Variation in Beta diversity Species turnover patterns between adjacent levels were different on each mountain according to habitat type (Fig. 3). On the Cofre de Perote for the forest the values of bt reached a maximum between 900 and 1340 m a.s.l. and then slowly decreased with increasing altitude (bt: 0.9 0.6). In this habitat the values of bt were influenced by the loss of species (Fig. 3a). In contrast, for pastures species turnover increased rapidly up to 1860 m a.s.l. (bt: 0.45 0.82) and stayed constant above 2000 m a.s.l. For the pasture of the Cofre de Perote, values of bt were the result of gaining species at certain altitudes (Fig. 3b). The comparison between forest and pasture habitats for each altitude shows that although species turnover between habitats tends to decrease with increasing altitude (bt: 0.52 0.2), the values of bt reflect the gain of species, particularly above 1800 m a.s.l. (Fig. 3c).

Table 3. Summary of the ANCOVA results. For St (total number of species per site) a Generalized linear model (GLM) was used with a Poisson error distribution (link function Log). Deviance values (ca x2) are given as a measure of the model’s fit. Sm denotes the mean number of species per trap. In both cases, the fit model was Y m Habitat Altitude Habitat Altitude o. * pB0.05*; pB0.01**; pB0.001***; ns not significant.

Factor

St Deviance (x2 approx.)

Sm F

Cofre de Perote Habitat Altitude Habitat Altitude Error

1 1 1 12

0.03ns 26.73*** 0.29ns 8.76

1.69ns 17.39*** 1.0ns

Chiles Volcano Habitat Altitude Habitat Altitude Error

1 1 1 9

10.10*** 38.32*** 0.09ns 12.85

8.69** 9.65** 1.63ns

Rı´o Cusiana Habitat Altitude Habitat Altitude Error

1 1 1 9

12.53*** 2.16ns 0.75ns 15.75

16.40** 2.75ns 2.67ns

(a)

Forest: a = 13.3; b = -3.9 Pasture: a = 12.7; b = -3.3

18 16 14

8 6 4 2 0 0 20

500

1000

1500

2000

(b)

2500

3000

3500

Forest: a = 17.0; b = - 5.3 Pasture: a = 8.2; b = - 3.4

Number of species

16 14 12 10 8 6 4 2 0 0

500

1000

1500

2000

(c)

2500

3000

3500

Forest: a = 16.1; b = -3.7 Pasture: a = 5.9; b = -0.4

18 16 14 12 10 8 6 4 2 0 0

500

1000

1500

2000

2500

3000

Altitude (m a.s.l.)

Fig. 2. Variation in total species richness (St) in each type of habitat as elevation increases: (a) Cofre de Perote, (b) Chiles Volcano, (c) Rı´o Cusiana. The black dots represent forest sites and the open rectangles represent the pastures. Lines indicate the fitted curve (forest continuous line; pasture dotted line).

Species turnover between adjacent altitudes on the Chiles Volcano, both in the forest and the pasture, show a similar pattern, decreasing at intermediate

altitudes and then reaching maximum values at the upper end of the gradient, although for the pasture the increase was more gradual (bt forest: 0.28 1.0; bt pasture: 0.55 1.0). In both habitats the values of bt were affected by the constant loss of species with increasing altitude (Fig. 3d, e). Likewise, when species turnover was compared between the forest and the pasture at each altitude, above 1800 m a.s.l. the similarity between habitats was lower and the values of bt reflect the loss of species (Fig. 3f). As for t he Chiles Volcano, at Rı´o Cusiana, species turnover between adjacent altitudinal levels in both forest and pasture decreases from low to intermediate altitudes around 1250 m a.s.l., and then rapidly increases towards the top of the gradient the (bt forest:. 0.24 0.9; bt pasture: 0.52 1.0). On Rı´o Cusiana, in contrast to what we observed on the other altitudinal gradients, the values of bt for both the forest and the pasture reflect the gain of species at the beginning of the gradient between 450 and 1250 m a.s.l. Above this the values of bt reflect the loss of species (Fig. 3g, h). The comparison between the forest and the pasture at each altitude indicate a strong species turnover between habitats, especially above 1500 m a.s.l., as well as values of bt that are strongly influenced by the loss of species along the entire altitudinal gradient (Fig. 3i). Variation in abundance and dominance On comparing the distribution of species abundance between habitats, above 1340 m a.s.l. on the Cofre

201

Cofre de Perote

Chiles Volcano 1 0 .8

Río Cusiana

(d)

1 0.8

0 .6 8

Number of species

0 -45

0 45

8 0.4

520 000 350 600 800 300 500-1 0-1 0-1 0-2 0-3 52 135 100 260 180

0 .8

(e)

0.8

-45

0 45

0-9

8 0.4

0. 8

450

900 1340 1860 2000 2340 3000

0.4

0-1 52

350

0-1 135

800

000 600 0-3 0-2 260 180

(f)

1 0.8

0.2 0

0-9 45

00 900

50 -12

50 12

00 -15

50 00 -17 -25 00 50 15 17

(i)

1 0.8

0.6

0.4 4

0.8 0.6

0. 6 8

0.2

5 50-

(h)

00 000 -2340 -3000 340 86 0 0-1 40-1 60-2 40 00 90 20 13 23 18

(c)

0.6

0 .2 0

0.2

00 50 00 00 50 00 0-9 -12 -15 -25 -17 -20 45 50 00 50 00 900 12 20 15 17

0. 4

0.4

0.2

8 4

0.6

0. 2

0. 6

0.8

0 860 340 000 -2340 -3000 -90 0-1 40-1 60-2 00 40 90 13 18 20 23

(b)

0.6

0 .4 4

(g)

0.4

0.2

βt

(a)

0.2 0 50

0 52

0 100

0 135

0 180

0 260

0 330

0.2 0

0 450

900

50 12

00 15

00 20

00 25

Altitude (m a.s.l.)

Fig. 3. Changes in bt diversity (Wilson and Shmida index; black dots) between adjacent altitudes in the forest (a, d, g) and pasture (b, e, h) along each altitudinal gradient. The bottom figures (c, f, i) show species turnover between habitats (forest vs pasture) at each altitudinal level. The filled bars indicate the number of species lost and the open bars indicated the number of species gained in each comparison.

de Perote, abundance was greater in pastures (Table 2). High on this mountain, the areas used for cattle were dominated by very few species (Fig. 4), while between 50 and 900 m a.s.l. the distribution of abundances was more even and the number of individuals captured was greater in the forest than in the pasture. On this mountain, Simpson’s index was greater for pastures for some altitudes and the change in diversity between habitats was significantly different from a random distribution (Table 4). In contrast, for the northern Andes independent of altitude, abundance was always greater in the forest (Table 2). For both the Chiles Volcano and Rı´o Cusiana, the range-abundance curves for the forest show a more even distribution than the open areas used for cattle do; areas where the dominance of only a few species increased (Fig. 4). On these mountains, Simpson’s index was generally greater at forest or was similar between habitats and, the changes in diversity were not significantly different from a random distribution (Table 4).

202

Discussion On the three altitudinal gradients studied, we observed a decrease in species richness as altitude increased; this phenomenon has been reported for different taxonomic groups on different mountains (Rahbek 1995 and references therein). However, the lack of comparative studies between taxonomic groups, and between natural and anthropogenic environments hinders efforts to properly contrast patterns of diversity with altitude (Lomolino 2001). In the case of dung beetles, this study indicates that in the northern Andes the rate of species loss with increasing altitude was much more pronounced in the forest than in the pasture. While on the Cofre de Perote (MTZ) the decrease was very similar for both habitat types. In these pastures, the decrease in species richness was lower owing to the presence of a set of species of northern affinity (Paleoamerican Montane and Nearctic Patterns) or of those that evolved on the Mexican High Plain (Altiplano Distribution Pattern). The latter is

Cofre de Perote 1000

50 m

900 m

450 m

2340 m

1340 m 2000 m

100

3000 m

1860 m 10

Chiles Volcano

1000

Abundance (Log 10)

1350 m 50 m 100

1000 m*

1800 m

520 m 10

RĂo Cusiana

1000

900 m

1500 m

1250 m 100

1750 m 2500 m 2000 m*

450 m

Abundance Range

Fig. 4. Dominance-diversity curves comparing the distribution of abundance for forest (black dots) and pasture (open squares) at each elevation along each gradient. Altitudes marked with an asterisk * indicate sites where it was not possible to collect in pastures (1000 m a.s.l. on the Chiles Volcano and 2000 m a.s.l. on the RÄąÂ´o Cusiana).

comprised of species with a heliophile habit that are of South American origin and were isolated in the MTZ a very long time ago (having migrated during the Oligocene-Miocene period, Halffter 1987). Consistent with the rate of species loss with increasing altitude, the opposite process species accumulation or g diversity in each type of habitat along each altitudinal gradient exhibited a similar pattern. For the Cofre de Perote, both habitats accumulated a similar number of species; while in the northern Andes species accumulation was

always lower for pasture than for the forest. As such, a first conclusion that can be drawn from these results is that the degree to which species richness decreases and the degree to which species are added in each type of habitat as altitude increases, indicate that the impact of cattle pasturing is different for each gradient, and results from the biogeographical differences between the mountain systems studied. At the local level, the pastures of the northern Andes always were less diverse than sites at equivalent altitudes

203

Table 4. Simpson’s index values (1/D) obtained for each type of habitat along each altitudinal gradient; d is the difference between forest and pasture in the Simpson’s index (see Data analysis). * Denotes elevations on the Chiles Volcano where it was not possible to calculate 1/D because abundance was52 individuals. Elevation (m a.s.l.)

Forest

Pasture

Number of times ½d½ simulated ½d½ observed

Cofre de Perote 50 450 900 1340 1860 2000 2340 3000

2.13 6.07 2.94 6.00 3.03 2.71 1.28 1.57

5.41 3.90 6.29 1.44 1.60 4.27 1.94 1.06

3.28 2.17 3.35 4.56 1.43 1.56 0.66 0.51

0 2 0 0 42 10 0 215

B0.0001 0.0002 B0.0001 B0.0001 B0.0042 0.001 B0.0001 0.02

7.67 4.70 4.58 5.68 5.90

2.19 3.33 4.79 5.00

5.48 1.37 0.89 0.90

0 3091 5788 7897

B0.0001 0.31 0.57 0.78

4.18 2.43 5.30 2.39 2.32 2.78 2.44

3.5 4.16 4.47 1.38 3.00 2.85

0.68 1.73 0.83 1.01 0.68 0.41

199 7992 3894 1566 2378 5860

0.02 0.79 0.39 0.15 0.23 0.58

Chiles Volcano 50 520 1000 1350 1800 2600* 3300* Rı´o Cusiana 450 900 1250 1500 1750 2000 2500

on the Cofre de Perote (MTZ). On the Cofre de Perote differences were accentuated at lower altitudes and less conspicuous on the upper parts of the altitudinal gradient. The latter was particularly notable at altitudes over 1800 m a.s.l. owing to the influence of the previously mentioned Holarctic and High Plain elements. This agrees with findings for different mountain regions in Europe where dung beetle richness and abundance, especially in open habitats, are not negatively correlated with altitude (Mene´ndez and Gutie´rrez 1996, Romero-Alcaraz and A´vila 2000). In temperate climates, such as the upper part of the Cofre de Perote, Scarabaeinae are restricted to or dominate open environments. This is a general phenomenon characteristic of the northern hemisphere, both latitudinally and altitudinally (Martin-Piera et al. 1992), and one that does not occur or is very limited in the northern Andes. There, the regional set of species than is adapted to forest conditions has more species that the set that is adapted to open habitats. The study by Amat et al. (1997) shows that the forests of the Sabana de Bogota´, Eastern Range (2800 2900 m a.s.l.) has up to 11 species while the pastures only have three, and these belong to genera that are widely diversified in the lowland forests. Based on these results it is also possible to conclude that if the reduction of forest in the upper slopes of the northern Andes continues as a conse-

204

quence of creating cattle pastures, the beetle fauna will become isolated in small forest remnants, as has been documented for birds and amphibians in fragmented areas above 1500 m a.s.l. in the mountains of Colombia and Ecuador (Kattan and A´lvarez-Lo´pez 1996, Marsh and Pearman 1997). Therefore, the integrity of these communities depends to a large extent on the connectivity of the forests along the altitudinal gradient, and on alternative land uses, both of which can buffer the impacts of cattle ranching on biodiversity (Pineda et al. 2005). The comparison of fauna composition at the tribe and genus level for each mountain and between habitats reveals notable differences: 32.5% of the species and 43% of the individuals on the Cofre de Perote belong to Onthophagini, while in the northern Andes this tribe represents no more than 8% of the species and 15% of the individuals. The present day distribution of Onthophagus (Onthophagini) is the result of an ancient process of invasion of the Americas from Asia, followed by intense diversification in North America, including Mexico. With 2000 species described, this genus is considered one of the most modern of the dung beetles and is supposed to have diversified during the Oligocene (ca 23 33 million years BP), a diversification that coincided with the expansion of pastures and the spread of mammals (Davis et al. 2002). Currently, the

representatives of this genus are ubiquitous members of beetle communities in areas where the forest has been cut at different altitudes in both Mexico and Central America (Halffter et al. 1995, Horgan 2002). In South America, however, this genus is restricted to distinct habitat types below 2000 m a.s.l. with few species at higher altitude in the mountains (Zunino and Halffter 1997). The other member of tribe Onthophagini found on the Cofre de Perote, Digitonthophagus gazella , is a notable example of the modern expansion process on the American continent and could serve as a model for understanding how Paleo-American tropical lineages expanded on this continent. This Afrotropical species was introduced in Texas in 1972 and in little more than 30 yr has made its way down to southern Nicaragua (Montes de Oca and Halffter 1998). According to Halffter et al. (1995), D. gazella is found in open habitats and is markedly associated with cattle dung. Its dispersal has been favored by deforestation and by the change in the use of large tracts of land to cattle pasture. Consequently, the invasion of introduced species is indicative of possible habitat deterioration (Kennedy et al. 2002). The results of this study and others (Halffter et al. 1995, Arrellano and Halffter 2003) indicate that the pastures of the Cofre de Perote have served as altitudinal dispersal routes for heliophile and thermophile species such as D. gazelle , and Euonicitellus intermedius (another species introduced in North America) from tropical lowland landscapes. So, the expansion of areas used for pasturing cattle in the mountains appears to have facilitated the expansion of those species adapted to open environments previously present in the region. This may be contributing to the homogenization of the fauna along altitudinal gradients, as reported for amphibians, reptiles and birds in Costa Rica’s mountains where this process is favored by an increase in temperature at higher altitudes as a consequence of global climate change (Pounds et al. 1999). Although we do not have a definitive image of horizontal colonization by beetles in the northern Andes mountain ranges, studies of plants and birds allows us to illustrate its relevance; as one moves up these mountains, the proportion of genera originating outside the tropics increases (Vuilleumier 1986, Gentry 2001). However, for the dung beetles and butterflies, colonization is mainly vertical (Decimon 1986, Escobar et al. 2006). The fauna of the intermediate and high altitudes on these mountains is a derivative of the found in the neighboring lowlands. This also occurs in the mountains of southeast Asia (Hanski and Niemela¨ 1990), Ecuador (Celis et al. 2004) and in some of the mountains of Costa Rica (Halffter and Reyes-Castillo unpubl.). Therefore for historical-biogeographical reasons, in the northern Andes species turnover between

adjacent altitudes mainly results from the loss of species. This is a product of the process of vertical colonization and can be explained by the restrictions imposed by altitude in environmental terms (decreasing temperature) and the reduction in food availability; conditions that require physiological adjustments if the higher altitudes of the mountains are to be colonized (Chown et al. 2002). There is an important difference in the species turnover between the two Andean transects. On the Chiles Volcano (located on the western slope of the Eastern Range), species turnover was dominated by the loss of species, while on the Rı´o Cusiana transect (located on the eastern slope of the Eastern Range), there was a gain of species below 1250 m a.s.l. the contact zone between the lowland fauna and the mountain forests. According to Lomolino (2001), the degree of overlap or juxtaposition between adjacent communities along altitudinal gradients contributes to explaining the type of relationship between species richness and altitude (monotonic model vs hump-shape model) and particularly, the degree of faunistic turnover between adjacent altitudinal bands. Precisely this was observed for five altitudes gradients between 08 and 78 North latitude on the Eastern cordillera of the Andes, Colombia (Escobar et al. 2005). Therefore, the differences in the species turnover patterns for opposite slopes in the northern Andes could result from the fact that their lowland dung beetle faunas differ in diversity and composition. This has been documented for the less diverse lowlands of the Pacific Plains on the western slope of the Andes (Peck and Forsyth 1982, Medina and Kattan 1996) and for the locations of the Amazonia-Orinoquia that are richest in species on the eastern slope of the Andes (Howden and Nealis 1975, Pulido et al. 2003). There appears to be an altitudinal gradient with respect to abundance, and it depends on habitat type. In general, abundance was much greater in forests than in pastures, while dominance increased in the pastures. However, the differences in abundance between the pastures of the different mountains above altitudes of 1750 m a.s.l. are marked. On the Cofre de Perote (MTZ) pastures were home to 39% of all the individuals collected, while in the northern Andes the proportion of individuals found in pastures was never higher than 10%. High total abundance values (biomass) have been recorded in the higher zones where species richness decreases in the mountains of Europe (Lumaret and Stiernet 1991) and in Asia (Hanski and Krikken 1991). In all of these studies, the dominance of a few species was found to increase in what seemed to be a compensating mechanism for adjusting populations to the available resources (Hanski and Cambefort 1991).

205

Although cattle dung can be abundant resource in the pastures of many tropical mountains, in this environment it dries out quickly and this modifies its microenvironmental and nutritional characteristics (Halffter 1991). Once changed, this dung can only be used by some species; species that are physiologically and behaviorally adapted to using the dung of medium sized and large herbivores in open areas. This is the case for many species of Holarctic (Onthophagus chevrolati , O. incensus) and Afrotropical (D. gazella and E. intermetus ) affinity in the MTZ or those of Neotropical affinity that exhibit wide ecological tolerance and are present in both mountain systems, such as Dichotomius colonicus , Ontherus mexicanus and Scatimus ovatus in the MTZ (Halffter et al. 1995, Arellano and Halffter 2003) and Dichotomius satanas , D achamas , Ontherus kirchii , O. brevicollis , Eurysternus marmoreus and E.caribaeus in the northern Andes (Amat et al. 1997, Escobar 2004). The limited presence of the Scarabaeinae in the pastures of the northern Andes highlights the following paradox: in spite of the abundance of food cow dung this environment is not available and effectively does nor exist for the majority of species that inhabit the native forest. This supports the concept that in Tropical America plant shade and its influence on microclimatic conditions on the ground are more important than a greater abundance of food (Halffter 1991).

Conclusion Although it was not possible to control factors such as the size of the mountains, the topography, the climatic variation, the configuration of the landscape and food supply, the results are revealing. The change in beetle community attributes that we observed in the face of an ecological change resulting from the transformation of forest to cattle pastures were clearly different along each altitudinal gradient. This suggests that processes of disturbance caused by human activity along altitudinal gradients can impact communities in different ways, depending on the geographic position of each mountain and particularly the biogeographical history of the group of species that inhabits it. This study contributes to the understanding that the response of communities to human activities (such as replacing forest with pastures, habitat loss and fragmentation) cannot be extracted from their regional context. Nor can they be understood without considering the biogeographical patterns and the evolutionary restrictions (e.g. habitat specialization) of species that belong to these communities (Ewer and Didham 2006).

Acknowledgements We are grateful to the subject editor of Ecography for valuable remarks and critiques. We thank

206

Bianca Delfosse for translating the article into English and Ute Kryger for offering valuable suggestions to the last version. Research in Colombia was supported by the Financiera Elećtrica Nacional (FEN) and by the Inst. Colombiano para el Desarrollo de la Ciencia y la Tecnologıá (COLCIENCIAS, project 2245-13-306-97). In Mexico, this research was financed by the Consejo Nacional de Ciencia y Tecnologia de Me´xico (CONACYT, project 37514-V), by the Comisioń Nacional para el Uso y Conocimiento de la Biodiversidad (CONABIO, projects 093-01 and EE005) and by Ministerio del Medio Ambiente y Recursos Naturales y el Consejo Nacinal de Ciencia y Tecnologia de Me´xico (SEMARNAT-CONACyT, project 2004-56-A1). Finally, the first author appreciates support from Univ. of Pretoria (Postdoctoral Fellowship Programme) to allow writing the last version of this article.

References Amat, G. et al. 1997. Patrones de distribucioń de escarabajos copro´fagos (Coleoptera: Scarabaeidae) en relictos del bosque altoandino, cordillera Oriental de Colombia. Caldasia 19: 191 204. Anon. 2002. Statistical database for agriculture. FAO, Rome. Arellano, L. and Halffter, G. 2003. Gamma diversity: derived from and a determinant of alpha diversity and beta diversity: an analysis of three tropical landscapes. Acta Zool. Mex. (n.s.) 90: 27 76. Brown, J. H. 2001. Mammals on mountainsides: altitudinal patterns of biodiversity. Global Ecol. Biogeogr. 10: 101 109. Celis, J. et al. 2004. Dung beetles (Coleoptera: Scarabaeinae) diversity in an altitudinal gradient in the Cucutu´ Range, Morona Santiago, Ecaudorian Amazon. Lyonia 7: 37 52. Challenger, A. 1998. Utilizacioń y conservacioń de los ecosistemas terrestres de Me´xico: Pasado, presente y futuro. CONABIO, Inst. de Biologıá, UNAM y Sierra Madre, Me´xico. Chown, S. L. et al. 2002. Physiological variation in insects: large-scale patterns and their implications. Biochem. Mol. Ecol. 531: 587 602. Colwell, R. K. 2005. EstimateS: statistical estimation of species richness and shared species from samples, v. 7.5.0. Bhttp://viceroy.eeb.uconn.edu/estimates . Colwell, R. K. and Coddington, J. A. 1994. Estimating the extent of terrestrial biodiversity through extrapolation. Phil. Trans. R. Soc. B 345: 101 118. Colwell, R. K. et al. 2004. Interpolating, extrapolating and comparing incidence-based species accumulation curves. Ecology 85: 2717 2727. Crawley, M. J. 2002. Statistical computing. Wiley. Davis, A. L. V. et al. 1999. Species turnover, community boundaries, and biogeographical composition of dung beetle assemblages across an altitudinal gradient in South Africa. J. Biogeogr. 26: 1039 1055. Davis, A. L. V. et al. 2002. Historical biogeography of Scarabaeinae dung beetles. J. Biogeogr. 29: 1217 1256.

Decimon, H. 1986. Origins of lepidopteran faunas in the high tropical Andes. In: Vuilleumier, F. and Monasterio, M. (eds), High altitude tropical biogeography. Oxford Univ. Press, pp. 500 533. Escobar, F. 2004. Diversity and composition of dung beetles (Scarabaeinae) assemblages in a heterogeneous Andean landscape. Trop. Zool. 17: 123 136. Escobar, F. et al. 2005. Altitudinal variation of dung beetle assemblages in the Colombian Andes. Global Ecol. Biogeogr. 14: 337 347. Escobar, F. et al. 2006. Assessing the origin of Neotropical mountains dung beetle assemblages (Scarabaeidae: Scarabaeinae): the comparative influence of vertical and horizontal colonization. J. Biogeogr. 33: 1793 1803. Errouissi, F. et al. 2004. Composition and structure of dung beetle (Coleoptera: Aphodiidae, Geotrupidae, Scarabaeidae) assemblages in mountain grasslands of the southern Alps. Ann. Entomol. Soc. Am. 97: 710 209. Ewer, R. M. and Didham, R. 2006. Confounding factors in the detection of species responses to habitat fragmentation. Biol. Rev. Camb. Philos. Soc. 81: 117 142. Ferrusquia-Villafranca, I. 1993. Geology of Mexico: a synopsis. In: Ramamoorthy, T. P. et al. (eds), Biological diversity of Mexico: origins and distribution. Oxford Univ. Press, pp. 3 107. Gentry, A. H. 2001. Patrones de diversidad y composicio´n floristica en los bosques de las montan˜as tropicales. In: Kappelle, K. and Brown, A. D. (eds), Bosques Nublados del Neotro´pico. Inst. Nacional de la Biodiversidad, pp. 63 84. Halffter, G. 1987. Biogeography of the mountain entomofauna of Mexico and Central America. Annu. Rev. Entomol. 32: 95 114. Halffter, G. 1991. Historical and ecological factors determining the geographical distribution of beetles (Coleoptera: Scarabaeidae: Scarabaeinae). Folia Entomol. Mex. 82: 195 238. Halffter, G. and Matthews, E. G. 1966. Natural history of dung beetles of the subfamily Scarabaeinae (Coleoptera: Scarabaeidae). Folia Entomol. Mex. 12 14: 195 238. Halffter, G. et al. 1995. Spatial distribution of three groups of Coleoptera along an altitudinal transect in Mexican Transition Zone and its biogeographical implications. Elytron 9: 151 185. Hanski, I. 1983. Distributional ecology and abundance of dung and carrion-feeding beetles (Scarabaeidae) in tropical rain forest in Sarawak, Borneo. Actas Zool. Fenn. 167: 1 45. Hanski, I. and Niemela¨, J. 1990. Elevational distributions of dung and carrion beetles in north Sulawesi. In: Knight, W. J. and Holloway, J. D. (eds), Insects and the rain forest of southeast Asia (Wallacea). Roy. Entomol. Soc., pp. 145 152. Hanski, I. and Cambefort, Y. 1991. Dung beetle ecology. Princeton Univ. Press. Hanski, I. and Krikken, J. 1991. Dung beetles in tropical forest in southeast Asia. In: Hanski, I. and Cambefort, Y. (eds), Dung beetles ecology. Princeton Univ. Press, pp. 179 197.

Henderson, P. A. and Seaby, R. M. 2002. Species diversity and richness v. 3.0. Pisces Conservational, Pennington. Horgan, F. G. 2002. Shady field boundaries and the colonization of dung by coprophagous beetles in central American pastures. Agricult. Ecosyst. Environ. 91: 25 36. Howden, H. F. and Nealis, V. 1975. Effects of deforestation clearing in a tropical rain forest on the composition of the coprophagous scarab beetle fauna (Coleoptera). Biotropica 7: 77 83. Huston, M. A. 1994. Biological diversity. Cambridge Univ. Press. Janzen, D. H. 1967. Why mountain passes are higher in the tropics. Am. Nat. 101: 233 249. Kappelle, K. and Brown, A. D. 2001. Bosques Nublados del Neotro´pico. Inst. Nacional de la Biodiversidad-INBio, Santo Domingo de Heredia, Costa Rica. Kattan, G. and .A´lvarez-Lo´pez, H. 1996. Preservation and management of biodiversity in fragmented landscape in the Colombian Andes. In: Schelhas, J. and Greenberg, R. (eds), Forest patches in tropical landscapes. Island Press, pp. 3 18. Kennedy, T. A. et al. 2002. Biodiversity as a barrier to ecological invasion. Nature 417: 636 638. Lobo, J. M. and Halffter, G. 2000. Biogeographical and ecological factors affecting altitudinal variation of mountainous communities of coprophagous beetles (Coleptera, Scarabaeoidea): a comparative study. Ann. Entomol. Soc. Am. 93: 115 126. Lomolino, M. V. 2001. Elevation gradients of species-density: historical and prospective views. Global Ecol. Biogegr. 10: 3 13. Lomolino, M. V. and Perault, D. R. 2004. Geographic gradients of deforestation and mammalian communities in a fragmented, temperate rain forest landscape. Global Ecol. Biogeogr. 13: 55 64. Lumaret, J. P. and Stiernet, N. 1991. Montane dung beetles. In: Hanski, I. and Cambefort, Y. (eds), Dung beetles ecology. Princeton Univ. Press, pp. 242 254. Magurran, A. E. 1988. Ecological diversity and its measurement. Princeton Univ. Press. Marsh, D. M. and Pearman, P. B. 1997. Effect of habitat fragmentation on the abundance of two species of Leptodactylid frogs in an Andean montane forest. Conserv. Biol. 11: 1323 1328. Martin-Piera, F. et al. 1992. Ecology and biogeography of dung-beetle communities (Coleoptera, Scarabaeoidea) in an Iberian mountain range. J. Biogeogr. 19: 677 691. Medina, C. A. and Kattan, G. H. 1996. Diversidad de coleo´pteros copro´fagos (Scarabaeidae) de la reserva forestal de Escalerete. Cespedesia 21: 89 102. Mene´ndez, R. and Gutie´rrez, D. 1996. Altitudinal effects on habitat selection of dung beetles (Scarabaeoidea: Aphodiidae) in the northern Iberian Peninsula. Ecography 19: 331 317. Montes de Oca, E. and Halffter, G. 1998. Invasion of Mexico by two dung beetles previously introduced into the United States. Stud. Neotrop. Fauna Environ. 33: 37 45. Murgueitio, E. 2003. Environmental impact of milk production systems in Colombia and alternative solutions. Livestock

207

Research for Rural Development 15: 1010, Bhttp:// www.cipay.org.co/Irrd/Irrd15/10/Murg1510.htm . Nichols, E. et al. in press. Dung beetle response to tropical forest modification and fragmentation: a quantitative literature review and meta-analysis. Biol. Conserv. Peck, S. B. and Forsyth, A. 1982. Composition, structure and comparative behaviour in a guild of Ecuadorian rain forest dung beetles (Coleoptera, Scarabaeidae). Can. J. Zool. 60: 1624 1634. Pineda, E. et al. 2005. Biodiversity in cloud forest and shade coffee: analysis of three indicator groups. Conserv. Biol. 19: 400 410. Pounds, J. A. et al. 1999. Biological response to climate changes on a tropical mountain. Nature 398: 611 615. Pulido, L. A. et al. 2003. Escarabajos copro´fagos (Coleoptera: Scarabaeidae: Scarabaeinae) del parque nacional natural Sierra del Chiribiquete, Caqueta´, Colombia. In: Onore, G. et al. (eds), Escarabeidos de Latinoame´rica: estado actual del conocimiento. Monografias Tercer Milenio, Vol. 3. Sociedad Entomolo´gica Aragonesa, pp. 51 58. Rahbek, C. 1995. The altitudinal gradient of species richness: a uniform pattern? Ecography 18: 200 205. Romero-Alcaraz, E. and .A´vila, J. M. 2000. Effect of altitude and type of habitat on the abundance and diversity of Scarabaeoid dung beetles (Scarabaeoidea) assemblages in a

Download the appendix as file E4818 from Bwww.oikos.ekol.lu.se/appendix .

208

Mediterranean area southern Iberian Peninsula. Zool. Stud. 39: 351 359. Solow, A. R. 1993. A simple test for change in community structure. J. Anim. Ecol. 62: 191 193. van der Hammen, T. 1995. Global change, biodiversity, and conservation of Neotropical montane forest. In: Churchill, S. P. et al. (eds), Biodiversity and conservation of Neotropicsl montane forests. Proc. of the Neotropicsl Montane Forest Biodiversity and Conservation Symp. New York Botanical Garden, pp. 603 607. van der Hammen, T. and Hooghiemstra, H. 2001. Historia y paeloecologia de los bosques montanos andinos tropicales. In: Kappelle, K. and Brown, A. D. (eds), Bosques Nublados del Neotro´pico. Inst. Nacional de la Biodiversidad INBio, pp. 63 86. Vuilleumier, F. 1986. Origins of the tropical avifauna in the high Andes. In: Vuilleumier, F. and Monasterio, M. (eds), High altitude tropical biogeography. Oxford Univ. Press, pp. 586 622. Wilson, M. V. and Shmida, A. 1984. Measuring beta diversity with presence-absence data. J. Ecol. 72: 1055 1064. Zar, J. H. 1996. Biostatistical analysis, 2nd ed. Prentice Hall. Zunino, M. and Halffter, G. 1997. Sobre Onthophagus Latreille, 1802 americanos. Elytron 11: 157 168.

ECOGRAPHY 29: 919 927, 2006

Elevational patterns of frog species richness and endemic richness in the Hengduan Mountains, China: geometric constraints, area and climate effects Cuizhang Fu, Xia Hua, Jun Li, Zheng Chang, Zhichao Pu and Jiakuan Chen

Fu, C., Hua, X., Li, J., Chang, Z., Pu, Z. and Chen, J. 2006. Elevational patterns of frog species richness and endemic richness in the Hengduan Mountains, China: geometric contraints, area and climate effects. Ecography 29: 919 927. We studied frog biodiversity along an elevational gradient in the Hengduan Mountains, China. Endemic and non-endemic elevational diversity patterns were examined individually. Competing hypotheses were also tested for these patterns. Species richness of total frogs, endemics and non-endemics peaked at mid-elevations. The peak in endemic species richness was at higher elevations than the maxima of total species richness. Endemic species richness followed the mid-domain model predictions, and showed a nonlinear relationship with temperature. Water and energy were the most important variables in explaining elevational patterns of non-endemic species richness. A suite of interacting climatic and geometric factors best explained total species richness patterns along the elevational gradient. We suggest that the mid-domain effect was an important factor to explain elevational richness patterns, especially in regions with high endemism. C. Fu (czfu@fudan.edu.cn), X. Hua, J. Li, Z. Chang, Z. Pu and J. Chen, Ministry of Education Key Laboratory for Biodiversity Science and Ecological Engineering, and Inst. of Biodiversity Science, Fudan Univ., Shanghai 200433, China.

Understanding variation in species richness has been a central aim of community ecology for decades. Although the latitudinal gradient in species richness is one of the most studied diversity patterns, research on montane biotas has continued to play a prominent role in understanding the distribution of organisms (Lomolino et al. 2006). Rahbek (1995, 2005) recognized three basic types of elevational diversity patterns, i.e. species richness decrease with elevation, peak at low elevation plateaus, and peak at mid-elevation. Area, climate and geometric constraints are the most frequently cited explanation for both a linear and a humped relationships between species richness and elevation (Rahbek 1995, 1997, Brown 2001, Lomolino 2001, McCain 2004, 2005, 2006, Bhattarai et al. 2004, CardeluÂ´s et al. 2006, Fu et al. 2006, Kluge et al. 2006).

The elevational gradient of species richness may be intricately related to species-area relationships (Lomolino 2001). Some studies found that area of elevational belts explained a large proportion of the variation in species richness (Rahbek 1997, Kattan and Franco 2004, Bachman et al. 2004, Fu et al. 2004). Elevational patterns of diversity were also commonly explained by water and energy (Bhattarai et al. 2004, McCain 2006, Fu et al. 2006). Recent studies demonstrated that the mid-domain effect (MDE) is also a powerful explanatory variable in elevational diversity patterns (McCain 2004, CardeluÂ´s et al. 2006, Kluge et al. 2006, Watkins et al. 2006). The mid-domain effect arises from geometric constraints on species ranges within a bounded domain (Colwell and Lees 2000, Colwell et al. 2004). Based on chance alone, the likelihood of the

Accepted 5 October 2006 Copyright # ECOGRAPHY 2006 ISSN 0906-7590 ECOGRAPHY 29:6 (2006)

919

elevational ranges of many species overlapping is higher at mid-elevations than for the lower and higher elevations (Colwell et al. 2004, Cardelu´s et al. 2006). MDE theory predicts that endemic species will show more of MDE peak than non-endemics (Lee et al. 1999, McCain 2004, Colwell et al. 2004) and that small-ranged species will show less of an MDE peak than large-ranged species (Lee et al. 1999, Cardelu´s et al. 2006, Dunn et al. 2006, Kluge et al. 2006, Watkins et al. 2006). Biogeographical variation in species richness and endemic richness is critical to our understanding and conservation of biological diversity (Grytnes and Vetaas 2002). Areas with high species richness may also have a large number of endemic species, but patterns in richness and endemism are not necessarily positively related (Whittaker et al. 2001). Recent studies have found that species richness patterns across elevational gradients for total species and endemic species were different for plants (Kessler 2002, Grytnes and Vetaas 2002) and freshwater fishes (Fu et al. 2004). Geographic range size affects determinants of geographic patterns in species richness (Jetz and Rahbek 2002, Mora and Robertson 2005). Recent studies have found that there were different patterns and processes for elevational diversity of large-ranged and small-ranged species (Cardelu´s et al. 2006, Kluge et al. 2006). In this study, we delineated elevational patterns of frog species richness and endemic richness in the Hengduan Mountains of China, and assessed the ability of area, climate, and geometric constraints to explain the elevational diversity patterns.

Methods Study area The Hengduan Mountains (238 338N, 978 1038E) of China occur within an ecotone between the Oriental (Indo-Malayan) region and the Palearctic region. They span east of Tibet, west of Sichuan province and northwest of Yunnan Province, and are a part of the Qinghai-Tibet Plateau (Fig. 1). The total area of the mountains is ca 0.56 million square kilometers, and elevation range from ca 200 to 7300 m a.s.l. based on a global digital elevation model (DEM) from B/http:// lpdaac.usgs.gov/gtopo30/gtopo30.asp /. The altitude of this area declines from northwest to southeast (Fig. 1). Most parts of the area are characterized by a series of paralleled mountain ranges and rivers from south to north, with a sharp altitudinal differentiation.

Richness data We compiled frog distribution and altitudinal data from secondary sources (Appendix 1). Species checklists and 920

the altitudinal limits of occurrences were listed in Appendix 2. Species accumulation curves and species richness estimates showed that sampling was adequate (see the details in Appendix 3). One species, Paa feae is excluded from the analyses because of lacking information on their elevational distribution. Endemic species are defined according to their distribution being limited to the Hengduan Mountains.

Area data To calculate the area at each elevational band in the Hengduan Mountains, we used a global digital elevation model (DEM), GTOPO30, with a horizontal grid spacing of 30 arc-seconds (ca 1 km2) from B/http:// lpdaac.usgs.gov/gtopo30/gtopo30.asp /. First, we extracted the map, which contains elevational information of Hengduan Moutains, from the global GTOPO30 map. This map was converted to Lambert-Azimuthal equal area projection map, and rasterized at 1 /1 km grid cells. Finally, we counted the number of grid cells (1 km2/grid cell) within each 200-m elevational interval based on the elevational value of each grid cell, and summed up to the area. The relationship between area and elevation is shown in Appendix 4.

Climatic variables In this study, two climatic variables, annual mean temperature and annual mean precipitation, were included, selected because they have been shown to be important correlates of broad-scale richness gradients. Temperature and precipitation data were obtained from 104 climate stations with a record length of 30 yr (1951 1980) from Climate Resource Database (B/http:// www.data.ac.cn/zrzy/g03.asp /) and Meteorological Administration of local or central goverment in China. Six 100-m elevation intervals lack climate stations, including those at 600 700, 800 900, 2400 2500, 2700 2800, 3500 3600, and 4000 4100 m. In these 100-m elevation zones, temperature and precipitation data were interpolated from the mean value of the nearest adjacent upper and lower climatic station’s record. The relationships between climatic variables and elevation are shown in Appendix 4.

Geometric constraints: the mid-domain effect To test the mid-domain effect (MDE), diversity patterns were compared to null model predictions using a novel, discrete MDE model that does not necessitate the use of interpolated ranges (Dunn et al. 2006). The simulation program was implemented in RangeModel software, ver. 5 (Colwell 2006). For each data set, expected mean ECOGRAPHY 29:6 (2006)

Fig. 1. The sketch map of the Hengduan Mountains in China.

richness and its 95% confidence interval over the domain based on 50 000 simulations sampled were used to assess the impact of spatial constraints on the elevational diversity gradients.

Data analysis We divided the range of elevation into 200-m bands between 400 and 5000 m, and calculated the total number of species in each band to examine the relationship between species richness and elevation. Following the methods of Rahbek (1997), we tested four models of the relationships between species richness (S) and area (A), i.e. S/A, S/log A, log S/A, and log S/log A, and selected the best fit S/A model based on comparisons of r values. To account for multicollinearity, we used a combination of multi-regression techniques recommended by Graham (2003). These include single and multiple ordinary least squares models (OLS models), single and multiple conditional autoregressive models (CAR models), and generalized linear models (GLM models). Recent studies have advocated using the CAR models to identify the predictive power of the hypotheses for explaining geographic richness pattern of different taxa (Jetz and Rahbek 2002, Mora and Robertson 2005). ECOGRAPHY 29:6 (2006)

The GLM models have also been used to relate species richness to explanatory variables along the elevational gradients (Bhattarai et al. 2004, Kluge et al. 2006). GLM models were performed with S-Plus 7.0. OLS and CAR models were performed in SAM 1.0 (Rangel et al. 2006). We used Akaikeâ&#x20AC;&#x2122;s information criteria (AIC) to compare the fit of the OLS and CAR models, and smaller AIC values indicated a better fit. We repeated all analyses for large and small ranges, separately, to compare the role of candidate predictors for range size classes (following Jetz and Rahbek 2002). We divided each dataset (the overall-species set and the taxon subsets of endemics and non-endemics) into the 50% of species with large ranges and the 50% of species with the small ranges. We used STATISTICA (ver. 6.0) for the graph representation.

Results Frog fauna Our synthesis found 94 frog species from the Hengduan Mountains (Appendix 2), distributed among 7 families and 29 genera. Among these, there are 37 species endemic to the mountains. The most species-rich 921

families are the Megophryidae (8 genera and 28 species), the Ranidae (10 genera and 35 species), and the Rhacophoridae (4 genera and 12 species). The frogs are distributed between 400 and 5000 m a.s.l.

Elevational diversity patterns Overall species of total, endemic and non-endemic frogs showed a humped relationship between species richness and elevation (Fig. 2). Peak in endemic overall species richness (2400 2600 m elevational zones, Fig. 2b) was at higher elevations than the maxima of non-endemic or total overall species richness (1200 1400 m elevational zones, Fig. 2a, c).

Partitioning species into range-size categories highlighted disparate contributions to overall species richness patterns (Fig. 2). There were higher correlation coefficients between large-ranged species richness and overall species richness (total, r2 /0.96; endemics, r2 / 0.97; non-endemics, r2 /0.96) than between smallranged species richness and overall species richness (total, r2 /0.86; endemics, r2 /0.90; non-endemics, r2 /0.95).

Species richness patterns and explained variables Total species richness was strongly correlated with temperate and precipitation across the data sets of overall, large-ranged and small-ranged species (Table 1,

Fig. 2. Species richness patterns along elevational gradients (black circles and lines) in the Hengduan Mountains including 95% simulation limits (lines only) of the mid-domain analyses from 50 000 simulations samples for overall species and separated for large-ranged and small-ranged species of (a) total, (b) endemic, and (c) non-endemic frogs in the Hengduan Mountains.

922

ECOGRAPHY 29:6 (2006)

Table 1. Relationships between species richness of total, endemic and non-endemic frogs and explained variables using simple ordinary least squares (OLS) and simple conditional autoregressive (CAR) models for overall species and separated for large-ranged and small-ranged species in the Hengduan Mountains. Variables

Total

Overall species

Large-ranged species

Small-ranged species

Endemics

Overall species

Large-ranged species

Small-ranged species

Non-endemics

Overall species

Large-ranged species

Small-ranged species

MDE Area MAP MAT MDE Area MAP MAT MDE Area MAP MAT MDE Area MAP MAT MDE Area MAP MAT MDE Area MAP MAT MDE Area MAP MAT MDE Area MAP MAT MDE Area MAP MAT

OLS models

CAR models 2

AIC

2.33* /0.97 6.53** 5.50** 3.05** /0.51 4.82** 4.42** 1.15 /1.84 10.60** 6.95** 8.73** /0.01 0.63 0.90 10.44** /0.27 0.68 0.90 3.90** 0.50 0.49 0.82 0.42 /1.19 11.69** 7.16** 1.36 /0.72 7.70** 6.19** /0.40 /1.66 11.54** 6.32**

137 141 117 122 117 125 108 110 92 90 51 66 66 101 100 100 41 83 82 82 40 52 52 52 133 132 87 105 105 106 76 83 99 96 53 74

0.21 0.04 0.66 0.59 0.31 0.01 0.53 0.48 0.06 0.14 0.84 0.70 0.78 B/0.01 0.02 0.04 0.84 B/0.01 0.02 0.04 0.42 0.01 0.01 0.03 B/0.01 0.06 0.87 0.71 0.08 0.02 0.74 0.65 B/0.01 0.12 0.86 0.66

2.56* /0.39 5.59** 4.74** 3.30** 0.07 4.11** 3.86** 1.28 /1.33 8.45** 5.55** 7.90** /0.06 0.28 0.76 9.53** /0.43 0.37 0.76 3.23* 0.59 0.15 0.71 0.39 /0.43 9.70** 5.88** 1.42 0.06 6.41** 5.25** /0.4 /0.96 10.00** 5.06**

127 144 134 134 102 127 122 121 89 96 65 76 45 102 102 103 11 84 85 85 41 55 56 56 133 135 102 112 100 109 94 94 102 100 38 75

0.61 0.15 0.46 0.46 0.72 0.15 0.33 0.35 0.35 0.15 0.77 0.64 0.93 0.19 0.17 0.15 0.97 0.19 0.16 0.14 0.52 0.13 0.11 0.09 0.24 0.17 0.79 0.69 0.43 0.17 0.56 0.55 0.13 0.18 0.95 0.73

Model fit was assessed using the Akaike information criteria (AIC), smaller values indicate a better fit. *p B/0.05; **p B/0.01. MDE, the mid-domain effect; MAP, mean annual precipitation; MAT, mean annual temperature.

Appendix 5). The best fit multiple models also identified MDE as a strong predictor of overall and large-ranged species, whereas the explanatory power of MDE was low for small-ranged species (Table 2, Fig. 2a). Area was the weakest predictors of total richness (Table 1 and 2, Appendix 5). Endemic richness was strongly correlated only with MDE, whereas all other variables showed weak linear correlations across the data sets of overall, large-ranged and small-ranged species (Table 1, Fig. 2b). The explanatory power of MDE was stronger for largeranged species than for small-ranged species (Table 1 and 2). Endemic richness also showed a nonlinear relationship with temperature (Appendix 5). Non-endemic species richness was strongly correlated with temperate and precipitation, whereas MDE and area were weak predictors of non-endemic species richness across the data sets of overall, large-ranged and small-ranged species (Table 1 and 2, Fig. 2c, Appendix 5). ECOGRAPHY 29:6 (2006)

Discussion Elevational diversity patterns Amphibians and reptiles generally show a monotonic decline in species richness with increasing altitude (Brown and Alcala 1961, Duellman 1988, Fauth et al. 1989, Nathan and Werner 1999), although a humped relationships between species richness and elevation have been observed in particular habitats (Heyer 1967, Fischer and Lindenmayer 2005, Fu et al. 2006). However, growing evidence suggested that mid-elevational peaks in species richness for a wide variety of taxa are perhaps more general (Rahbek 1995, 2005, Lomolino 2001). This study showed a diversity peak at midelevations for frog species richness in the Hengduan Mountains. Other studies from the adjacent regions of the Hengduan Mountains, plant diversity in the Nepal Himalaya and the Indian Western Himalaya, and small mammal diversity in the Mt. Qilian also reported similar elevational diversity patterns (Grytnes and Vetaas 2002, 923

924 Table 2. Multiple ordinary least squares (OLS) and multiple conditional autoregressive (CAR) regressions for explained variables and species richness of total frogs, endemics and nonendemics including overall species and separated for large-ranged and small-ranged species in the Hengduan Mountains. Model A included all explained variables, model B all variable except the mid-domain effect (MDE). Magnitudes of t -values indicate variable importance in the models. Overall species

Total

Endemics

Non-endemics

MDE Area MAP MAT AIC Model fit MDE Area MAP MAT AIC Model fit MDE Area MAP MAT AIC Model fit

t t t t R2 t t t t R2 t t t t R2

Large-ranged species

OLS models

CAR models

OLS models

4.9** 0.6 2.9** 0.3 105 0.87 10.1** /1.0 /1.8 2.0 65 0.86 0.8 1.3 4.7** /0.5 91 0.89

/ 1.3 1.8 0.6 121 0.70 0.7 /0.8 1.1 105 0.08 1.6 4.7** /0.4 88 0.89

4.9** 0.7 3.0** 0.8 98 0.93 9.4** /1.1 /1.9 2.1 34 0.97 0.8 1.4 4.8** 0.001 100 0.89

/ 1.2 1.9 0.9 139 0.51 0.5 /0.9 1.2 109 0.20 1.5 4.7** 0.1 104 0.84

5.7** 1.4 2.1 0.8 91 0.86 12.9** /2.1* /0.8 1.1 38 0.91 2.8* 2.0 2.9** 0.6 72 0.86

B 1.7 0.9 1.0 111 0.59 0.4 /0.6 0.8 88 0.06 2.3* 2.1* 0.9 77 0.80

Small-ranged species

CAR models

OLS models

CAR models

1.7 0.9 1.4 126 0.42 0.03 /0.6 0.9 92 0.18 2.4* 2.2* 1.5 95 0.67

2.2* /1.2 4.5** /1.1 54 0.88 4.3** 0.2 /2.4* 2.5* 42 0.59 2.4* 0.1 7.6** /2.8* 49 0.93

/0.4 4.3** /1.0 56 0.85 1.4 /1.2 1.6 55 0.15 /0.6 6.7** /2.6* 52 0.90

2.1 /1.3 4.4** /0.6 60 0.89 3.9** 0.3 /2.6* 2.7* 41 0.71 2.0 0.05 7.3** /2.2* 62 0.91

/0.6 4.3** /0.7 74 0.76 1.4 /1.5 1.8 60 0.21 /0.5 6.7** /2.0 43 0.95

5.8** 1.6 2.1 1.3 78 0.94 11.9** /2.3* /0.8 1.1 /5 0.98 2.8* 2.3* 2.9* 1.2 72 0.90

Model fit was assessed using the Akaike information criteria (AIC), smaller values indicate a better fit. *p B/0.05; **p B/0.01. Abbreviations expressed as in Table 1.

ECOGRAPHY 29:6 (2006)

Li et al. 2003, Bhattarai et al. 2004, Oommen and Shanker 2005). More studies on local and region diversity of amphibians and reptiles are needed to make generalizations on elevational richness patterns of these groups at global scale. In the Hengduan Mountains, species richness patterns across the elevational gradient for total frogs and endemics were different, and the maxima in diversity for endemic frogs peaked at higher elevations. Similar patterns were also reported in plants (Kessler 2002, Grytnes and Vetaas 2002) and freshwater fishes (Fu et al. 2004). Kessler (2002) thought that reduced surface area and more divided topography lead to more isolated populations and hence higher speciation rates with increasing elevation. Increased isolation and reduced dispersal might have resulted in increased differentiation and higher endemism with increasing elevation (Brown 2001). Recent studies reported that small-ranged and largeranged species showed markedly different species richness patterns, and geographic patterns in species richness were mainly based on large-ranged species because their larger number of distribution records had a disproportionate contribution to the species richness counts (Lee et al. 1999, Jetz and Rahbek 2002, Lennon et al. 2004, Cardelu´s et al. 2006, Dunn et al. 2006, Kluge et al. 2006, Kreft et al. 2006, Watkins et al. 2006). Va´zquez and Gaston (2004) further found that this pattern was most clearly demonstrated by endemic species. In this study, we also found that large-ranged species contributed more to overall richness pattern than small-ranged species in the data sets of total species and endemics. However, there were equal contributions for largeranged and small-ranged non-endemics to overall nonendemic richness. Below we discuss how area, climate and MDE may influence elevational patterns of frog species richness in the Hengduan Mountains.

Area effects Area may be a crucial parameter determining elevational diversity patterns because area generally declines with increasing elevation (Rahbek 1997, Lomolino 2001). The influence of area in determining regional species richness in altitudinal ranges has been shown for different taxa (Kattan and Franco 2004, Bachman et al. 2004, Fu et al. 2004). In this study, we found that area as a single predictor was less important to explaining richness patterns along an elevational gradient. The reasons for this may be from the special relationships between area and elevation. Area showed a bimodal relationship with elevation in the Hengduan Mountains (Appendix 4). ECOGRAPHY 29:6 (2006)

The mid-domain effect Many studies reported that the mid-domain effect (MDE) was an important variable to explain species richness patterns along elevational gradients (McCain 2004, Bachman et al. 2004, Oommen and Shanker 2005, Cardelu´s et al. 2006, Kluge et al. 2006, Watkins et al. 2006). The frog datasets in the Hengduan Moutains also showed that MDE accounted for a significant proportion of elevational patterns of total species richness. Endemic frog richness in the Hengduan Moutains was strongly correlated with MDE, whereas MDE was the weak predictor of non-endemic frog richness. It confirmed that MDE theory predicts that endemic species will show more of MDE peak than non-endemics (Lee et al. 1999, McCain 2004, Colwell et al. 2004). In this study, we found that the explanatory power of MDE was stronger for large-ranged species than for small-ranged species in the datasets of total frogs and endemics. It confirmed that MDE theory predicts that small-ranged species will show less of an MDE peak than large-ranged species (Lee et al. 1999, Jetz and Rahbek 2002, Colwell et al. 2004, 2005, Mora and Robertson 2005, Cardelu´s et al. 2006, Dunn et al. 2006, Kluge et al. 2006, Watkins et al. 2006).

Climatic effects A few studies have shown that water inputs played a prominent role in richness patterns of amphibian in Iberia (Schall and Pianka 1977) and frogs in Australia and the United States (Schall and Pianka 1978), and others found that energy input constrainted amphibian richness in North America (Currie 1991, Allen et al. 2002). And a recent study found that water-energy dynamics determined amphibian richness pattern in Europe (Rodrı´guez et al. 2005). Lomolino (2001) hypothesized that many components of climate and local environments varied along the elevational gradients and ultimately created the variation in species richness. In this study, climatic variables, annual mean precipitation and annual mean temperature were also found to be important predictors for frog species richness when total species and non-endemic species were considered. Other studies also observed that water and temperature were correlated with elevational richness in plants (Bhattarai et al. 2004, Kro¨mer et al. 2005) and reptiles (Fu et al. 2006). For endemic frogs in the Hengduan Mountains, water and temperature were only weakly correlated with patterns of species richness.

Conclusions This study has provided some answers to the questions presented at the outset. 1) We could reject the argument 925

that frog species richness has a monotonic decreasing relationship along the elevational gradient and replaced this with an alternative unimodal hypothesis. 2) Elevational richness patterns for total frogs, endemics and non-endemics exhibited different patterns, and endemic richness peaked at higher elevations. 3) Land area explained a small amount of the variation in species richness of total frogs, endemics and non-endemics. 4) Endemic species richness followed the mid-domain model predictions, and showed a nonlinear relationship with temperature. 5) Water and energy were the most important variables in explaining elevational patterns of non-endemic species richness. 6) A suite of interacting climatic and geometric factors best explained total species richness patterns along the mountains elevational gradient. Acknowledgements This study was financially supported by ‘‘Sustaining project of career development of young teachers’’.

References Allen, A. P. et al. 2002. Global biodiversity, biochemical kinetics, and the energetic-equivalence rule. Science 297: 1545 1548. Bachman, S. et al. 2004. Elevational gradients, area and tropical island diversity: an example from the palms of New Guinea. Ecography 27: 299 310. Bhattarai, K. R. et al. 2004. Fern species richness along a central Himalayan elevational gradient, Nepal. J. Biogeogr. 31: 389 400. Brown, J. H. 2001. Mammals on mountainsides: elevational patterns of diversity. Global Ecol. Biogeogr. 10: 101 109. Brown, W. C. and Alcala, A. C. 1961. Populations of amphibians and reptiles in the submontane and montane forest of Cuernos Negros, Philippine Islands. Ecology 42: 628 636. Cardelu´s, L. C. et al. 2006. Vascular epiphyte distribution patterns: explaining the mid-elevation richness peak. J. Ecol. 94: 144 156. Colwell, R. K. 2006. RangeModel: a Monte Carlo simulation tool for assessing geometric constraints on species richness. Ver. 5. User’s guide and application published at B/http:/ viceroy.eeb.uconn.edu/RangeModel /. Colwell, R. K. and Lees, D. C. 2000. The mid-domain effect: geometric constraints on the geography of species richness. Trends Ecol. Evol. 15: 70 76. Colwell, R. K. et al. 2004. The mid-domain effect and species richness patterns: what have we learned so far? Am. Nat. 163: E1 E23. Colwell, R. K. et al. 2005. The Mid-Domain effect: there’s a baby in the bathwater. Am. Nat. 166: E149 E154. Currie, D. J. 1991. Energy and large-scale patterns of animaland plant-species richness. Am. Nat. 137: 27 49. Duellman, W. E. 1988. Patterns of species diversity in anuran amphibians in the American tropics. Ann. Missouri Bot. Gard. 75: 79 104. Dunn, R. R. et al. 2006. The river domain: why are there more species halfway up the river? Ecography 29: 251 259. Fauth, J. E. et al. 1989. Elevational patterns of species richness, evenness, and abundance of the Costa Rican leaf litter herpetofauna. Biotropica 21: 178 185. Fischer, J. and Lindenmayer, D. B. 2005. The sensitivity of lizards to elevation: a case study from south-eastern Australia. Div. Distrib. 11: 225 233.

926

Fu, C. et al. 2004. Patterns of diversity, altitudinal range and body size among freshwater fishes in the Yangtze River basin, China. Global Ecol. Biogeogr. 13: 543 552. Fu, C. et al. 2006. Elevational gradients of diversity for lizards and snakes in the Hengduan Mountains, China. Biodiv. Conserv., in press. Graham, M. H. 2003. Confronting multicollinearity in ecological multiple regression. Ecology 84: 2809 2815. Grytnes, J. A. and Vetaas, O. R. 2002. Distribution of vascular plant species richness and endemic richness along the Himalayan elevation gradient in Nepal. Global Ecol. Biogeogr. 11: 291 301. Heyer, W. R. 1967. A herpetofauna study of an ecological transect through the Cordillera de Tilaran, Costa Rica. Copeia 1967: 259 271. Jetz, W. and Rahbek, C. 2002. Geographic range size and determinants of avian species richness. Science 297: 1548 1551. Kattan, G. H. and Franco, P. 2004. Bird diversity along elevational gradients in the Andes of Colombia: area and mass effects. Global Ecol. Biogeogr. 13: 451 458. Kessler, M. 2002. The elevational gradient of Andean plant endemism: varying influences of taxon-specific traits and topography at different taxonomic levels. J. Biogeogr. 29: 1159 1165. Kluge, J. et al. 2006. What drives elevational patterns of diversity? A test of geometric constraints, climate and species pool effects for pteridophytes on an elevational gradient in Costa Rica. Global Ecol. Biogeogr. 15: 358 371. Kreft, H. et al. 2006. The significance of geographic range size for spatial diversity patterns in Neotropical palms. Ecography 29: 21 30. Kro¨mer, T. et al. 2005. Diversity patterns of vascular epiphytes along an elevational gradient in the Andes. J. Biogeogr. 32: 1799 1809. Lee, D. C. et al. 1999. A null model for species richness gradients: bounded range overlap of butterflies and other rainforest endemics in Madagascar. Biol. J. Linn. Soc. 67: 529 584. Lennon, J. J. et al. 2004. Contribution of rarity and commonness to patterns of species richness. Ecol. Lett. 7: 81 87. Li, J. S. et al. 2003. Elevational gradients of small mammal diversity on the northern slopes of Mt. Qilian, China. Global Ecol. Biogeogr. 12: 449 460. Lomolino, M. V. 2001. Elevational gradients of species diversity: historical and prospective views. Global Ecol. Biogeogr. 10: 3 13. Lomolino, M. V. et al. 2006. Biogeography, 3rd ed. Sinauer. McCain, C. M. 2004. The mid-domain effect applied to elevational gradients: species richness of small mammals in Costa Rica. J. Biogeogr. 31: 19 31. McCain, C. M. 2005. Elevational gradients in diversity of small mammals. Ecology 86: 366 372. McCain, C. M. 2006. Could temperature and water availability drive elevational richness patterns? A global case study for bats. Global Ecol. Biogeogr., in press. Mora, C. and Robertson, D. R. 2005. Causes of latitudinal gradients in species richness: a test with fishes of the tropical eastern Pacific. Ecology 86: 1771 1782. Nathan, R. and Werner, Y. L. 1999. Reptiles and breeding birds on Mt. Hermon: patterns of altitudinal distribution and species richness. Israel J. Zool. 45: 1 33. Oommen, M. A. and Shanker, K. 2005. Elevational species richness patterns emerge from multiple local mechanisms in Himalayan woody plants. Ecology 86: 3039 3047. Rahbek, C. 1995. The elevational gradient of species richness: a uniform pattern? Ecography 18: 200 205. Rahbek, C. 1997. The relationship between area, elevation, and regional species richness in Neotropical birds. Am. Nat. 149: 875 902. ECOGRAPHY 29:6 (2006)

Rahbek, C. 2005. The role of spatial scale and the perception of large-scale species-richness patterns. Ecol. Lett. 8: 224 234. Rangel, T. F. L. V. B. et al. 2006. Towards an integrated computational tool for spatial analysis in macroecology. Global Ecol. Biogeogr. 15: 321 327. ´ . et al. 2005. Energy, water and large-scale Rodrı´guez, M. A patterns of reptile and amphibian species richness in Europe. Acta Oecol. 28: 65 70. Schall, J. J. and Pianka, E. R. 1977. Species densities of reptiles and amphibians on the Iberian Peninsula. Don˜ana. Acta Vertebr. 4: 27 34. Schall, J. J. and Pianka, E. R. 1978. Geographical trends in numbers of species. Science 201: 679 686.

Va´zquez, L. and Gaston, K. J. 2004. Rarity, commonness and patterns of species richness: the mammals of Mexico. Global Ecol. Biogeogr. 13: 535 542. Watkins, J. E. Jr et al. 2006. Vascular epiphyte distribution patterns: explaining the mid-elevation richness peak. J. Ecol. 94: 144 156. Whittaker, R. J. et al. 2001. Scale and species richness: towards a general, hierarchical theory of species diversity. J. Biogeogr. 28: 453 470.

Subject Editor: Carsten Rahbek.

Download the appendix as file E4802 from B/www.oikos.ekol.lu.se/appendix /.

ECOGRAPHY 29:6 (2006)

927

ECOGRAPHY 15: 177-183. Copenhagen 1992

Latitudinal and elevational variation in fruiting phenology among western European bird-dispersed plants Marcelino Fuentes

Fuentes, M 1992 Latitudinal and elevational variation in fruiting phenology among western European bird-dispersed plants - Ecography 15 177-183 I try to test the prediction that bird-dispersed plants should produce fruits when fruit-eating birds are most abundant by reviewing some phenological data of fleshy fruit production in westem Europe The prediction that fruit ripening dates m populations of the same species should occur later at lower latitudes and elevations, to coincide with the maximum abundance of fruit-eating birds, is not supported by the data The patterns of seasonal variation in the total number and biomass of fruits, but not in the proportion of species in fruit, in communities at different latitudes and elevations do coincide with patterns of seasonal abundances of avian frugivores 1 suggest that this coincidence is due to the greater relative abundance (and contribution to total fruit production) in each locality of species that fruit at times of the year when birds are most abundant These species may have achieved a demographic advantage by getting more seeds dispersed than species that ripen fruits in other seasons M Fuentes, Area de Ecologia, Facuttad de Biologia, Umv de Santiago, E-15071 Santiago de Compostela, Spam

The second seasonal pattern of bird abundance is found in the Mediterranean basin lowlands, where large In areas where strong seasonal fluctuations of bird numbers of migrants (including many frugivorous speabundance occur, it is expected that, m general, bird- cies) coming from northern and higher elevation areas dispersed plants should tend to produce fruits at times join sedentary populations m autumn-winter, producing of the year when fniit-eating birds are most abundant a peak in bird numbers dunng this season (see Herrera (Snow 1971, Thompson and WiUson 1979, Stiles 1980) 1982, 1984, Costa 1984, Jordano 1985, Cuadrado 1986 Two distinct patterns of seasonal variation in the for southem Spain, Luis and Purroy 1980 for the Baabundance of fruit-eating birds have been reported for leanc Islands, Debussche and Isenmann 1989 for southwestern Europe. One is typical of high latitudes and em France, Lovei 1989 for southern Italy) In these elevations and is characterized by minimum abundances wintenng grounds fruit-eatmg bird species make up a from September-October until spnng, as a consequence larger proportion of bird communities than in more of birds migratmg towards warmer regions (see Hogstad northem and mountainous areas (Jordano 1985, Telle1%7, Alatalo 1978 for Scandinavia, Eybert 1973 for ria et al. 1988, see also Willson 1986 for eastern North northern France; Frochot 1973, Cordonnier 1976 for America). A pattern similar to the first one has been found for central France; Purroy 1975, Santos and Suarez 1983 for northern Spam; and Pens Alvarez 1984, Guitian 1984, the northeastern part of North America and one similar Obeso IWI for mountainous areas of the Ibenan Penin- to the second one for the southeastem part of North sula). In more southem localities or at lower elevations, America and around the Gulf of Mexico (Thompson the autumn decrease m bird numbers occurs later, since and WiUson 1979, Stiles 1980, Skeate 1987). Some authors have proposed that, to enhance the the climatic conditions are more favorable.

Introduction

Accepted 24 January 1991 ECOORAPHY 12 ECOGRAPHY 15 2 (1992)

177

Spnng migrant birds scarcely eat these fruits (see above references), which may be seen as surplus of winter production I do not think their case is worth discussing in the context of the predictions tested here For objective (2) I have used all the phenological studies of western European plant communities I could find For objective (3) I have represented all the published data I could find on the phenology of total fruit abundance, except those offered by Guitian (1984) and Jordano (1984) Their data are only refered to in the text because from the information provided by them, I was not able to draw graphics similar to the ones used throughout this paper Obeso (1985) found similar fruit production phenologies in two study sites in the same locality I have also found similar phenologies in my two sites of La Barosa For brevity, I have represented only one of each data sets I estimated the curves for the phenology of total fruit biomass by combining data of individual species phenologies and average masses for each fruit species When fruit masses were not given in the onginal sources I used data from Herrera (1987) I could not construct biomass curves for Obeso's (1985) data because of the difficulty m computing individual Sfjecies phenologies from his graphics Different authors used different methods to record fruiting phenology and, theoretically, there is a possibility of obtaining patterns that are but artifacts derived from inconsistent methodologies The authors used difMethods ferent methods to calculate the dates of peaks and of For objective (1), above, I have compared the dates whole periods of npe fruit availability, l e fruit counts when there is a maximum of npe fruit among different in tagged plants, fruit counts in permanent plots, or populations of the same species I have considered all qualitative observations as in Debussche et al (1987) published information, provided the dates were given and Snow and Snow (1988) Most authors considered a with sufficient accuracy (e g month), as well as some species to be "in fruit" when some npe fruit were preunpublished data that I collected m two northwestern sent in the plants, while others required that a minimum Spanish spiny shrublands (La Barosa) These shrub- proportion be npe, e g 10% of the total crop Delands consisted mainly of Crataegus monogyna, Rosa bussche et al (1987) and Snow and Snow (1988) commicrantha, Rubus ulmifolius, Quercus ilex and Lonicera bined information from wide regions over several years, etrusca For latitudinal variation I have compared local- while the rest of the authors obtained their data in ities at low elevations (< 600 m a s I) For elevational limited areas with relatively homogeneous vegetation, variation I have compared different sites withm re- and gave fruit npenmg dates for different years sepstricted regions (particularly the Sierra de Cazorla area, arately The data reviewed to test prediction (3) were in southeastern Spam). In the Results, I have consid- obtained using very similar methods, except that Soreered that two npt-travi availability peak dates were nsen's (1981) study site was more extensive and indifferent when they were separated by at least a fort- ternally more diverse than those of the rest of the authors. Sorensen (1981) and Herrera (1984 for El Viso) night calculated the abundance of the different fruit species Fruit consumption by birds is well documented for all by combining phenological data and fruit crop estimates the plant species considered (see Herrera 1984, Jordano from tagged plants, and cover estimates for each plant 1984, Guitian 1984, Obeso 1985, Snow and Snow 1988, species. Herrera (1984 for Hoyos de Munoz) periodFuentes 1990, and references therein), except for Rus- ically counted all fruits present in permanent plots. The cus aculeatus, which is nevertheless included because it other studies (mcluding my own unpubl. one of La presents all the typical characteristics of the fruits eaten Barosa) combmed phenological data from tagged by birds (Herrera 1987,1989). Mammals are considered plants, and abundance data from fruit counts m ranto have a minor impact, m comparison to birds, on the domly placed plots. In all cases, I believe that the methseed dispersal of most plant species reviewed (see above odological differences between authors are very slight references) Sometimes, fruits of certam winter-npen- and certainly not hkely to produce, per se, the kind of mg species remain in good condition until early spnng. probability of seed dispersal by frugivorous birds, plants in the northern regions should fruit in early autumn, when many birds switch from an insectivorous diet to a mixed or almost completely frugivorous diet, and before most birds leave for southern areas (Morton 1973. Thompson and Willson 1979, Stiles 1980, Herrera 1982, 1984) Maximum fruit production should occur gradually later at lower latitudes, and take place in winter in bird wintering grounds such as the Mediterranean basin or around the Gulf of Mexico (Snow 1971, Thompson and Willson 1979, Stiles 1980, Skeate 1987, Snow and Snow 1988) A similar reasoning can be applied to elevational gradients fruit ripening should occur later at lower elevations In this paper I will try to test this hypothesis by reviewing the phenological data of fleshy fruit production in western Europe Specifically, I will test the predictions by examining 1) the fruit ripening dates in populations of the same species, 2) the patterns of seasonal variation in the proportion of species bearing npe fruit m different plant communities, and 3) the patterns of seasonal vanation m total fruit abundance (measured in number and biomass of fruits) in different plant communities, situated at different points along latitudinal and elevational gradients

178

ECOGRAPHY 15.2 (1992)

Table 1 Latitude and elevation (in m a s 1 ) of the localities considered Locality

Latitude

Elevation (m a s I )

5r45'

< 200 < 100

Snow and Snow 1988 Sorensen 1981

43''40'

50-950

Debussche and Isenmann 1985 Debussche et al 1987

42''37' 42''50'

600 1350

Fuentes 1990 Guitian 1984

37°26' 37" 9' 37° 1' 37°59' 37''56'

100 10 10 1150 1350

Herrera Jordano Herrera Herrera Obeso

South England Aylesbury Whytham South France Montpellier Cedex North Spain La Barosa Ancares South Spam El Viso Hato Raton Donana Hoyos de Munoz Roblehondo

latitudinal and elevational patterns expected to support or refute the predictions stated above The latitude and elevation of the different localities and the author references are given m Table 1 The annual cycles of abundance of fruit-eatmg birds m the regions considered agree with the patterns outlined in the Introduction, l e high numbers in summer and early autumn, and low in winter, in the more northern regions and mountains, and the reverse in circummediterranean lowland sites (Snow 1971, Herrera 1984, Jordano 1985, Obeso 1985, Debussche and Isenmann 1989, Fuentes unpubl data but see Guitian 1984)

Reference

1984 1984 1986 1984 1985

Therefore, the set of localities is reasonably adequate for testing the hypothesis

Results Fruit ripening dates in different populations of the same species Only 3 {Asparagus acuttfohus, Crataegus monogyna and Ruscus aculeatus) out of 20 species considered suppiort the prediction that southern populations would ripen

Table 2 Fruiting peaks of populations of single species growing at different latitudes The bars separate dates of different years The number indicates first or second fortnight of^ the month

Asparagus acuttfohus Bryonia cretica Cornus sanguinea Crataegus monogyna Daphne gnidium Euonymus europaeus Hedera helix Lomcera periclymenum Osyrts alba Phillyrea angusttfolia Pistacm lentiscus Prunus spinosa Rhamnus alaternus Rosa canina Rubia peregnna Rubus ulnufoltus Ruscus aculeatus Sambucus nigra Snulax aspera Tamus communis

South England (Whytham)

South France (Montpellier)

North Spam (La Barosa)

South Spain (El VISO) (Hato Raton)

Oct Jul Sep Sep

_ -

_ — Sep

2Oct Sep* 1 Oct -

1 Dec 1 Apr 2 Sep — _

2 Oct _

2 Jan 2 Sep + lOct _

lNov

Sep Nov Jan Sep Oct Nov Oct Jul Oct Oct Aug Oct Aug Oct Sep

2 Aug 1 Sep/1 Oct 2 Oct —

1 Sep — -

2 Aug — —

2 Nov/1 Nov 2 Sep — 2 Sep

Dec Jun — Oct

2 Nov/1 Dec 2 Sep/1 Aug

1 Dec/1 Nov 2 Aug/2 Sep -

2 Aug/1 Jul 2 Sep/1 Sep 2 Jul/2 Aug 2 Nov/2 Oct — -n Sep 2 Aug/1 Aug 2 Nov/1 Dec 1 Oct/1 Oct 2 Aug/1 Jul

' Aylesbury + Rubus fhtticosus n* ECOGRAPHY 15 2 (1992)

179

Table 3 Fruiting peaks of populations of single species growing at different elevations in southern Spain The bars separate dates of different years The number indicates first or second fortnight of the month

Crataegus monogyna Daphne gnidium Juniperus oxycedrus* Lonicera implexa Phillyrea angustifolia Pistacia lentiscus Rhamnus alaternus Rosa canina Rubia peregrma Rubus utmifolius Smilax aspera

Roblehondo

H Mufioz

1 Sep -

_ Nov

Aug-Sep

Nov-Feb

— -

2 Aug 2 Oct

Oct Oct Nov Nov

Dec

2 Sep

Oct Nov

El Viso

Hato Raton 1 Dec/1 Nov 2 Aug/2 Sep

Aug-Oct

Sep Dec Jun Oct

—

2 Jul/2 Aug 2 Nov/2 Oct

-/I Sep 2 Aug/1 Aug 1 Oct/1 Oct

* the fruiting periods, not the peaks, are given

their fruits later than northern ones (Table 2) In 14 species, the southern populations ripen their fruits earlier than the northern ones Two species do not show latitudinal variation in fruiting phenology and one (Ptstacta lenttscus) shows no clear pattern

In 7 out of 11 species, the populations from higher elevation sites npen their fruits later than those of lower elevation ones, thus contrary to initial expectation (Table 3) Only m 3 cases does the reverse occur, and in one no clear pattern emerges In Rubus ulmtfohtis.

LOWLAND LOCALITIES

SOUTH ENGLAND (/Veotxry)

SOUTH ENGLAND (W»v»wni)

SOUTH FRANCE (MontpeHer)

NORTH SPAIN ( U Barosa)

SOUTH SWUN (Doflana)

SOUTH gptm {B vho)

HIGHLAND LOCALITIES

100

CC NORTH SBMN (Arcanis)

aOUTH SFMN (H. M l A » )

SOUTH SFMN (RoUahondD)

100

50 0

J J A S O N D J F M A M SOUTH SmN {Hmi Rattn 1881/82)

J J A S O N D J F M A M

SOUTH s m N (Han Rat6n 1882/83)

J J A S O N O J F M A M

Fig 1 Curves of the number of species (expressed as percent of the total number of species) beanng npe fruit m communities situated at different latitudes and elevations Thin hnes are the fruiting curves (number of fruits expressed as proportion relative to the peak) of all species combined See Table 1 for geographic information about these communities ECOGRAPHY 15 2 (1992)

100 A t

\ / /

, \ \ Y

LATITUDINAL VARIATION (Number)

sites of comparable latitude A similar trend was found by Herrera (1985, Fig 1) for three localities in the Sierra de Cazorla situated at 1150, 1350 and 1800 m a s 1 However, at < 100 m a s 1 the peaks occur on similar or earlier dates than at the 1150 m site

i z

100 LATITUDINAL VARIATION (Biomass)

Dates of maximum total fruit abundance

0 Z

100

ELEVATIONAL VARIATION (Number)

o Q.

otr

J J A S O N D J F M A M Fig 2 (A) Number of npe fruits, of all species combined, expressed as proportion relative to the maximum abundance of ripe fruits, m three lowland localities situated at different latitudes ( ) Whytham, southern England, (â&#x20AC;&#x201D;) La Barosa, northern Spam, ( ) El Viso, southern Spain (B) Biomass of ripe fnuts, of all species combined, expressed as proportion relative to the maximum quantity (in biomass) of npe fruits, in the same three locahties as in A (C) Number of npe fruits, all species combined, expressed as proportion relative to the maximum abundance of npe fnuts, in three localities situated at different elevations in southern Spain ( ) Roblehondo (1350 m), (â&#x20AC;&#x201D;) Hoyos de Munoz (1150 m), ( ) El Viso (100 m)

maximum npe fruit availability occurs at the same time in highland (Sierra de Ancares) and lowland (La Barosa) populations in northern Spain

Dates of manmum proportion of species in fruit in different communities No clear pattern emerges from latitudinal comparisons of the dates of maximum proportion of species in fruit (Fig 1). Furthermore, the vanability of dates of maximum proportion of species in fruit is often greater withm the same latitude tfian between latitudes. On the other hand, the maximum proportion of species m fruit occurs earlier in tfie two highest sites in the north and south of Spain than in the two lower elevation ECOORAPHY

When curves of the total fruit availability (in number. Fig 2A, and estimated biomass. Fig. 2B), of all species combmed, are compared for three localities at different latitudes, a pattern consistent with mitial predictions is seen The clearest differences are between southern Spam and southern England These differences also hold true when data from a Donana (southern Spam) scrubland are compared with data from England (Jordano 1985, Fig 1) The dates of maximum fruit number m the northern Spanish shrublands fall in an intermediate position between those of more extreme latitudes This also holds true for the dates of maximum fruit biomass in one of the sites, but not in the other, in which maximum biomass occurs on the same dates as in England (Fig 2B) The maximum number of fruit in a 1350 m a s 1 locahty in the Cazorla mountains occurs, earlier than in a 1150 m locality in the same region, as predicted However, there is no time span between the later and a 100 m elevation site m southwestern Spam (Fig 2C) In a northwestern Spanish highland site (Sierra de Ancares, 1400 m) the maximum fruit availability also occurs earlier than in a nearby lowland locality (La Barosa, 600 m) (see Guitian 1984, Fig 0 17)

Discussion Some {xjssible shortcomings of this study (such as the relatively small number of localities for which information IS available, the short duration of studies, or the broad variation observed among nearby locations and among years) make the conclusions somewhat provisional and prompt the need of further studies. However, some of the patterns revealed by this study deserve consideration. The theoretical expectation that, within species, fruiting peaks should occur later at lower latitudes is not supported by the data for western Europe Actually, the prevailing trend observed runs contrary to expectations, most species ripen their fruits earlier at lower latitudes. This contrasts with the assertion of Stiles (1980) for eastern North America. An analogous prediction for elevational vanation (later npenmg at lower elevations) IS also not supported by the data, and the reverse again app>ears to be the rule There is a possibility that np)enmg dates are adapted in each population to particular local conditions, such as abundance cycles of certain

181

frugivores that could depart from the rather general abundance patterns outhned m the Introduction However, this does not explain why fruit ripening tends to occur earlier in lower latitudes and elevations This finding may be better explained by some factor that vanes predictably with latitude and elevation Thus, the variation in fruiting phenology may be caused by differences in flowenng phenology (Pnmack 1985, 1987, Rathcke and Lacey 1985) or m the timing of fruit development and/or maturation (Rathcke and Lacey 1985, Duke 1990) induced by latitudinal and elevational climatic variation No clear pattern emerges when comparing the dates of maximum proportion of species in fruit among localities situated at different latitudes, contrary to the pattern reported for eastern North America (Thompson and Willson 1979, Skeate 1987), but peaks tend to occur earlier at higher elevations than at lower ones in southern Spam This indicates that many species that produce fruits when fruit-eating birds are relatively scarce are nevertheless able to persist successfully in plant communities The variation in the dates of maximum fruit availability among communities situated along latitudinal and elevational gradients agrees with the variation in the dates of maximum abundance of avian fruit consumers This match between fruiting and bird abundance is not accomplished, as we have seen, either by intraspecific vanation, or by variation m the frequencies of the species ripening their fruits at different times of the year (at least for latitudinal variation) A possible explanation is that the abundance of fruit-eating birds has shaf>ed the fruit availability curves through a demographic, rather than evolutionary, process At a given locality, the shape of the fruit abundance curve is usually determined by the disproportionate influence of a single, very abundant sf)ecies, which produces most fruits in the habitat In northern and high elevation communities the curve is dominated by an early-fruiting species {Sambucus ntgra in southern England (Sorensen 198f), Berberts hispantca in Sierra de Cazorla highlands (Obeso 1985)) and in southern and low elevation localities by a late-npenmg one (Ptstacta lenttscus in southern Spanish lowlands (Herrera 1984, Jordano 1984)) The greater abundance of these species may be, at least partly, due precisely to their fruiting when birds are most abundant m their particular localities These species may gain a demographic advantage by getting more seeds dispersed than species that njien fruits m other seasons, and thus they shift the community fniitmg curve in a way that matches fnigivore abundance phenology (see also Herrera 1985). This kind of ecological sorting of populations of different speaes, and not adaptive change withm species, may underly many allegedly coevolutionary adjustments among animals and plants (Janzen f985).

182

Acknowledgements - I thank J Guitian. C M Herrera. P Jordano and M F WiUson, for commenting upon earlier versions of the manuscript C M Herrera and L F Turnes kindly helped to improve my English 1 thank my parents for their financial support This study was funded by grants from Xunta de Gahcia (Tercer Cido and XUGA-8030789) and the Spanish Mimsterio de Educacion y Ciencia (FPI and PB 86-0453)

References Alatalo, R V 1978 Bird community energetics in a boreal coniferous forest - Holarct Ecol 1 367-376 Cordonnier. P 1976 Etude du cycle annuel des avifaunes par la methode des "points d'ecoute" - Alauda 44 169-180 Costa. L 1984 Composicion de las comunidades de aves en pmares del Parque Nacional de Donana (suroeste de Espana) - Donana Acta Vert 11 151-183 Cuadrado. M 1986 La comumdad de aves de un acebuchar del sur de Espana durante el penodo mvernal y de cria Donana Acta Vert 13 71-85 Debussche, M and Isenmann, P 1985 Frugivory of transient and wintering European robins Erithacus rubecula in a Mediterranean region and its relationship to ornithochory - Holarct Ecol 8 157-163 - and Isenmann. P 1989 Fleshy fruit characters and the choices of bird and mammal seed dispersers in a Mediterranean region - Oikos 56 327-338 - . Cortez. J and Rimbault, I 1987 Vanation in fleshy fruit composition in the Mediterranean region the importance of npenmg season, life-form, fruit type and geographical distribution - Oikos 49 244-252 Duke, N C 1990 Phenological trends with latitude in the m&ngxo\c XKS Avicennia marina - J Ecol 78 113-133 Eybert, M C 1973 Le cycle annuel des oiseaux dans trois stades evolutifs d'une pmede de Bretagne - Terre Vie 27 507-522 Frochot, B 1973 L'evolution seasonniere de l'avifaune dans Una futaie de chenes en Bourgogne - Terre Vie 25 145182 Fuentes, M 1990 Relaciones entre pajaros y frutos en un matorral del norte de Espana variaciones estacionales y diferencias con otras areas geograficas - Ardeola 3753-66 Guitidn, J 1984 Ecologia de la comumdad de passenformes de un bosque montano de la Cordillera Cant^brica occidental - Ph D Thesis, Univ Santiago Herrera, C M 1982 Seasonal vanation in the quahty of fruits and diffuse coevolution between plants and avian dispersers - Ecology 63 773-785 - 1984 A study of avian frugivores. bird-dispersed plants, and their interaction m Mediterranean scrublands - Ecol Monogr 54 1-23 - 1985 Habitat-consumer interactions m frugivorous birds In Cody, M L (ed ), Habitat selection in birds Academic Press, Orlando, FL, pp 341-365 - 1987 Bird-dispersed plants of the Iberian Peninsula a study of fruit characteristics - Ecol Monogr 57 305-331 - 1989 Frugivory and seed dispersal by carnivorous mammals, and assoaated frtut characteristics, in undisturbed Mediterranean habitats - Oikos 55 250-262 Herrera, J 1986 Flowenng and fruiting phenology in the coastal shniblands of Donana, south Spam - Vegetatio 68 91-98. Hogstad, O 1%7 Seasonal fluctuation in bird populations within a forest area near Oslo (southern Norway). - Nytt Mag Zool 15:81-96 Janzen, D.H 1985 On ecological ftttmg -Oikos45 308-310 ECOGRAPHY 15 2 (1992)

Jordano, P 1984 Relaciones entre plantas y aves frugi'voras en el matorral mediterraneo del area de Donana - Ph D Thesis, Umv Sevilla - 1985 El ciclo anual de los pasenformes frugi'voros en el matorral mediterraneo del Sur de Espana lmportancia de su invernada y vanaciones interanuales - Ardeola 32 69-94 Lovei, G L 1989 Passerine migration between the Palaearctic and Africa - Current Ornithology 6 143-174 Luis, E and Purroy, F J 1980 Evolucion estacional de las comunidades de aves en la lsla de Cabrera (Baleares) Studia Oecol 1 181-223 Morton, E S 1973 On the evolutionary advantages and disadvantages of fruit eating in tropical birds - Am Nat 107 8-22 Obeso, J R 1985 Comunidades de Passenformes y frugivorismo en altitudes medias de la Sierra de Cazorla - Ph D Thesis, Umv Oviedo - 1987 Comunidades de passeriformes en bosques mixtos de altitudes medias de la Sierra de Cazorla - Ardeola 34 37-59 Pens Alvarez, S J 1984 Avifauna mvernante y nidificante en la Sierra de Bejar (Sistema Central, Provincia de Salamanca) - Studia Oecol 3 219-230 Primack, R B 1985 Patterns of flowering phenology in communities, populations, individuals, and single flowers - In White, J (ed ), The population structure of vegetation Dr W Junk, Dordrecht, The Netherlands, pp 571-593 - 1987 Relationships among flowers, fruits, and seeds - Ann Rev Ecol Syst 18 409-430

ECOGRAPHY 15 2 (1992)

Purroy, F J 1975 Evolucion anual de la avifauna de un bosque mixto de coniferas y frondosas en Navarra - Ardeola 21 669-697 Rathcke, B and Lacey, E P 1985 Phenological patterns of terrestrial plants - Ann Rev Ecol Syst 16 179-214 Santos, T and Suarez, F 1983 The bird communities of the heathlands of Palencia The effects of coniferous plantations - Proc VII Int Cong Bird Census Work, pp 172179 Skeate, S T 1987 Interactions between birds and fruits in a northem Florida hammock community - Ecology 68 297309 Snow, B and Snow, D 1988 Birds and berries - T & AD Poyser, Waterhouses, England Snow, D W 1971 Evolutionary aspects of fruit-eatmg by birds -Ibis 113 194-202 Sorensen, A E 1981 Interactions between birds and fruits in a British woodland - Oecologia (Berl ) 50 242-249 Stiles, E W 1980 Patterns of fruit presentation and seed dispersal in bird-dissemmated woody plants in the eastern deciduous forest - Am Nat 116 670-688 Telleria, J L , Santos, T and Carrascal, L M 1988 La invernada de los pasenformes (Orden Passenformes) en la Peninsula Ibenca - In Telleria, J L (ed ), Invernada de aves en la Peninsula Ibenca SEO, Madrid, pp 153-166 Thompson, J N and Willson, M F 1979 Evolution of temperate fruit/bird interactions phenological strategies Evolution 33 973-982 WiUson, M F 1986 Avian frugivory and seed dispersal in eastern North Amenca - Current Ornithology 3 223-279

183

Ecography 34: 85 93, 2011 doi: 10.1111/j.1600-0587.2010.06250.x # 2011 The Authors. Journal compilation # 2011 Ecography Subject Editor: Francisco Pungnaire. Accepted 18 January 2010

Demographic processes of upward range contraction in a long-lived Mediterranean high mountain plant Luis Gimeńez-Benavides, Marıá Jose´ Albert, Jose´ Marıá Iriondo and Adriań Escudero L. Gimeńez-Benavides (luis.gimenez@urjc.es), M. J. Albert, J. M. Iriondo and A. Escudero, A´rea de Biodiversidad y Conservacioń, Univ. Rey Juan Carlos-ESCET, Tulipań s/n. ES-28933 Mo´stoles, Madrid, Spain.

We analyzed demographic data of a long-lived high mountain Mediterranean plant, Silene ciliata Poirret, over a 4-yr period. Selected populations were located at contrasting altitudes at the southernmost margin of the species (Sierra de Guadarrama, central Spain), representing a local altitudinal range at the rear edge of its overall distribution. Previous studies have suggested that differences in the reproduction and performance of individuals at upper and lower populations may have implications for population dynamics. We used matrix analysis to assess their demographic behaviour. Life Table Response Experiments were used to identify the life history stages most relevant to observed differences in population growth rates between populations. Transition matrices revealed great spatio-temporal variability in demographic traits. Seedling recruitment was very low each year in all populations. Maximum longevity of S. ciliata individuals in the lower peripheral population was much lower compared to the central population, probably due to higher adult mortality. Population growth rate (l) showed a declining trend at the lowest altitude and a relatively stable trend at the central population. Long-term simulations also indicated a great risk of quasi-extinction at the lowest population. Our results suggest that rear edge populations of S. ciliata at Sierra de Guadarrama are suffering demographic processes that may be leading to the latitudinal displacement of the species’ range.

Peripheral populations of plant species are important reservoirs of intraspecific genetic diversity and evolutionary potential (Lesica and Allendorf 1995, Hampe and Petit 2005, Jump and Pen˜uelas 2005), as well as functional drivers of ecosystem stability (Eriksson 2000). Studies of individual plant performance across a species’ range frequently find lower survival and/or reduced fecundity at range margins compared to the range center (Jump and Woodward 2003, Gimeńez-Benavides et al. 2007a, b, Marcora et al. 2008). However, a major concern is whether reductions in fitness components really affect population growth and persistence. In fact, many times the persistence of plant species is not crucially dependent on reproductive success and seedling establishment (Pico and Riba 2002, Garcıá 2008, Iriondo et al. 2008). Differences in life-history traits, such as life-span, will largely determine the species dependence on sexual regeneration. In long-lived perennial species, the impact of limited seed output on population maintenance is difficult to determine due to the complexity of recruitment, but a tradeoff between sexual regeneration and persistence of already established individuals has been suggested (Garcıá and Zamora 2003). Typical examples of persistence due to longevity and/or vegetative reproduction are more frequent in stressful and unstable environments such as arid, alpine

and rocky habitats (Grime 2001, Garcı´a and Zamora 2003) where geographical limits of reproduction do not necessarily coincide with actual range limits (Gaston 2003). Therefore, to get a complete view of the factors shaping geographical range limits, the components of individual performance must be integrated into population dynamic models across species’ distributions (Angert 2006, Foden et al. 2007). This task is especially relevant today, when ongoing climate warming and other anthropogenic impacts, such as habitat fragmentation and changes in land use, are currently threatening peripheral populations. Despite this, there are few detailed comparisons of population dynamics of central vs marginal populations (Nantel and Gagnon 1999, Stokes et al. 2004, Angert 2006, Samis and Eckert 2007). Moreover, while demographic and evolutionary traits underlying the expansion of species at their leading edge have been more extensively studied during the last few decades (Petit et al. 2004), population dynamics responsible for range contractions at the rear margins of species’ distributions have not received sufficient attention (Hampe and Petit 2005). This general lack of mechanistic studies is even greater in high mountain environments, even though they provide an excellent opportunity for the study of range margins (Angert 2006, Ko¨rner 2007).

In mountain plants, conditions for regeneration and survival are hierarchically arranged within their distribution range. Firstly, they are more suitable in the latitudinal centre of their distribution area than in the periphery. Secondly, they also appear structured within each mountain island. Similar to latitudinal range displacements, recent altitudinal shifts in the abundance and distribution of species inhabiting mountain environments have been documented during the last few decades. Several studies have revealed contemporary alterations of species richness in high summits (Grabherr et al. 1994, Gottfried et al. 1999, Virtanen et al. 2003, Walther et al. 2005, Pauli et al. 2007, Erschbamer et al. 2009). The migration of lowland plant species to higher elevations forces subsequent displacements of alpine species (Theurillat and Guisan 2001). Altitudinal shifts in vegetation belts and distribution ranges of species have already been documented (Walther et al. 2002, Klanderud and Birks 2003, Pen˜uelas and Boada 2003, Lesica and McCune 2004). Therefore, there is an urgent need for accurate forecasting of the consequences of this process. Important progress on species’ distribution modelling has recently been made, but most of these models do not explicitly take into account the essential mechanisms operating at individual and population scales (Thuiller et al. 2008, Morin and Thuiller 2009). These factors may cause important bias and inaccuracy in current projection models, and are a probable cause of the divergence found among coarse and fine resolution models when compared (Trivedi et al. 2008). At least at the population level it seems basic to monitor and model demographic trends at rear populations. Despite this necessity, demographic studies of high mountain plants are still scarce compared to those of lowland species, and few studies have documented the populationlevel dynamics driving altitudinal displacements (Doak and Morris 1999, Diemer 2002, Angert 2006). In the present work, we analyzed demographic data of a long-lived Mediterranean high mountain plant, Silene ciliata (Caryophyllaceae), at different altitudes within its southernmost margin of distribution (central Spain). During the last 45 yr, mean air temperature has increased by 1.88C in this area, and days of snowcover per year have decreased by 19.7 d (Gime´nez-Benavides et al. 2007a). In addition to direct impacts of climate warming on the species’ performance, the area has suffered a substantial bottom-up shrub encroachment (Sanz-Elorza et al. 2003) with potential consequences for the persistence of S. ciliata rear populations. The final objective of the present work is to assess whether the reproductive and recruitment failure observed at the rear edge of the species (Gime´nez-Benavides et al. 2007a, 2008) results in regressive population dynamic in the lower population compared to higher altitude populations. We argue that the breakdown of sexual regeneration at lower limits could only be balanced out by a long lifespan and reduced adult mortality. Otherwise, the species could suffer a high risk of peripheral extinction and altitudinal range contraction under the present global warming context. We analyzed demographic performance of the species over a 4-yr period using transition matrix models and long-term simulations. Specifically, the questions addressed were: 1) are S. ciliata populations at the rear altitudinal edge experiencing a declining population trend? 2) Are the upper and lower populations along an altitudinal gradient driven by the 86

same demographic processes? And, if they differ, 3) what are the vital rates responsible for the observed differences in population growth rates at different altitudes? 4) Do populations at different altitudes differ in their probability of quasi-extinction in the long term?

Methods Plant species and study site Silene ciliata (Caryophyllaceae) is a long-lived perennial plant that grows in main mountain ranges of the Balkan Peninsula, the Appenines, the Massif Central in France and the northern half of the Iberian Peninsula, covering a latitudinal range from 408N to 468N (Tutin et al. 1995). In mountain ranges of central Spain, where the species reaches its southernmost margin, it grows from 1900 m (treeline zone) up to the highest summits (ca 2600 m). The species typically grows in a compact cushion-shape. Its flowering period extends from late June to early-mid September. Flowering stems (1 33 per adult plant) are 15 cm in height and bear 1 5 flowers. Hand-crossing experiments indicate that S. ciliata is a self-compatible species. However, passive autogamy is restricted by a pronounced protandry so it requires pollinators (Gimeńez-Benavides et al. 2007a). Although many alpine species are highly clonal (Forbis 2003), no evidence of vegetative propagation was observed in this species when several individuals were dug up (Gimeńez-Benavides unpubl.). The study area was in the Sierra de Guadarrama (Penãlara Natural Park), a mountain range located in central Spain, 50 km north of Madrid city (408N, 38W). Mean annual precipitation at Navacerrada Pass weather station (1800 m, 8 km southwest of the study site) is 1350 mm, and is concentrated from late autumn to early winter. A marked drought season occurs from late May to October (Fig. 1). Snowfall generally begins in October and the snow-free season begins in May June (Palacios et al. 2003). This work is part of a broader study of factors controlling the distribution and performance of S. ciliata along the local altitudinal gradient. Three populations were selected for this demographic approach. The first population was located in the vicinity of Laguna Chica (hereafter Laguna), a small glacial lake situated in a moraine deposit in the treeline zone (1970 m). Vegetation is dominated by a dense shrub cover of Cytisus oromediterraneus and Juniperus communis subsp. alpina intermingled with a low-dense stand of Pinus sylvestris. Here, S. ciliata is displaced by the shrub species and only grows in small, isolated pasture patches dominated by Festuca curvifolia. The second population was on the Dos Hermanas peak (hereafter Dos Hermanas), a summit flat area situated at 2250 m, dominated by a Cytisus-Juniperus shrub formation and patchy xerophytic fellfields of Festuca curvifolia. This fellfield community bears extreme winds and a relatively short snowcover period, and is characterized by the abundance of cushion plants (Escudero et al. 2004). The third population was located at the summit of the highest peak of the mountain range, the Penãlara peak (hereafter Penãlara), at 2440 m. This area is dominated by the Festuca fellfield and shrub species are scarce. As a consequence of

Figure 1. Climatic data at the Navacerrada Pass weather station (40846?N, 4819?W; 1860 m, located 8 km south-west of study sites). Columns and lines represent monthly mean precipitation and temperature, respectively, during the period 1946 2006 and in the study years, 2003 to 2006.

reported temperature increase and reduction of snow cover, probably combined with a moderate reduction in livestock grazing, the Cytisus-Juniperus shrub belt is encroaching and replacing the Festuca cryophilic pastures colonized by S. ciliata (Sanz-Elorza et al. 2003). Census scheme Data were collected from 2003 to 2006. One permanent plot was established in each population for demographic monitoring. Plot size varied between populations due to differences in microhabitat characteristics and plant density but populations were larger enough and similar in slope and orientation to be considered representative at each altitude considered. At Laguna (1970 m), we initially monitored all plants available within a small population (128 individuals in a 9 m2 plot), while at Dos Hermanas (2250 m) and Penãlara (2440 m) we tagged 266 (7 m2 plot) and 168 individuals (10 m2 plot), respectively, within a larger, continuous population. All plants found within each plot were mapped to allow subsequent location. Plants were monitored every year at the end of the reproductive season (September October). Plant size was estimated as maximum cushion diameter. Total number of inflorescences per plant was counted in a single visit as an estimate of reproductive output. Previous studies suggest that inflorescence number is a good surrogate of fruit production (Pearson’s r 790, 549 and 580, pB0.0001, for Laguna, Dos Hermanas and Penãlara respectively, Gimeńez-Benavides et al. 2007a).

Silene ciliata seedlings emerge at the beginning of the growing season, suffering extremely high mortality during summer (Gime´nez-Benavides et al. 2007b). Thus, all seedlings found within the plots at the end of the growing season were registered as a basis for estimating annual recruitment rates. Unfortunately, the plot at Pen˜alara was vandalized during the second year of study, so parameter estimations of only one transition could be obtained. The possible uncertainty in plant population structure and demographic parameters derived from the establishment of a single plot per population was assessed by comparing them with five extra plots of 5 m2 randomly placed at each population. Plant size of every plant in these extra plots was measured in 2005, and size structures of the permanent plots were compared with those of the respective extra plots from the same altitude by cross-tabs and chi-square test. Stage classification We established five stage classes, one seedling class and four reproductive classes. The first corresponded to seedlings of about 1 cm in diameter that germinated in spring. Seedlings that survived into the next growth period grew into a reproductive class. Reproductive classes were obtained by classification of individuals by k-means clustering (Hartigan 1975), except seedlings, using pooled data from all populations: small (plants of 1.5 2.5 cm diameter), medium (3 4.5 cm), large (5 8 cm) and extra-large (]8.5 cm). 87

Matrix construction We constructed Lefkovitch matrices for each population and time interval using estimates of inflorescence production, seedling recruitment and transition probabilities between reproductive stages. Transitions were built from the underlying vital rates (survival, growth and fertilitity), following Morris and Doak (2002). When sample size of some stages was small (n B13) mainly seedlings and small reproductive plants at Laguna survival rates were obtained from average transition frequencies across all years for each population (Menges and Dolan 1998, Angert 2006). The reproduction terms in the matrix were estimated as follows. Mean number of inflorescences per class was used to calculate the proportional contribution of each adult class to total reproductive effort. Thus, the reproduction term for each reproductive class in each transition was estimated following the equation: Fi;t ;t 1

sdlt 1 Ri;t ; 4 X (Ri;t ni;t ) i 1

where Fi,t,t 1 is the reproduction matrix element of class i (small, medium, large or extra-large) for the period t to t 1, sdlt 1 is the total number of seedlings censused in the population in time t 1, Ri,t is the proportional contribution of class i to total reproductive effort, and ni,t is the number of individuals in class i surviving at time t. Thus, seedlings censused in the following year were allocated among the four reproductive classes according to their proportional reproductive effort in the previous year. These estimations assumed that seedlings in time t 1 germinated from seeds produced in time t, as occurs in the absence of a permanent soil seed bank. Field and lab assays

showed that germination capacity of S. ciliata seeds can reach 100% over a one year period (GimeÂ´nez-Benavides et al. 2005, 2007b). Thus, we assumed that soil seed bank does not play an important role in the dynamics of S. ciliata populations. A recent seedbank study conducted on this fellfield community also supports the absence of a permanent seedbank in this species (GarcÄąÂ´a-Camacho 2009). Transition matrix models project population size according to the equation: x(t 1) Ax(t ); where x is a vector of the number of individuals in different plant stages (stage distribution at time t) and A is a matrix of probabilities and fertilities that defines the survival and reproduction of individuals in each stage between time t and time t 1 (i.e. matrix elements). The transition matrix A is derived from a life cycle graph that shows the possible transitions between stages (Caswell 2001). The life cycle graph for our model system is shown in Fig. 2. The dominant eigenvalue for each transition matrix was used to calculate the finite rate of increase, l. Bootstrapped matrices were generated by randomly sampling individuals with replacement within stage classes, using a Matlab routine (Matlab 7.0, MathWorks). Two-thousand bootstrapped transition matrices were used to obtain bias-corrected 95% confidence intervals for l. Maximum plant longevity was estimated for each population separately following Forbis and Doak (2004). A starting vector of one seedling and zero reproductive individuals was multiplied by the mean matrix with all fecundities set to zero (Caswell 2001). Year by year, the resulting vector was multiplied by the mean matrix until the summed probability of survival for all stage classes reached 0.01.

Figure 2. Life cycle of Silene ciliata. Each arrow represents a one-year transition. S denotes survival (stasis in the same class, and growth or regression to a different class) and F fertility.

Life Table Response Experiments

X (l ) (dh) 1l (ai;j ai;j ) l(l ) :l(dh) 1ai;j A m i;j

where l(dh) is the population growth rate in Dos Hermanas, and Am is the mean matrix from both populations. The first factor of the sum (in parenthesis) denotes the differences, and the result of multiplying them by the appropriate sensitivity are the contributions (Caswell 2001). We carried out an LTRE analysis for each time interval. All matrix analyses were performed with Matlab 7.0. Probability of quasi-extinction We calculated the probability of reaching a quasi-extinction threshold for each population by computer simulations. We used the Matlab routine developed by Morris and Doak (2002), which estimates the quasi-extinction time cumulative distribution function for a structured population in a stochastic environment. Proportions of individuals were equal to the initial stage structure of each population. Environmental stochasticity was implemented by using the three available transition matrices for each population. It was based on the climatic variability among the periods of time considered. The years 2004, 2005 and 2006 had milder temperatures and summer rainfall (data not shown, Navacerrada Pass weather station, 40846?N, 4819?W, 1860 m), whereas in 2003 the most severe European heatwave took place (Scha¨r et al. 2004) (Fig. 1). Maximum temperatures of the complete climate series were reached in summer, no rainy days were recorded in July, and only one in August (Gime´nez-Benavides et al. 2007a). Therefore, two contrasting scenarios were considered: in the ‘‘realistic’’ scenario the probability of occurrence of an extremely warm and dry year (first transition matrix) was fixed to 0.1 whereas in the ‘‘warming’’ scenario this probability was substantially increased by making the three transition matrices equally probable. We set the quasiextinction threshold at 50 individuals and time horizon at 250 yr. Pen˜alara was not included in the simulations because only one transition was available. The starting population density was the number of individuals of the initial plots, i.e. 128 and 266 individuals for Laguna and Dos Hermanas, respectively.

40 Individuals (%)

We used Life Table Response Experiments (LTREs) to identify the matrix elements most relevant to the observed differences in population growth rate between populations (Caswell 2001). We focused on a one-way fixed design where the two populations (Laguna and Dos Hermanas) were of interest in themselves. We chose the Dos Hermanas matrix as the reference matrix, i.e. a baseline for measuring population effects. Thus, population growth rate in Laguna was defined as:

Laguna (Low) Dos Hermanas (Middle) Peñalara (High)

0 Seedling

Small

Medium

Large

Extra-large

Stage classes

Figure 3. Size structure of S. ciliata populations according to stage classes established in transition matrices. Data from the three permanent plots set in 2003. n 128 individuals in Laguna, 266 individuals in Dos Hermanas and 168 individuals in Pen˜alara.

large and extra-large classes. On the contrary, Dos Hermanas and Pen˜alara populations were consisted of smaller plants, with a majority of medium and large plants and B15% of extra-large individuals. The seedling class was scarcely represented in all populations. The results obtained in our permanent plots did not differ in terms of size structure from the surrounding extra plots (low: x2 13.4, p 0.640, medium: x2 22.2, p 0.332, high: 19.4, p 0.248), thus the permanent plots could be considered representative of their corresponding population. Matrix analyses Transition matrices also revealed great differences in demographic traits between populations (Supplementary material Table S1). Adult survival was higher at Dos Hermanas for all stage classes and all time intervals. Mean survival per class increased at Dos Hermanas from small to extra-large plants, while mean survival was higher for large plants at Laguna. Seedling survival showed high variability between years and populations. The climatically extreme year (2003 2004 transition) showed higher mortality only for seedling and extra-large plants in Laguna (Supplementary material Table S1). Population growth rates (l) showed a declining trend at the lowest altitude and relatively stable population dynamics at the central and highest population. Bootstrapping confidence intervals of l were much wider at Laguna than at Dos Hermanas (Fig. 4). Estimated maximum plant longevity inferred from average transition matrices was 147 yr at Dos Hermanas and only 23 yr at Laguna. Life Table Response Experiments

Results Plant size structure showed great differences between populations (Fig. 3). Laguna population was dominated by large plants, with 70% of individuals belonging to

LTRE analyses revealed greater differences in matrix elements between populations for the seedling stage, in addition to greater variability in differences (Fig. 5a). However, the higher seedling survival at Laguna slightly contributed to differences in l between populations (Fig. 5b). Lower 89

1.10 Laguna (Low) Dos Hermanas (Middle) Peñalara (High)

0.8 0.6

2003–04 2004–05 2005–06

0.4 LTRE differences

Median population growth rate

1.05

(a)

0.95

0.2 0 –0.2 –0.4 –0.6

0.9

–0.8

(b)

0.85

Seedling

Small

Medium

Large

Extra-large

0.03 Fecundity Survival

0.02

0.75 2003–04

2004–05

2005–06

Time intervals

LTRE contributions

0.01

0.8

0 –0.01 –0.02 –0.03 –0.04 –0.05

Figure 4. Lambda values and 95% confidence intervals (from 2000 bootstrapped transition matrices) for each population and time interval. Dashed line highlights a stable population (l 1).

–0.06

Seedling

Small

Medium

Large

Extra-large

–0.07 Stage classes

Probability of quasi-extinction The probability of reaching a threshold density of 50 individuals was very different between Laguna and Dos Hermanas under both scenarios. Such probability reached 100% after only ca 16 yr at Laguna, irrespective of the scenario considered. However, at Dos Hermanas 100% probability of quasi-extinction was reached in about 240 yr under the ‘‘realistic’’ scenario, and was reached 50 yr earlier under the ‘‘warming’’ scenario (Fig. 6).

Discussion Demographic processes at contrasting altitudes Our study clearly revealed that the S. ciliata population at the lowest altitude presented a pronounced decline and showed a demographic behaviour highly different from those at the central and highest altitudes. These differences were great enough to suggest that the studied peripheral population of S. ciliata at its lower margin is not able to withstand present conditions, which may lead to its local extinction. This scenario was supported by the long-term simulations showing great quasi-extinction hazard at the lowest altitude, irrespective of the climatic scenario projected (Fig. 6). In the central population, finite rate of increase was relatively stable, and time to quasi-extinction was much longer than at the rear edge, indicating that 90

Figure 5. LTRE differences (a) and contributions (b) to differences in l between populations, per each stage class. Contributions were grouped per vital rate (survival and fecundity). Bars represent the mean value9SD for all time intervals. The Dos Hermanas matrix was used as the reference matrix.

population declining is not likely to be occurring at this altitude. However, quasi-extinction probability under a warming scenario (with a higher chance of extreme years) showed that this population is also vulnerable to changing conditions. The larger confidence intervals of l at Laguna 1 Cumulative probability of quasi-extinction

survival of all reproductive stages at Laguna most greatly contributed to differences in l. Fecundity contributions increased with stage class but were always substantially lower than survival contributions (Fig. 5b).

0.9 Laguna (Low) 0.8

Dos Hermanas (Middle)

0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0

100

150

200

250

Years into the future

Figure 6. Average of ten simulated cumulative distribution functions of 5000 simulations, for the time to reach a quasiextinction threshold of 50 individuals for both populations of S. ciliata. The plot only shows the ‘‘warming’’ scenario, in which all time intervals are equally probable.

compared to those at Dos Hermanas reflected higher uncertainty in population growth estimates, as a consequence of higher variability in plant fates. These results were expected because marginal populations are usually near the limit of their physiological tolerance and, consequently, are more vulnerable to environmental stochasticity (Gaston 2003). Although we could only provide one transition, the most favourable trend is expected at the highest population (Penãlara peak), based on the positive finite rate of increase (l 1.059) and the significantly higher values of adult survival (100% for all reproductive stages). Seedling recruitment observed in the field was highly variable and mean values were very low for all the studied years. This is a common feature in long-lived, high mountain species (Forbis 2003, Forbis and Doak 2004). Furthermore, recruitment in Mediterranean mountains is seriously limited by high seedling mortality during summer drought (Castro et al. 2004, 2005, Cavieres et al. 2005). The climatic diagram presented in Fig. 1 precisely showed that mean monthly temperature during the growing season of all the studied years was 1.5 28C over the mean of the last 60 yr. Moreover, mean monthly precipitation was severely reduced just in July, the harsher month of the summer period. The studied period is therefore representative of the warming trend occurred in the last half of the century in this area (Wilson et al. 2005, Gimeńez-Benavides et al. 2007a). Previous results in S. ciliata showed that recruitment was mainly limited by low seed production and low seedling emergence and survival, probably due to environmental harshness in summer, in contrast to biotic factors such as flower and fruit predation, which had a minor effect on the probability of plant recruitment (Gimeńez-Benavides et al. 2008). Although a reciprocal sowing experiment detected some evidence of local adaptation in seedling establishment along this gradient (Gimeńez-Benavides et al. 2007b), the probability of recruitment (estimated as the probability of an ovule becoming a 2-yr-old plant) was 20 to 40-fold higher in medium and higher populations compared to the lower population (Gimeńez-Benavides et al. 2008). Reduced fecundity at species’ distribution limits has been widely observed, and may be the main factor responsible for lower population densities and aged population structures (Garcıá et al. 2000, Dorken and Eckert 2001, Jump and Woodward 2003, Marcora et al. 2008). As predicted by life history theory (Grime 2001), extended longevity is expected to allow long-term persistence of remnant populations in harsh environments with high interannual climate variation, while waiting for eventual recruitment episodes (Morris and Doak 1998, Garcıá and Zamora 2003, Garcıá 2008). In fact, our LTRE analyses suggested that observed differences in l between populations, much more pronounced at the seedling stage, were only slightly explained by the fecundity term (Fig. 5b). Thus, fecundity didn’t seem an important factor explaining differences in population growth rate between altitudes. For this reason, demographic trends of long-lived plants at their rear edge populations cannot simply be inferred from their current recruitment rates, as they are more determined by adult mortalities (Hampe and Petit 2005). Moreover, long-lasting longevity have been proved to buffer the variability in vital rates associated with the variability in climate, hence reducing the vulnerability of long-lived species to climate change (Morris et al. 2008).

Interestingly, the effects of plant size on survival did not follow the same trend at both altitudes. A decrease in plant survival of extra-large plants occurred at the lower population, while this stage reached 100% probability of survival at the central population. Size-dependence of demographic fates defined as transition probabilities has been commonly observed in a great variety of plants and environments (Hortvitz and Schemske 1995). Greater adult survival has also been observed in other alpine cushion-form Caryophyllaceae such as Silene acaulis (Morris and Doak 1998), Minuartia obtusiloba and Paronychia pulvinata (Forbis and Doak 2004). Smaller size stages are expected to be more vulnerable to losses of above-ground tissues during adverse environmental conditions (e.g. seasonal drought), and consequently are expected to show higher rates of mortality. However, our results suggest that environmental conditions at the lower limit may also be affecting survival of large plants by seriously reducing their life span. Indeed, maximum longevity of S. ciliata individuals, inferred from transition matrices, was very low at Laguna population (23 yr) compared to Dos Hermanas (147 yr). In general, maximum longevity of S. ciliata is relatively low when compared to S. acaulis, another cushion plant from arcticalpine habitats (Morris and Doak 1998). The lifespan of S. acaulis may extend over 300 yr and demographic studies carried out in this species did not detect mortality among large plants (Morris and Doak 1998). The results found at Pen˜alara peak the highest population of the Guadarrama mountain range were more in accordance with this pattern (100% adult survival, resulting in a maximum longevity of over 350 yr). Evidence of changes in plant life-history along altitudinal gradients has been reported previously (Ko¨rner 2007). Von Arx et al. (2006) detected significantly older plants and lower growth rates at higher altitudes by means of herb-chronology in three forb species, corresponding to a more conservative life-history. Population dynamics of Silene ciliata: towards an altitudinal range contraction Our results highlight the relevance of survival and longevity for dissecting the processes that may be driving such distinct population dynamics. Together with reproductive limitations (Gime´nez-Benavides et al. 2007a, 2008), rear edge populations of S. ciliata at Sierra de Guadarrama are suffering other demographic processes, resulting in low adult survival, which may force them to an altitudinal range shift. Populations inhabiting the rear edge will become completely extirpated if current demographic processes prevail. As upward shift is unviable, since the species actually colonizes the highest summits of the major mountain ranges in this southern margin, range shifts will therefore be irremediably associated with a decrease in habitat area. Biotic causes, apart from direct climatic effects on survival and reproduction, are undoubtely involved in the expected habitat contraction. The encroachment of the high mountain xerophytic pastures (the main niche of S. ciliata) by montane shrub species already detected in the area (Sanz-Elorza et al. 2003), is probably one of the major sources of risk. Remnant patches of suitable habitat are currently colonized by small-sized populations dominated 91

by adult and senescent plants with extremely low proportions of seedlings and juveniles and a high degree of isolation. Reduced population sizes and isolation are common factors limiting individual reproductive performance (Leimu et al. 2006), leading to a feedback process towards local extinction. Under this scenario, long-term persistence would only be possible by the longevity of established individuals, but our results highlight that longevity is also seriously eroded at this range margin. These findings contrast with the assumption that populations of many alpine species are not likely to be affected substantially by climate warming due to their long lifespan (Steinger et al. 1995, Diemer 2002). Evolutionary implications in response to climate change As noted above, extreme longevity coupled with occasional recruitment episodes may support the demographic stasis and even growth of perennial plant populations. However, a more important consequence of population dynamics governed by extreme longevity arises in the context of global change. The evolutionary adaptation of populations to changes in environmental conditions varies over both space and time as a consequence of natural selection operating on fitness components, and eventually fixed by sexual regeneration. Rapid climate change may act as a potent agent of natural selection within populations and, in this context, the adaptive potential of a given population will be partially ruled by the frequency of sexual regeneration, being annual plants the quickest to adapt because of their short generation time (Jump and Pen˜uelas 2005). By contrast, in long-lived perennials with delayed regeneration time, the lag of adaptation will be significantly longer. Moreover, in S. ciliata an extremely low recruitment rate is coupled to size-dependent reproduction in stressful years, especially in its lowland rearing edge (Gimeńez-Benavides et al. 2007a, 2008). In years of extreme summer drought, small-sized individuals have a much lower flowering probability, seriously limiting opportunities for adaptive selection. In conclusion, although high longevity is the last strategy to assure the long-term persistence of remnant populations of S. ciliata, it is also critically reduced at its lowland range limit. This situation is affecting the population growth rate, eventually forcing the upward shift and the contraction of its regional distributional range. Further demographic studies are required to gather the necessary data to move from simple bioclimatic niche models to process-based models that take into account both climate change and population dynamics. Acknowledgements The authors especially thank Pedro QuintanaAscencio for his help with Matlab programming and the staff of Parque Natural de las Cumbres, Circo y Lagunas de Penãlara who gave them permission to work in the area. They also thank Nuria Ortega, Vera Ortega and Rau´l Garcıá-Camacho who helped with the field work and Lori De Hond for her linguistic assistance. This work was supported by projects ISLAS (CGL2009-13190-C0301), SIL-HAD (CGL2009-08755) and LIMITES (CGL2009-

07229) funded by the Ministerio de Ciencia e Innovacio´n (Spain) and REMEDINAL2 funded by Comunidad de Madrid.

References Angert, A. L. 2006. Demography of central and marginal populations of monkeyflowers (Mimulus cardinalis and M. lewisii). Ecology 87: 2014 2025. Castro, J. et al. 2004. Seedling establishment of a boreal tree species (Pinus sylvestris) at its southernmost distribution limit: consequences of being in a marginal, Mediterranean habitat. J. Ecol. 92: 266 277. Castro, J. et al. 2005. Alleviation of summer drought boots establishment success of Pinus sylvestris in a Mediterranean mountain: an experimental approach. Plant Ecol. 181: 191 202. Caswell, H. 2001. Matrix population models: construction, analysis, and interpretation. Sinauer. Cavieres, L. A. et al. 2005. Nurse effect of the native cushion plant Azorella monantha on the invasive non-native Taraxacum officinale in the high-Andes of central Chile. Perspect. Plant Ecol. Evol. Syst. 7: 217 226. Diemer, M. 2002. Population stasis in a high-elevation herbaceous plant under moderate climate warming. Basic Appl. Ecol. 3: 77 84. Doak, D. and Morris, W. 1999. Detecting population-level consequences of ongoing environmental change without long-term monitoring. Ecology 80: 1537 1551. Dorken, M. E. and Eckert, C. G. 2001. Severely reduced sexual reproduction in northern populations of a clonal plant, Decodon verticillatus (Lythraceae). J. Ecol. 89: 339 350. Eriksson, O. 2000. Functional roles of remnant plant populations in communities and ecosystems. Global Ecol. Biogeogr. 9: 443 450. Erschbamer, B. et al. 2009. Short-term signals of climate change along an altitudinal gradient in the South Alps. Plant Ecol. 202: 79 89. Escudero, A. et al. 2004. Patch dynamics and islands of fertility in a high mountain Mediterranean community. Arct. Antarct. Alp. Res. 36: 518 527. Foden, W. et al. 2007. A changing climate is eroding the geographical range of the Namib Desert tree Aloe through population declines and dispersal lags. Divers. Distrib. 13: 645 653. Forbis, T. A. 2003. Seedling demography in an alpine ecosystem. Am. J. Bot. 90: 1197 1206. Forbis, T. A. and Doak, D. F. 2004. Seedling establishment and life history trade-offs in alpine plants. Am. J. Bot. 91: 1147 1153. Garcıá, D. and Zamora, R. 2003. Persistence and multiple demographic strategies in long-lived mediterranean plants. J. Veg. Sci. 14: 921 926. Garcia, D. et al. 2000. Geographical variation in seed production, predation and abortion in Juniperus comunis throughout its range in Europe. J. Ecol. 88: 436 446. Garcia, M. B. 2008. Life history and population size variability in a relict plant. Different routes towards long-term persistence. Divers. Distrib. 14: 106 113. Garcıá-Camacho, R. 2009. Evaluacioń del e´xito reproductive de Armeria caespitosa en el contexto de cambio clima´tico en la alta montanã mediterrańea. PhD thesis, Univ. Rey Juan Carlos, Madrid, Spain. Gaston, K. J. 2003. The structure and dynamics of geographical ranges. Oxford Univ. Press. Gimeńez-Benavides, L. et al. 2005. Seed germination of high mountain Mediterranean species: altitudinal, interpopulation and interannual variability. Ecol. Res. 20: 433 444.

Gimeńez-Benavides, L. et al. 2007a. Reproductive limits of a lateflowering high-mountain Mediterranean plant along an elevational climate gradient. New Phytol. 173: 367 382. Gimeńez-Benavides, L. et al. 2007b. Local adaptation enhances seedling recruitment along an altitudinal gradient in a high mountain Mediterranean plant. Ann. Bot. 99: 723 734. Gimeńez-Benavides, L. et al. 2008. What shapes the altitudinal range of a high mountain Mediterranean plant? Recruitment probabilities from ovule to seedling stage. Ecography 31: 731 740. Gottfried, M. et al. 1999. A fine-scaled predictive model for changes in species distribution patterns of high mountain plants induced by climate warming. Divers. Distrib. 5: 241 251. Grabherr, G. et al. 1994. Climate effects on mountain plants. Nature 369: 448. Grime, J. P. 2001. Plant strategies, vegetation processes, and ecosystem properties. Wiley. Hampe, A. and Petit, R. J. 2005. Conserving biodiversity under climate change: the rear edge matters. Ecol. Lett. 8: 461 467. Hartigan, J. A. 1975. Clustering algorithms. Wiley. Horvitz, C. C. and Schemske, D. W. 1995. Spatiotemporal variation in demographic transitions for a neotropical understory herb: projection matrix analysis. Ecol. Monogr. 65: 155 192. Iriondo, J. M. et al. 2008. Conservation of plant populations. Myths and paradigms. In: Valladares, F. et al. (eds), Unity in diversity. Reflections on ecology after the legacy of Ramon Margalef. Fundacioń BBVA, pp. 247 267. Jump, A. S. and Woodward, F. I. 2003. Seed production and population density decline approaching the range-edge of Cirsium species. New Phytol. 160: 349 358. Jump, A. S. and Pen˜uelas, J. 2005. Running to stand still: adaptation and the response of plants to rapid climate change. Ecol. Lett. 8: 1010 1020. Klanderud, K. and Birks, H. J. B. 2003. Recent increases in species richness and shifts in altitudinal distributions of Norwegian mountain plants. Holocene 13: 1 6. Ko¨rner, C. 2007. The use of ‘altitude’ in ecological research. Trends Ecol. Evol. 22: 569 574. Leimu, R. et al. 2006. How general are positive relationships between plant population size, fitness and genetic variation? J. Ecol. 94: 942 952. Lesica, P. and Allendorf, F. W. 1995. When are peripheral populations valuable for conservation? Conserv. Biol. 9: 753 760. Lesica, P. and McCune, B. 2004. Decline of arctic alpine plants at the southern margin of their range following a decade of climatic warming. J. Veg. Sci. 15: 679 690. Marcora, P. et al. 2008. The performance of Polylepis australis trees along their entire altitudinal range: implications of climate change for their conservation. Divers. Distrib. 14: 630 636. Menges, E. S. and Dolan, R. W. 1998. Demographic viability of Silene regia in midwestern prairies and relationships with fire management, genetics, geography, population size, and isolation. J. Ecol. 86: 63 78. Morin, X. and Thuiller, W. 2009. Comparing niche- and processbased models to reduce prediction uncertainty in species range shifts under climate change. Ecology 90: 1301 1313. Morris, W. F. and Doak, D. F. 1998. Life history of the long-lived gynodioecious cushion plant Silene acaulis (Caryophyllaceae), inferred from size-based population projection matrices. Am. J. Bot. 85: 784 793.

Morris, W. F. and Doak, D. F. 2002. Quantitative conservation biology. Theory and practice of population viability analysis. Sinauer. Morris, W. F. et al. 2008. Longevity can buffer plant and animal populations against changing climatic variability. Ecology 89: 19 25. Nantel, P. and Gagnon, D. 1999. Variability in the dynamics of northern peripheral versus southern populations of two clonal plant species, Helianthus divaricatus and Rhus aromatica. J. Ecol. 87: 748 760. Palacios, D. et al. 2003. Distribution and effectiveness of nivation in Mediterranean mountain, Pen˜alara (Spain). Geomorphology 54: 157 178. Pauli, H. et al. 2007. Signals of range expansions and contractions of vascular plants in the high Alps: observations (1994 2004) at the GLORIA master site Schrankogel, Tyrol, Austria. Global Change Biol. 13: 147 156. Pen˜uelas, J. and Boada, M. 2003. A global change-induced biome shift in the Montseny mountains (NE Spain). Global Change Biol. 9: 131 140. Petit, R. J. et al. 2004. Ecology and genetics of tree invasions: from recent introductions to Quaternary migrations. For. Ecol. Manage. 197: 117 137. Pico´, X. F. and Riba, M. 2002. Regional-scale demography of Ramonda myconi: remnant population dynamics in a preglacial relict species. Plant Ecol. 161: 1 13. Samis, K. E. and Eckert, C. G. 2007. Testing the abundant center model using range-wide demographic surveys of two coastal dune plants. Ecology 88: 1747 1758. Sanz-Elorza, M. et al. 2003. Changes in the high-mountain vegetation of the central Iberian Peninsula as a probable sign of climate warming. Ann. Bot. 92: 273 280. Scha¨r, C. et al. 2004. The role of increasing temperature variability in European summer heatwaves. Nature 427: 332 336. Steinger, T. et al. 1995. Long-term persistence in a changing climate: DNA analysis suggests very old ages of clones of alpine Carex curvula. Oecologia 105: 94 99. Stokes, K. E. et al. 2004. Population dynamics across a parapatric range boundary: Ulex gallii and Ulex minor. J. Ecol. 92: 142 155. Theurillat, J.-P. and Guisan, A. 2001. Potential impact of climate change on vegetation in the European Alps: a review. Clim. Change 50: 77 109. Thuiller, W. et al. 2008. Predicting global change impacts on plant species’ distributions: future challenges. Perspect. Plant Ecol. Evol. Syst. 9: 137 152. Trivedi, M. R. et al. 2008. Spatial scale affects bioclimate model projections of climate change impacts on mountain plants. Global Change Biol. 14: 1089 1103. Tutin, T. et al. 1995. Flora Europaea Vol. 1 5. Electronic dataset supplied by Pankhurst, R. J. Virtanen, R. et al. 2003. Long-term changes in alpine plant communities in Norway and Finland. In: Nagy, L. et al. (eds), Alpine biodiversity in Europe. Ecological Studies 167. Springer, pp. 411 421. von Arx, G. et al. 2006. Evidence for life history changes in high-altitude populations of three perennial forbs. Ecology 87: 665 674. Walther, G.-R. et al. 2002. Ecological responses to recent climate change. Nature 416: 389 395. Walther, G.-R. et al. 2005. Trends in the upward shift of alpine plants. J. Veg. Sci. 16: 542 548. Wilson, R. J. et al. 2005. Changes to the elevational limits and extent of species ranges associated with climate change. Ecol. Lett. 8: 1138 1146.

Download the Supplementary material as file E6250 from <www.oikos.ekol.lu.se/appendix>. 93

ECOGRAPHY 26: 291–300, 2003

Species-richness patterns of vascular plants along seven altitudinal transects in Norway John Arvid Grytnes

Grytnes, J. A. 2003. Species-richness patterns of vascular plants along seven altitudinal transects in Norway. – Ecography 26: 291– 300. Altitudinal richness patterns were investigated along altitudinal gradients located in northern Norway (two transects) and along a west– east gradient in southern Norway (five transects). The transects were sampled for vascular plant species richness using a uniform sampling method. Each transect consisted of 38 – 48 5 ×5 m sample plots regularly spaced from sea level or valley bottom to a local mountain top. In five transects species richness peaked at mid-altitudes, whereas in the two northern transects species richness decreased with altitude. The observations were qualitatively evaluated in relation to the influence of the area of the species pool, hard boundaries, temperature and precipitation, and mass effect. The observed patterns cannot be fully accounted for by any of these factors. However, the altitude of the peak in species richness was above the forest-limit for all the humped relationships, which may suggest that species richness above the forest-limit might be enhanced by a mass effect from forest taxa. The two monotonic relationships found in the north may be caused by the relatively low number of alpine species at these sites. The monotonic pattern may result from a decrease in ‘‘forest species’’ towards the mountain tops. J. A. Grytnes ( jon.grytnes@bot.uib.no), Dept of Botany, Uni6. of Bergen, Alle´gaten 41, N-5007 Bergen, Norway.

Understanding altitudinal species-richness patterns is important for the management of species diversity in a world that may become warmer due to human impact. Large environmental variation within a small geographical area makes altitudinal gradients ideal also for investigating several ecological and biogeographical hypotheses (Ko¨rner 2000). Altitudinal gradients have therefore become increasingly popular for investigating patterns in species richness in recent years. At present it is not clear if there are any common patterns of species richness and altitude. Rahbek (1995) reviewed studies on altitudinal species-richness patterns and found that approximately half the studies revealed a hump-shaped pattern. Also monotonically decreasing species richness with altitude or no relationship before a decrease in richness with altitude are common altitudinal richness patterns. Studies since this review confirm that both unimodal (e.g. Rahbek 1997, Fleishman et al. 1998, Heaney 2001, Md. Noor 2001) and monotonically

decreasing (e.g. Patterson et al. 1998, Odland and Birks 1999, Ohlemu¨ller and Wilson 2000, Austrheim 2002) patterns of richness occur with altitude. Rahbek (1995) suggests that differences between studies may at least partially result from failure to correct for area and sampling effort (see also Terborgh 1977, McCoy 1990, Whittaker et al. 2001). There is a major lack of studies on altitudinal richness that are directly comparable and this makes it difficult to assess whether the variation in patterns are real or due to sampling artefacts. Lomolino (2001a) calls for studies comparing altitudinal trends between mountain ranges using the same sampling regimes for all transects and standardised plot sizes within transects. In this study I investigate the pattern of vascular plant species richness along seven altitudinal gradients. All are sampled with the same method and identical plot size, ensuring that the patterns both within and between transects are not a result of different sampling

291

strategies. The main aim of the study is to describe statistically the altitudinal pattern in species richness in different areas in Norway. In addition a discussion of some selected potential causes for the observed pattern is presented by setting up tentative predictions for what is expected if each of these factors is important in determining the richness pattern.

Methods and study area Transects were sampled along seven hillsides at different geographical locations in Norway. Two transects were in northern Norway, and the remaining five make a gradient from west to east in southern Norway. The transects were chosen after looking at detailed maps (scale 1:50 000) to find transects with a relatively long continuous hillside with little variation in aspect. Transects were selected to be in areas of homogeneous bedrock type (Holtedahl and Dons 1960). Some important characteristics of the selected transects are summarised in Table 1. Sampling was done during the summers of 1999, 2000, and 2001. Plots were placed along the transects starting as far down as possible with the criterion that plots should have a similar aspect throughout the transect, and not be placed in clear-cuts or other areas clearly influenced by humans. All transects are, however, influenced by grazing by domestic animals, mainly sheep. The highest altitude sampled is either a local top, or where vegetation cover became discontinuous. In one case weather conditions (30 cm of fresh snow) and time limitations prevented sampling at higher altitudes even though the vegetation was continuous for at least another 100 altitude m (Kvitingskjølen).

Plot size (or grain size) is 5 × 5 m. All vascular plants were recorded in each plot. Nomenclature and taxonomy follows Lid and Lid (1994) with the exception that Taraxacum and Hieracium species were only identified to genus level. Depressions and ridges were avoided when the plots were placed, as were mires. This was done to avoid sampling extreme wet and extreme dry places. Generally, distance between plots is 20 vertical metres with some deviations depending on the length of the gradient, steepness of hillside, etc. The number of plots in a transect varies from 38 (Lynghaugtinden) to 48 (Gra˚ heivarden and Tronfjellet), which is considered sufficient for a statistical evaluation of the altitudinal species richness pattern for each transect. The relatively small grain size compared to other studies on altitudinal variation in species richness was used because it was possible to have a large number of samples within each transect and to sample several transects. The small plot size also ensures a fairly complete list of species from each plot. Sampling at scales not usually sampled may also give extra information and suggest how processes may influence the patterns. However, there may be other processes explaining species richness patterns if larger areas are sampled than if smaller grain sizes are used (Rahbek and Graves 2001, Whittaker et al. 2001). In addition, the fine scale of the sampling area makes the estimate of species richness sensitive to the number of individuals (Gotelli and Colwell 2001). To evaluate if the small grain size created a different pattern than would have been detected with larger grain sizes, two approaches were used. First, testing the altitudinal pattern using larger sample sizes in four of transects, and second, using a nested sampling to evaluate if the altitudinal richness pattern is dependent on sampling area.

Table 1. A summary of some important characteristics for each transect. Location of each transect is indicated by county, latitude, and longitude (see also Fig. 1). The lowest and highest points locally corresponding to the hard boundaries along the hillside where the transect is sampled are given. Note that this does not always correspond to the highest and lowest points sampled (cf. Fig. 2). A crude description of the bedrock geology is taken from Holtedahl and Dons (1960). Approximate forest limit as observed in the field is indicated. Aspect of the sampled transect is given. Annual precipitation and mean January and July temperature from the nearest station are from Førland (1993) and Aune (1993), respectively (altitude of the climatic station is given in brackets). Mountain

Trollan

County

Latitude, Lowest– Bedrock Longitude highest point (m)

Bø in Vestera˚ len 68°46%N, 14°33%E Lynghaugtinden Bø in Vestera˚ len 68°40%N, 14°32%E Horndalsnuten Voss 60°39%N, 06°38%E Gra˚ heivarden Jondalen 60°13%N, 06°15%E Grjothøi Skja˚ k 61°52%N, 08°08%E Kvitingskjølen Lom 61°45%N, 08°43%E Tronfjellet Tynset/Alvdal 62°11%N, 10°41%E

292

0–543 0–504 300–1461 0–1271 450–1952 360–2060 600–1666

Forest Aspect Annual Jan°C limit precipitation (mm (m))

Acidic 20 granite Acidic 200 granite Acidic 700 supracrustal Acidic 600 granite Acidic 950 gneiss Basic plu1000 tonic Basic Cam- 900 bro-Silurian

1505 (3)

Jul°C

−1.4 (12)

12.1

WNW 1017 (12)

−1.4 (12)

12.1

1555 (590)

−4.6 (590) 11.8

2200 (342)

278 (372)

−9.4 (378) 13.9

NNW

321 (382)

−9.4 (378) 13.9

500 (485)

−11.4 (485) 12.5

0.8 (1)

14.3

ECOGRAPHY 26:3 (2003)

For the first approach larger areas were sampled in addition to the ordinary 5 ×5 m plots in four transects. In one transect (Grjothøi) a plot of 10 × 10 m surrounding each of the ordinary plots was used. In three other transects (Horndalsnuten, Kvitingskjølen, and Tronfjellet) a 25 m walk to each side of the main plot perpendicular to the slope of the hillside was made for approximately half the ordinary plots. All species within one metre were noted and added to the 5 × 5 m plot, i.e. the richness of the large-scale plots are the sum of the 5×5 m plots plus the extra species in the 50 × 2 m walk. The second approach is evaluating if the altitudinal richness pattern is plot-size sensitive or not along one of the seven transects (Grjothøi). If the emerging richness pattern is a result of different number of individuals in the plots along the altitudinal gradient (Gotelli and Colwell 2001), this will be clearest at small grain sizes (fewest number of species). At larger grain sizes the number of individuals will increase in all plots and due to the relationship between the number of individuals and the number of species, the effect of the number of individuals on the altitudinal richness pattern will be dampened. In one transect (Grjothøi), a nested design of samples were used to allow a test of the sensitivity of the altitudinal richness pattern on sampling area. Quadrat samples of 0.25 m2 (four contiguous samples), 1 m2, 25 m2 (four contiguous samples), and 100 m2 were sampled at each altitudinal level (a total of ten samples at each altitude level) and multiple regression was used to detect if any significant change in the altitude pattern could be found with sampled area. This was statistically evaluated by testing if the interaction between sampled area and altitude was significant after including both variables in multiple regression. Area was log-transformed prior to analyses.

Statistical methods A Generalised Linear Model (GLM; McCullagh and Nelder 1989) was made for each individual transect. The most common patterns described for altitudinal richness, and the main interest in this study, are whether richness is monotonically or unimodally related to altitude. GLM using first- or second-order polynomials was therefore tested against no relationship and against each other. However, a GLM with a first- or second-order polynomial is restricted to be linear or symmetrically unimodal, whereas several authors describe a plateau before a decrease or an asymmetric hump. Therefore a more complex pattern described by non-parametric regression may give valuable additional information when inspecting the graphical output. A Generalised Additive Model (GAM; ECOGRAPHY 26:3 (2003)

Fig. 1. Map showing the geographical location of the seven transects.

Hastie and Tibshirani 1990) with a cubic smoother spline using 3, 4, or 5 degrees of freedom was tested against the best GLM model. As the response variable is counts (number of species) a Poisson distribution is assumed and a log link is used for all regressions. This gave satisfactory results when inspecting diagnostic plots. F-tests were used to evaluate statistical significance.

Predictions Numerous hypotheses have been proposed to explain both a linear and humped relationship between richness and altitude (recently reviewed by Brown and Lomolino 1998, Brown 2001, Lomolino 2001a). Hard boundaries (Colwell and Hurtt 1994, Colwell and Lees 2000, Grytnes and Vetaas 2002), area (MacArthur 1972, Rahbek 1995, 1997, Odland and Birks 1999), climate (Odland and Birks 1999), and mass effect (Shmida and Wilson 1985, Kessler 2000) are commonly discussed when altitudinal species richness patterns are considered. Although no quantitative measurements are available to test the different hypotheses directly, a qualitative evaluation of the predictions made for each of these is possible. This section therefore makes tentative predictions for the different transects based on these four factors. Note that even if this section sets up predictions and these predictions are evaluated in the Discussion section, the transects are far too few to consider this as a rigorous statistical evaluation of the different factors included here. 293

Hard boundaries Random placement of species optima and species ranges between an upper hard boundary (mountain top) and a lower hard boundary (sea level or valley bottom) will give a humped richness pattern along altitude. This effect has been demonstrated by simulations and analytical modelling in several studies (Colwell and Hurtt 1994, Willig and Lyons 1998, Koleff and Gaston 2001, Grytnes and Vetaas 2002). The simulations show that hard boundaries result in a symmetrical humped relationship in the middle of the gradient. It is also evident from these simulations that species richness decreases most steeply as the boundaries are approached. In this study the valley bottom or sea level and the mountain top define the hard boundaries (cf. Table 1). Following the results of simulations, humped relationships are predicted for all transects and maximum species richness is predicted to occur at or near the middle of the gradient (Table 1). Richness should also decrease most strongly as the hard boundaries are approached.

Area of species pool The relationship between species richness and area is well known (Rosenzweig 1995, Lomolino 2001b). In this study the sampled area is the same along the transect, hence the direct effect of plot area is cancelled out. However, as several studies have demonstrated, the number of species in a sample is dependent on the number of species in the species pool (Ricklefs 1987, Cornell 1999). The concept of species pool has been used with various meanings (Grace 2001). Some take the species present in a larger area around the focus area to be the species pool for the focus area (Caley and Schluter 1997), whereas other incorporates the constraints set by the environment (Gough et al. 1994, Dupre´ 2000, Grytnes and Birks 2003). Taking the latter approach here gives a potential role for the area through the effect on the number of species in the species pool. This interpretation is in accordance with Terborgh (1973) and Taylor et al. (1990), who were among the first to discuss the importance of species pools (although Terborgh (1973) did not use this term). They argue that higher species richness will be found in plots that have the most common habitat. The main environmental gradient in this study coincides with the altitudinal gradient. The areas covered by an altitudinal interval vary along the gradient. This area may then influence plot species richness through two steps. First, the area of an altitude interval influences the number of species at large scale (the species pool) and, second, the plot species richness depends on the size of the species pool (see also Lomolino 2001a). 294

The transects differ in the distribution of area for the altitude intervals as the mountains are not equally shaped. An accurate estimate of area per altitudinal interval is not available. Therefore only a qualitative inspection of maps at a scale of 1:50 000 is used here. Some of the mountains are more or less dome-shaped in a matrix of large area lowlands, which gives a monotonically decreasing area with altitude at both the local and the regional scale (Ko¨ rner 2000). These mountains are Lynghaugtinden, Trollan, and Tronfjellet. The other mountains are hillsides going down to a narrow valley or to the sea in an area dominated by high mountain plateaux. The area of the altitudinal bands in these mountains therefore depends on the slope of the hillside. Gentle-sloped hillsides result in large areas per altitude metre, and steep hillsides result in small areas per altitude metre. Horndalsnuten and Kvitingskjølen have approximately the same slope all along the transect and go down to a rather narrow valley. The area is therefore approximately equal for each interval along these two transects. At Gra˚ heivarden the hillside is steep and remains steep until it goes into the fjord. The slope becomes gentler, and hence the areas with similar climatic environments become larger above 1000 m. At Grjothøi the steepest slopes are towards the lowest altitudes while the slopes are gentler between 1200 and 1500 m. The predictions from this would therefore be that Lynghaugtinden, Trollan, and Tronfjellet should have a monotonically decreasing richness with altitude, Kvitingskjølen and Horndalsnuten should have no trend in richness, and Grjothøi and Gra˚ heivarden should have no trend until 1200 and 1000 m, respectively, and above these altitudes richness should increase.

Temperature and rainfall Temperature linearly decreases with altitude whereas precipitation has a more erratic pattern with altitude, but generally increases towards higher altitudes. Several studies have demonstrated that temperature or precipitation may influence species richness at broader scales (Currie and Paquin 1987, Leathwick et al. 1998, Odland and Birks 1999, Grytnes et al. 1999, Ohlemu¨ ller and Wilson 2000). Assuming that these variables have a positive effect on species richness along altitude in the transects in this study a monotonically decreasing trend for richness with altitude is expected (not necessarily linear if precipitation is important). The pattern should be similar for all transects. The two climatic variables in combination have often been seen as an indirect estimate of productivity (Currie 1991, O’Brien 1993, Odland and Birks 1999). Focussing on the effect that temperature and precipitation have on productivity, and assuming a positive relationship between productivity and species richness (for a discusECOGRAPHY 26:3 (2003)

sion of this see Rahbek (1997) and Waide et al. (1999)), will give different predictions for the continental versus the oceanic transects. In the oceanic transects (Lynghaugtinden, Trollan, Horndalsnuten, and Gra˚ heivarden), precipitation is probably not limiting in any part of the gradient (annual precipitation between 1017 and 2200 mm; Table 1). This means that temperature probably controls the richness in these areas more than precipitation does and a linearly decreasing richness pattern with altitude can be expected. In the continental transects (Grjothøi, Kvitingskjølen, and Tronsfjellet), however, the precipitation may be limiting in the lower part of the transect (annual precipitation between 278 and 500 mm; Table 1) and temperature in the upper part as precipitation increases and temperature decreases. In these cases a unimodal pattern of species richness can be expected. With the present data it is impossible to say where the optimum richness should be expected along the gradient if a combination of temperature and precipitation is important for the continental transects.

Mass effect Mass effect is the establishment of species in sites where a self-maintaining population cannot exist (Shmida and Wilson 1985). Other closely related terms that have been used to explain the same effect include source – sink effect (Pulliam 1988, 2000) and rescue effect (Brown and Kodric-Brown 1977, Stevens 1989). The most obvious way the mass effect can influence the altitudinal richness pattern is through a feedback among zonal communities. This will increase species richness in transition zones between two bordering communities (Lomolino 2001a). The most noticeable transition along altitudinal gradients is the transition from forest to alpine communities. Many species have their distribution more or less restricted to above or below the forest-limit (Hofgaard and Wilmann 2002). The interchange of species between forest and alpine

communities at the forest limit may cause richness to increase at altitudes corresponding to the forest-limit. The prediction is therefore that species richness will peak at or near the forest-limit, which differs between transects (Table 1). Only the Trollan transect has no forest along the whole transect and here no hump is predicted from the transition-zone or mass effect hypotheses.

Results The results of the regressions are summarised in Table 2 and show that when relating species richness to altitude using the 25 m2 plots a statistically significant second-order polynomial model was found in five of the seven transects (Horndalsnuten, Gra˚ heivarden, Grjothøi, Kvitingskjølen, Tronfjellet, Fig. 2). In none of these five transects was the first-order polynomial model statistically significant when tested against the null model of no relationship (lowest p-value for the first-order polynomial is 0.064 at Horndalsnuten). The two transects in northern Norway (Lynghaugtinden and Trollan) showed a linear relationship with altitude and were the only transects where a second-order polynomial model did not statistically significantly improve the fit. Generally the models explain a large part of the deviance (Table 2). For the four transects where larger grain sizes were tried in addition to the 25 m2 plots, the pattern for the larger grain sizes confirms the patterns from the smaller grain size. In this study both the general pattern and the placement along altitude where maximum species richness is estimated are independent of the grain sizes used (cf. Fig. 2 with Fig. 3). The GAM models significantly improved the fit over the GLM models in four of the transects (Table 2). The Lynghaugtind transect clearly has a break point at 250 m where species richness suddenly drops from ca 22 species to ca 12 species. This is captured by the GAM and not by the GLM model. The reason for a signifi-

Table 2. Summary of the regression models between species richness and altitude for each transect (LS indicates large-scale plots). GLM and GAM models with the respective degrees of freedom for each model given and the p-value of each model (given in brackets) refer to a test against no relationship for the GLM models and for the GAM models the p-value refers to a test against the given GLM model. NS = not significant (p \ 0.05). Transect

Null deviance

GLM model

Lynghaugtinden Trollan Horndalsnuten Horndalsnuten (LS) Gra˚ heivarden Grjothøi Grjothøi (LS) Kvitingskjølen Kvitingskjølen (LS) Tronfjellet Tronfjellet (LS)

80.17 74.93 32.52 35.27 67.50 104.26 140.41 65.11 50.47 311.20 212.23

1 1 2 2 2 2 2 2 2 2 2

ECOGRAPHY 26:3 (2003)

( \ 0.001) (B0.001) (0.02) (0.027) ( \ 0.001) (0.0018) ( \ 0.001) (0.039) (\ 0.001) ( \ 0.001) (\ 0.001)

GLM residual deviance

GAM model

GAM residual deviance

34.80 30.74 26.77 19.29 33.75 78.50 95.84 56.01 20.50 119.30 42.37

5 ( \ 0.001) 4 (0.0045) NS Not applied 5 ( \ 0.001) 5 (0.048) Not applied NS Not applied NS Not applied

15.79 21.80 – – 18.58 65.27 – – – – –

295

Fig. 2. Scatter plots of species richness in relation to altitude in the ordinary 5 Ă&#x2014;5 m plots for the seven transects. The unbroken line is the statistically best GLM model of a monotonic (two transects) or unimodal (five transects) relationship. Where a GAM model statistically improved the GLM model a broken line describing this model is added (see Table 2). The vertical broken line indicates the altitude of the forest-limit for each transect, and the midpoint between valley bottom (or sea level) and mountaintop is indicated by * on the altitude axis.

296

ECOGRAPHY 26:3 (2003)

Fig. 3. Scatter plots of species richness along altitude for the large-scale plots. In Grjothøi a plot of 10 ×10 m is used whereas in the three remaining transects the species in an area of 50 × 2 m is added to the ordinary 5 ×5 m plots. The line refers to the best fitted GLM model (see Table 2).

cant improvement of the GAM models over a linear GLM model in the Trollan transect is the plateau in richness from 300 –400 m (probably due to some species-poor plots around 300 m). The GAM model for the Gra˚ heivarden transect indicates a sharp peak in richness rather than a smooth unimodal curve indicated by the GLM model. In the Grjothøi transect the GAM model indicates a bimodal rather than the unimodal altitudinal pattern. This is due to some species-poor plots in the dense belt of Betula nana between 1100 and 1200 m. The GAM models were not applied for the large-scale plots because there are few samples at this scale. Testing the effect of the interaction term between sampling area and altitude on species richness in the Grjothøi transect after accounting for altitude and area indicates that the richness pattern is independent of sampling area within the scale sampled here (F = 0.45, p=0.64, n =480). Even though the samples at each altitude level are not strictly independent the result is so clear that there is no question about whether the pattern is independent of scale within the range of scales used. ECOGRAPHY 26:3 (2003)

Discussion Even if a standard sampling regime with a standard plot size is used, both humped and monotonically decreasing trends of richness appear. Five of the seven transects have a humped altitudinal richness pattern, whereas the two transects in northern Norway show a monotonically decreasing richness with altitude. This is in accordance with the studies reviewed by Rahbek (1995) who finds that a unimodal pattern is the most common and that a linear decreasing trend with altitude is also frequently observed. Previously, differences in altitudinal richness patterns have been attributed to differences in sampling methods (McCoy 1990, Rahbek 1995). In the present study the difference in patterns cannot be due to differences in sampling or plot sizes. The relatively small grain size used in this study may result in the patterns observed here having different causes than in other studies where larger grain sizes were used. For example, the importance of local biological interactions may be more important at smaller grain sizes, or the pattern may be a result of different numbers of individuals at different altitudes (Gotelli and Colwell 2001). While this may be true, the fact that 297

the pattern is the same when sampling larger areas suggests that the pattern is not due to small grain sizes and is therefore comparable to other studies using larger grain sizes. It is probable that small grain size may only obscure the relationship between altitude and richness. This is indicated by the larger explanatory power of the regressions at the large scale compared to the small scale (seen by comparing the fraction of explained deviance of the two scales derived from Table 2). In other words, a humped relationship is evident in spite of the small grain size rather than because of the small grain size. The independence of grain size is also demonstrated for the Grjothøi transect when testing if the pattern changes significantly when different grain sizes are used. The monotonic altitudinal richness patterns are found in two altitudinal transects with many similar features. The geographical distance between the two transects is short (Fig. 1). This implies similar climatic conditions, similar geology, and similar species pool. In addition, both mountains have short altitudinal gradients (Table 1). The finding of monotonic patterns in both transects strengthens confidence that the observed pattern is real. On the other hand, when comparing these two transects with the others and discussing the predictions, the similarities between the transects imply that the two transects might be treated as pseudoreplicates (Hurlbert 1984), hence we should be careful not to treat them as two independent pieces of ‘‘evidence’’ for or against a specific hypothesis. The predictions made from assuming hard boundaries as the only important factor are that species richness peaks at intermediate altitudes, i.e. intermediate between valley bottom (or sea level) and mountain top. This intermediate peak is marked by an asterisk on the graphs (Fig. 2). Inspection of the figures shows that the transects with unimodal relationships usually have the maximum estimated species richness close to the intermediate altitudinal point. Moreover, species richness at Tronfjellet (where richness is sampled from the valley bottom to the mountain top) seems to decrease more steeply as the hard boundaries are approached. This supports the hypothesis of hard boundaries as an important factor. However, at the other two transects where richness is sampled close to both hard boundaries the pattern is linear and does not correspond to the predictions from hard boundaries. At the remaining four transects with a humped pattern, richness is not sampled close enough to the hard boundaries to evaluate if richness decreases more steeply towards these boundaries. However, the hump at Gra˚ heivarden is clearly due to a peak between 650 and 850 m, and does not behave as the predictions from simulations that assume hard boundaries (e.g. Grytnes and Vetaas 2002). To summarise, indications in support of hard boundaries are found in four transects, but indications contradicting the importance of hard 298

boundaries are found in the remaining three transects (Fig. 3). The only transects where species richness behaves as predicted if area was the only important factor are the two northern transects, namely Lynghaugtinden and Trollan. Here a monotonically decreasing trend is expected and is observed. For the other transects, the predicted pattern and observed pattern do not coincide. The importance of species-pool area (gamma richness sensu Whittaker 1960) for plot species richness is therefore probably relatively minor. Lomolino (2001a) reaches a similar conclusion when discussing alternative explanations for altitudinal richness patterns. The predictions made assuming that temperature or rainfall is important assume that all richness should have a monotonic trend with altitude for all transects. This is not found. Assuming that a combination of the two variables are important and that precipitation is limiting for richness at the lower part of the transects in the continental areas and not in the oceanic is partly supported as the two linear relationships observed are in oceanic areas and all the continental transects have a unimodal relationship. However, the two oceanic transects in southern Norway (Horndalsnuten and Gra˚ heivarden) contradict the predictions as they show a unimodal relationship between richness and altitude whereas a linear relationship is expected. These transects are also the transects with the highest annual precipitation (Table 1). Hence the predictions are only partly supported. Mass effect may also enhance species richness in transition zones, especially in the transition zone defined by the forest-limit. The forest-limit is indicated on the figures and, except for the two monotonic patterns, the estimated maximum species richness is generally well above the forest-limit. This contradicts the interpretation of feedback effects among communities as outlined in the Predictions section. However, only a few alpine species are able to survive in shaded conditions, whereas forest species may be able to survive in the more open areas above the forest-limit. Field observations support this as several species with their main distribution below the forest-limit were found in the plots above the forest-limit, but typically not growing as vigorously as lower down. Examples of such species include Geranium syl6aticum, Geum urbanum, Trientalis europaea, and also small specimens of Betula pubescens and Pinus syl6estris. One of the two transects where a monotonic pattern is found is predicted to show a linear pattern as the forest-limit is almost at sea level (Trollan), whereas for the other transect (Lynghaugtinden) the expected humped pattern does not correspond to the observed linear pattern. However, for both these transects, an alternative hypothesis can be proposed for the lack of a humped pattern. In these northern oceanic parts the ECOGRAPHY 26:3 (2003)

species pool for the highest altitudes may be small, partly due to the effect of the oceanic climate (Moe 1995), but also due to the relative low maximum altitudes of the mountain tops in the area. The plant species on the top of these mountains may therefore be mainly species with their main distribution in the lowlands. To evaluate this possibility I counted the number of presences of each species at the highest 200 m, and compared with the lowest 300 m (340 m for Trollan). Only 9 species clearly have more presences in the upper part compared to the lower part for Trollan (Arctostaphylos alpinus has 13 occurrences in the 19 plots at the highest part and only 1 at the lower part, Juncus trifidus also has 13 above and 1 below, Salix herbacea 18:1, Silene acaulis 5:0, Cerastium alpinum 5:1, and Carex bigelowii 6:0, and three species have one or two occurrences in the upper part and none in the lower part). For the other 57 species there is no difference or more individuals of each species are found in the lower part. In Lynghaugtinden 10 species have more presences in the upper 200 m than in the lower 300 m (Silene acaulis has 13 occurrences in 16 plots above and 0 below, Loiseleuria procumbens 3:0, Antennaria dioica, 4:0, Rubus chamaemorus 4:1, Juncus trifidus 12:6, Arctostaphylos alpinus 6:2, Carex brunnescens 6:0, and three species have one or two occurrences above and none below 300 m), whereas 60 species have no difference or have their main distribution in the lower part. This means that, on average, ‘‘alpine species’’ (main distribution above 300/340 m) makes up a small part of the richness in any plot and hence the pattern of ‘‘forest species’’ (main distribution below 300/340 m) completely masks any trend in ‘‘alpine species’’. The total pattern of richness therefore corresponds to the pattern of ‘‘forest species’’ richness. Another possibly important factor, especially when plots are as small as used in this study, is environmental heterogeneity, especially heterogeneity of moisture within the plot. At low altitudes above-ground water is mostly concentrated in large rivers. At high altitudes small streams run almost everywhere, especially early in the growing season when snow is melting. This may enhance heterogeneity in two ways at high altitudes. First, very different moisture regimes may occur within small distances, and second, small streams create local disturbances and different soil conditions (e.g. different soil depth and grain size differences in soil). As species at high altitudes are generally of a relatively small size, fine-scaled heterogeneity created in this way may be an important determinant for differences in richness. Contrary to the importance of heterogeneity on species richness Austrheim (2001) found an increase in soil nutrient heterogeneity at higher altitudes in semi-natural grasslands in Norway, but this was not reflected in species richness. The moderate disturbance made by small streams may itself also enhance richness in plots at higher altitudes. The importance of moderate disturECOGRAPHY 26:3 (2003)

bance in enhancing species richness has been discussed by Huston (1979, 1994). The main conclusion from this study is that altitudinal richness patterns vary between transects even if the same sampling regime is used between different transects. Five of the seven altitudinal transects studied have a unimodal relationship between species richness and altitude, whereas the two remaining transects have a negative linear relationship. The qualitative evaluation of four causal hypotheses considering all seven transects together gave strongest support for the importance of mass effect, whereas there are only weak support for the area of species pool hypothesis. Further studies increasing the number of transects with similar methods should be done to give clearer support or rejection of the possible causal hypotheses. Acknowledgements – I am grateful to Gunnar Austrheim, H. J. B. Birks, Einar Heegaard, Carsten Rahbek, Ole R. Vetaas, and Vigdis Vandvik for reading the manuscript and giving valuable comments. Thanks are also due to Hans H. Blom, Louise Lindblom, Einar Heegaard, Per G. Ihlen, Anne E. Bjune, and Halvard H. Eggen for assistance in the field. Financial support was given by Grolle Olsen’s legat and NFR-grant no. 127 594/720.

References Aune, B. 1993. Temperaturnormaler normalperiode 1961 – 1990. – Det Norske Meteorologiske Institutt, Universitestforlaget, Oslo. Austrheim, G. 2001. Heterogeneity in semi-natural grasslands: the importance of the elevational gradient. – Nord. J. Bot. 21: 291 – 303. Austrheim, G. 2002. Plant diversity patterns in semi-natural grasslands along an elevational gradient in southern Norway. – Plant Ecol. 161: 193 –205. Brown, J. H. 2001. Mammals on mountainsides: elevational patterns of diversity. – Global Ecol. Biogeogr. 10: 101 – 109. Brown, J. H. and Kodric-Brown, A. 1977. Turnover rates in insular biogeography: effect of immigration and extinction. – Ecology 58: 445 – 449. Brown, J. H. and Lomolino, M. V. 1998. Biogeography, 2nd ed. – Sinauer. Caley, M. J. and Schluter, D. 1997. The relationship between local and regional diversity. – Ecology 78: 70 – 80. Colwell, R. K. and Hurtt, G. C. 1994. Nonbiological gradients in species richness and a spurious Rapoport effect. – Am. Nat. 144: 570 – 595. Colwell, R. K. and Lees, D. C. 2000. The mid-domain effect: geometric constraints on the geography of species richness. – Trends Ecol. Evol. 15: 70 – 76. Cornell, H. W. 1999. Unsaturation and regional influences on species richness in ecological communities: a review of the evidence. – Ecoscience 6: 303 – 315. Currie, D. J. 1991. Energy and large scale patterns of animaland plant-species richness. – Am. Nat. 137: 27 – 49. Currie, D. J. and Paquin, V. 1987. Large-scale biogeographical patterns of species richness of trees. – Nature 329: 326 – 327. Dupre´ , C. 2000. How to determine a regional species pool: a study in two Swedish regions. – Oikos 89: 128 – 136. Fleishman, E., Austin, G. T. and Weiss, A. D. 1998. An empirical test of Rapoport’s rule: elevational gradients in montane butterfly communities. – Ecology 79: 2482 – 2493.

299

Førland, E. J. 1993. Nedbørnormaler normalperiode 1961 – 1990. – Det Norske Meteorologiske Institutt, Universitestforlaget, Oslo. Gotelli, N. J. and Colwell, R. K. 2001. Quantifying biodiversity: procedures and pitfalls in the measurement and comparison of species richness. – Ecol. Lett. 4: 379 – 391. Gough, L., Grace, J. B. and Taylor, K. L. 1994. The relationship between species richness and community biomass: the importance of environmental variables. – Oikos 70: 271 – 279. Grace, J. B. 2001. Difficulties with estimating and interpreting species pools and the implications for understanding patterns of diversity. – Folia Geobot. 36: 71 –83. Grytnes, J. A. and Vetaas, O. R. 2002. Species richness and altitude: a comparison between simulation models and interpolated plant species richness along the Himalayan altitude gradient, Nepal. – Am. Nat. 159: 294 – 304. Grytnes, J. A. and Birks, H. J. B. 2003. The influence of scale and species pool on the relationship between vascular plant species richness and cover in an alpine area in Norway. – Plant Ecol., in press. Grytnes, J. A., Birks, H. J. B. and Peglar, S. M. 1999. Plant species richness in Fennoscandia: evaluating the relative importance of climate and history. – Nord. J. Bot. 19: 489 – 503. Hastie, T. J. and Tibshirani, R. J. 1990. Generalized additive models. – Chapman and Hall. Heaney, R. H. 2001. Small mammal diversity along elevational gradients in the Philippines: an assessment of patterns and hypotheses. – Global Ecol. Biogeogr. 10: 15 –39. Hofgaard, A. and Wilmann, B. 2002. Plant distribution pattern across the forest-tundra ecotone: the importance of treeline position. – Ecoscience 9: 375 –385. Holtedahl, O. and Dons, J. A. 1960. Geological map of Norway (bedrock). – Norges Geologiske Undersøkelse, No. 208. Hurlbert, S. H. 1984. Pseudoreplication and the design of ecological field experiments. – Ecol. Monogr. 54: 187 – 211. Huston, M. A. 1979. A general hypothesis of species diversity. – Am. Nat. 113: 81 –101. Huston, M. A. 1994. Biological diversity. The coexistence of species on changing landscapes. – Cambridge Univ. Press. Kessler, M. 2000. Upslope-directed mass effect in palms along an Andean elevational gradient: a cause for high diversity at mid-elevations? – Biotropica 32: 756 – 759. Koleff, P. and Gaston, K. J. 2001. Latitudinal gradients in diversity: real patterns and random models. – Ecography 24: 341 – 351. Ko¨ rner, C. 2000. Why are there global gradients in species richness? Mountains might hold the answer. – Trends Ecol. Evol. 15: 513 –514. Leathwick, J. R., Burns, B. R. and Clarkson, B. D. 1998. Environmental correlates of tree alpha-diversity in New Zealand primary forests. – Ecography 21: 235 – 246. Lid, J. and Lid, D. T. 1994. Norsk flora, 6th ed. – Det Norske Samlaget, Oslo. Lomolino, M. V. 2001a. Elevational gradients of species-density: historical and prospective views. – Global Ecol. Biogeogr. 10: 3 –13. Lomolino, M. V. 2001b. The species –area relationship: new challenges for an old pattern. – Prog. Phys. Geogr. 25: 1 – 21. MacArthur, R. H. 1972. Geographical ecology, patterns in the distribution of species. – Harper and Row.

300

McCoy, E. D. 1990. The distribution of insects along elevational gradients. – Oikos 58: 313 – 322. McCullagh, P. and Nelder, J. A. 1989. Generalized Linear Models, 2nd ed. – Chapman and Hall. Md. Nor, S. 2001. Elevational diversity patterns of small mammals on Mount Kinabalu, Sabah, Malaysia. – Global Ecol. Biogeogr. 10: 41 – 62. Moe, B. 1995. Studies of the alpine flora along an east-west gradient central-western Norway. – Nord. J. Bot. 77 – 89. O’Brien, E. M. 1993. Climatic gradients in woody plant species richness: towards an explanation based on an analysis of southern Africa’s woody flora. – J. Biogeogr. 20: 181 – 198. Odland, A. and Birks, H. J. B. 1999. The altitudinal gradient of vascular plant species richness in Aurland, western Norway. – Ecography 22: 548 – 566. Ohlemu¨ ller, R. and Wilson, J. B. 2000. Vascular plant species richness along latitudinal and altitudinal gradients: a contribution from New Zealand temperate rainforests. – Ecol. Lett. 3: 262 – 266. Patterson, B. D. et al. 1998. Contrasting patterns of elevational zonation for birds and mammals in the Andes of southeastern Peru. – J. Biogeogr. 25: 593 – 607. Pulliam, H. R. 1988. Sources, sinks, and population regulation. – Am. Nat. 132: 652 – 661. Pulliam, H. R. 2000. On the relationship between niche and distribution. – Ecol. Lett. 3: 349 – 361. Rahbek, C. 1995. The elevational gradient of species richness: a uniform pattern? – Ecography 18: 200 – 205. Rahbek, C. 1997. The relationship among area, elevation, and regional species richness in neotropical birds. – Am. Nat. 149: 875 – 902. Rahbek, C. and Graves, G. R. 2001. Multiscale assessment of patterns of avian species richness. – Proc. Natl. Acad. Sci. USA 98: 4534 – 4539. Ricklefs, R. E. 1987. Community diversity: relative roles of local and regional processes. – Science 235: 167 – 171. Rosenzweig, M. L. 1995. Species diversity in space and time. – Cambridge Univ. Press. Shmida, A. and Wilson, M. W. 1985. Biological determinants of species diversity. – J. Biogeogr. 12: 1 – 20. Stevens, G. C. 1989. The latitudinal gradient in geographic range: how so many species coexist in the tropics. – Am. Nat. 132: 240 – 256. Taylor, D., Aarssen, L. and Loehle, C. 1990. On the relationship between r/K selection and environmental carrying capacity: a new habitat templet for plant life history strategies. – Oikos 58: 239 – 250. Terborgh, J. 1973. On the notion of favorableness in plant ecology. – Am. Nat. 107: 481 – 501. Terborgh, J. 1977. Bird species diversity on an Andean elevational gradient. – Ecology 58: 1007 – 1019. Waide, R. B. et al. 1999. The relationship between productivity and species richness. – Annu. Rev. Ecol. Syst. 30: 257 – 300. Whittaker, R. H. 1960. Vegetation of the Siskiyou Mountains, Oregon and California. – Ecol. Monogr. 30: 279 – 338. Whittaker, R. J., Willis, K. J. and Field, R. 2001. Scale and species richness: towards a general hierarchical theory of species diversity. – J. Biogeogr. 28: 453 – 470. Willig, M. R. and Lyons, S. K. 1998. An analytical model of latitudinal gradients of species richness with an empirical test for marsupials and bats in the New World. – Oikos 81: 93 – 98.

ECOGRAPHY 26:3 (2003)

ECOGRAPHY 28: 209 /222, 2005

The elevational gradient in Andean bird species richness at the local scale: a foothill peak and a high-elevation plateau Sebastian K. Herzog, Michael Kessler and Kerstin Bach

Herzog, S. K., Kessler, M. and Bach, K. 2005. The elevational gradient in Andean bird species richness at the local scale: a foothill peak and a high-elevation plateau. / Ecography 28: 209 /222. A monotonic decline in species richness with increasing elevation has often been considered a general pattern, but recent evidence suggests that the dominant pattern is hump-shaped with maximum richness occurring at some mid-elevation point. To analyse the relationship between species richness and elevation at a local scale we surveyed birds from lowlands to timberline in the Bolivian Andes. We divided the transect into 12 elevational belts of 250 m and standardized species richness in each belt with both individual- and sample-based rarefaction and richness estimation. The empirical data were then correlated to four explanatory variables: 1) area per elevational belt, 2) elevation (also representing ecosystem productivity), 3) a middomain effect (MDE) null model of geometrically constrained empirical range sizes, and 4) a hump-shaped model derived empirically for South American birds representing the regional species pool hypothesis. Local species richness peaked at ca 1000 m elevation, declined sharply to ca 1750 m, and then remained roughly constant. Elevation was the best single predictor, accounting for 78 /85% of the variance in the empirical data. A multiple regression model with elevation, area, and MDE explained 85 /90% of the variance. Monte Carlo simulations showed that the richness peak at 1000 m is the result of an overlap of two distinct avifaunas (lowland and highland) and that the correlation to MDE in the multiple regression was likely spurious. We recommend complementing correlation analyses involving MDE predictions with an examination of the distribution of range midpoints. The steep decline at mid-elevations was mainly due to a rapid loss of lowland species. The high-elevation plateau is striking and unexpected, but has also been found previously. It cannot be explained at present and exemplifies that despite several decades of research elevational gradients are still not well understood. S. K. Herzog (skherzog@compuserve.com), Inst. fu¨r Vogelforschung ‘‘Vogelwarte Helgoland’’, An der Vogelwarte 21, D-26386 Wilhelmshaven, Germany. / M. Kessler and K. Bach, Albrecht-von-Haller-Inst. for Plant Sciences, Systematic Botany, Untere Karspu ¨ le 2, D-37073 Go¨ttingen, Germany.

The relationship between species richness and elevation has received considerable attention in the ecological literature. Equivalent to the latitudinal gradient in species richness, a monotonic decline in the number of species with increasing elevation has often been considered a general pattern (Brown and Gibson 1983, Begon et al. 1990, Rohde 1992, Stevens 1992).

However, Rahbek (1995) argued that this generalization is largely a result of too much emphasis on, and misinterpretation of, a few early studies (Terborgh 1977) combined with ‘‘citation inbreeding’’. In an extensive literature review including studies on a wide range of taxonomic groups, biomes, and spatial scales, Rahbek (1995) found empirical support for several

Accepted 17 November 2004 Copyright # ECOGRAPHY 2005 ISSN 0906-7590 ECOGRAPHY 28:2 (2005)

209

different elevational richness patterns, with a dominance of hump-shaped relationships where maximum species richness occurs at some mid-elevation point. However, Rahbek (1995) also noted that most studies had methodological problems because they did not account for the effect of sampling effort and/or area on patterns of species richness. Failure to standardize data to account for sampling effort (Gotelli and Colwell 2001) and area (Rahbek 1995, 1997) can cause artefactual results. To overcome such pitfalls, Rahbek (1997) analysed a data set largely devoid of sampling biases on the elevational distribution of neotropical land birds. After controlling for the surface area of each elevational belt, the emerging regional-scale relationship between species richness and elevation was indeed hump-shaped with a maximum at ca 1000 m. Stotz et al. (1996) approached the problem from a different angle. As not only surface area but also habitat diversity is greater in lowland Amazonia than on adjacent Andean slopes, Stotz et al. (1996: 37) only considered those species among lowland birds that inhabit terra firme forest and obtained a similar result as Rahbek (1997), namely a hump-shaped specieselevation curve with a maximum in the foothill zone between 500 and 1000 m. Thus, both analyses indicate that the enormous bird species richness of the Amazonian lowland is largely a consequence of its huge surface area and increased habitat diversity, and that if one controls for those factors, the lower Andean slope emerges as the zone of highest bird species richness. Recent work on other groups of organisms also documented hump-shaped relationships between species richness and elevation, e.g. for trees in Costa Rica (Lieberman et al. 1996), for several plant groups in Bolivia (Kessler 2000c, 2001b), for ants in the western United States (Sanders 2002), and for non-volant small mammals in the Philippines (Heaney 2001), Malaysia (Md. Nor 2001), and Costa Rica (McCain 2004). However, several of those studies interpolated species presence between maximum and minimum observed elevations, which may cause an artificial hump in the species-elevation curve (Grytnes and Vetaas 2002). By contrast, Patterson et al. (1998) found a trough-shaped, but non-significant, relationship for murid rodents in Peru with richness maxima in the lowlands and in the highlands and a minimum at mid-elevations. For bats and birds, Patterson et al. (1998) reported a smooth, slightly curvilinear decline of richness with elevation, but their data were not standardized for area or habitat diversity. Patterson et al. (1998) remarked, however, that preliminary analyses of the same bird data set revealed a hump-shaped pattern once the effect of increased Amazonian habitat diversity was removed. These studies seemingly illustrate that mountainsides do not simply mirror the latitudinal diversity gradient (Brown 2001) and that there is no universal agreement in 210

the shape of the elevational pattern. Brown (2001) further argued that synthetic theories explaining the common patterns of species richness on elevational gradients have not been developed yet. Recent advances in null model theory, however, have added a new twist to an old story. Colwell and Hurtt (1994) proposed several models that simulate range size and randomise range placement within one-dimensional bounded geographical domains (e.g. an elevational gradient). Based entirely on stochastic processes these null models produce symmetrical curves with a mid-domain peak in species richness. These simulations are the ancestor of analytical null models (Willig and Lyons 1998, Lees et al. 1999, Jetz and Rahbek 2001) and analytical-stochastic null models (Colwell 2000, applied by Sanders 2002, McCain 2003, 2004) that have been tested against empirical data and extended to two-dimensional domains (Jetz and Rahbek 2001). These studies provide strong support for what has come to be called the mid-domain effect (MDE; Colwell and Lees 2000, Colwell et al. 2004). By contrast, other authors (Bokma and MoÂ¨nkkoÂ¨nen 2000, Koleff and Gaston 2001, Bokma et al. 2001, Hawkins and Diniz-Filho 2002, Diniz-Filho et al. 2002, Zapata et al. 2003) have dismissed a significant influence of MDE on spatial richness patterns, criticizing MDE model assumptions as unrealistic, conceptually flawed, or internally inconsistent. Colwell et al. (2004) contend that much of the MDE criticism is based on misunderstandings of the nature of null models and of MDE models in particular, the use of inappropriate frequency distributions of geographical range sizes, and an all-ornothing approach searching for single-factor explanations. To properly evaluate the role of MDE in shaping species richness patterns Colwell et al. (2004) recommend that studies should a) quantitatively assess the relative importance of MDE instead of simply testing it as a null hypothesis, b) use re-sampling of the empirical range size frequency distribution (rather than theoretical range size frequency distributions or empirical range midpoints) for making MDE predictions, and c) apply multivariate analyses considering candidate explanations for richness patterns in addition to MDE. Of the 16 onedimensional MDE studies found by Colwell et al. (2004) in an exhaustive literature review only one (McCain 2004) meets all three criteria. In the present study, we follow the recommendations of Colwell et al. (2004) in analysing bird survey data from a perhumid elevational gradient from lowlands to timberline in central Bolivia. Lowland surveys were restricted to terra firme forest, and data analyses controlled for survey effort. Using multiple regression analysis we correlated the resulting pattern of species richness with the following explanatory variables: 1) area, 2) elevation (also representing potential evapotranspiration and mean canopy tree height), 3) MDE using the empirical version of the Colwell and Lees ECOGRAPHY 28:2 (2005)

(2000) constrained range-size null model, and 4) the regional species pool using Rahbek’s (1997) humpshaped model for South American birds.

Study area We surveyed birds in and adjacent to Carrasco National Park on the east-Andean slope in the central Bolivian department of Cochabamba (Fig. 1). The area spans an elevational gradient from 200 to 4800 m. Its natural vegetation consists of evergreen, perhumid forest that once extended up to elevations of ca 4200 m. Due to centuries of human impact in the high Andes, the current timberline is lowered and generally found around 3400 m, but remnant forest patches remain at higher elevations (Kessler and Herzog 1998, Kessler 1999, 2000b). Foothill and lowland forests have been cleared extensively in recent decades, primarily for timber extraction, coca plantations, and road construction (Henkel 1995). Data on the vegetation of the study transect can be found in Ibisch (1996), Navarro (1997), and Kessler (2000a, 2001b). We studied a continuous elevational gradient at 950 / 3400 m in the Serranıá de Callejas (Fig. 1) using a ca 4-m wide gravel road abandoned in the 1980s. The adjacent forest was almost entirely in a natural state; few disturbed areas largely had reverted to montane forest, and the road partly had been invaded by woody plants. Human activity in the Serranıá de Callejas increased drastically below 1000 m, and we collected data in two foothill area’s. Between the villages of El Palmar and Villa Tunari (Fig. 1; 300 /900 m) we selected several sites that contained some of the area’s best-preserved primary or mature secondary forest. The second area at Rıó Ichoa-Cerro Len˜e (Fig. 1; 300 /750 m) contained pristine forest and was reached by helicopter. A seismic line (a straight trail just wide enough for one person) established several weeks before represented the only visible anthropogenic disturbance; no timber extraction or hunting had occurred. We surveyed lowland forest in the Valle del Sacta (Fig. 1; 220 m), a 5600-ha area of

Fig. 1. Location of Carrasco National Park in Bolivia and distribution of survey sites within the study area. ECOGRAPHY 28:2 (2005)

primary and mature secondary terra firme forest. This area is separated from Andean forests by a ca 10-km wide belt of human settlements. It is located in the southernmost extension of Amazonian evergreen lowland forest, which forms a 10 /30-km wide wedge at the base of the central Bolivian Andes. Further to the northeast evergreen forest is replaced by seasonally flooded savannah vegetation (Ribera et al. 1996). Geologically, the bedrock mainly consists of sandstones, lutites, and quartzitic rocks of Devonian and Ordovician age (Montes de Oca 1997, Ergueta and Go´mez 1997) with limited areas of granitic intrusives and calcareous rocks at 2000 /2200 m along the study road and some white sand areas at 450 /500 m. At Villa Tunari the geological substrate is composed of nutrientpoor white sandstones, whereas it consists of clayey red alluvial deposits in flat areas and more sandy soils on ridges in the Valle del Sacta. Mean annual precipitation at the region’s only reliable climatic station in Villa Tunari is 5676 mm (10 yr data; Ibisch 1996). Most precipitation falls from November to May, but even from June to October every month receives /100 mm. Ibisch (1996) estimated /3500 mm mean annual precipitation at 2200 m in the somewhat sheltered valley of Sehuencas (Fig. 1). Mid-elevation (1000 /3000 m) slopes directly exposed to incoming clouds in the northwest of the park likely receive /8000 mm (Kessler 1999). Mean annual precipitation at Sacta is /3000 mm (Acebey pers. comm.), with additional moisture provided by frequent morning fog. Mean annual temperatures are 24.68C at Villa Tunari and 12 /158C at 2200 m in Sehuencas (Ibisch 1996), with an annual variability of monthly means of ca 58C. Nocturnal frosts occur down to 2000 m (Ibisch 1996, unpubl.), especially during periodic influxes of southern polar winds during the austral winter (Fjeldsa˚ et al. 1999).

Methods Field surveys We studied the continuous Serranı´a de Callejas transect (950 /3400 m) from June to September 1996 and sections of it (950 /1500 m, 3000 /3400 m) again in October 1997. The El Palmar-Villa Tunari area (300 /900 m) and the Valle del Sacta (220 m) were studied in September and October 1996, and Rı´o Ichoa-Cerro Len˜e (300 /750 m) in September 1997. No surveys were conducted above timberline (3400 m). To factor out the effect of increased lowland habitat diversity on species richness, lowland field work was restricted to terra firme forest. The survey method is detailed in Herzog et al. (2002) and only briefly summarized here. While walking slowly and quietly from dawn to mid-day and often again from late afternoon to after dusk along roads, trails and 211

212

55 65 64 59 70 69 53 61 58 55 64 62 62 82 73 63 74 74 b

Number of 10-species lists (samples) compiled. Mean number of individuals per 10-species list (sample)9/SD. c Observed total species richness. d MMMean estimate of species richness after the maximum number of samples. e Sample-based MMMean estimate of species richness standardized for survey effort. f Observed species richness after 337 individuals. g Individual-based MMMean estimate after maximum number of individuals. h Individual-based MMMean estimate after 337 individuals.

82 121 109 103 162 146 100 139 139

85 125 111

72 68 82 78 71 60 76 66 88 71 84 79 129 120 174 158 149 149

132 119

20 18.59/6.0 369 56 20 18.59/6.8 369 61 21 19.99/6.5 417 57 20 20.59/8.2 409 58 36 17.69/6.4 633 76 20 16.99/6.9 337 63 49 15.69/4.9 765 106 56 14.79/3.9 821 110 51 14.19/3.2 719 133 27 12.59/1.9 337 100

Lists 45 80 12.69/2.7 14.79/14.0 Individuals per listb Total individuals 569 1175 Scobs 119 141 Sample-based rarefaction d 153 161 MMMean max MMMean stde 141 136 Individual-based rarefaction Sobs stdf 98 96 146 156 MMMean maxg 137 132 MMMean stdh

3000 /3249 1000 /1249 1250 /1499 1500 /1749 1750 /1999 2000 /2249 2250 /2499 2500 /2749 2750 /2999 500 /749 300-499 220

We divided the transect into 14 elevational belts of 250 m (B/250 m, 250 /499 m, 500 /749 m, etc.; Table 1). We restricted surveys to forest areas, but included data from small areas of natural (e.g. landslides) or anthropogenic (e.g. small coca fields) disturbance. Species occurring naturally at forest edge were included in all analyses; species depending on aquatic habitats were excluded. For analysing the relationship between species richness and elevation, data from the El Palmar-Villa Tunari area were excluded for two reasons (but they were included in the analysis of range midpoints). First, human disturbance probably introduced biases that we were unable to control for. Second, individual survey sites with suitable habitat were spaced relatively far apart and measured species richness included a strong component of beta diversity (sensu Whittaker 1972), rather than alpha diversity as on the remaining transect. Thus, the Rı´o Ichoa-Cerro Len˜e data were used for the 300 / 499-m and 500 /749-m belts, and no data were available for the 750 /999-m belt. Although surveys extended to 3400 m, the accessible area in the 3250 /3499-m belt was too small for meaningful data analysis and it was excluded. For the Serranı´a de Callejas only data collected in 1996 were used in this step of the analysis. Due to topographic complexity and variations in the amount of accessible habitat, it was not feasible to standardize survey area across all elevational belts. Survey distance (length of transect line) varied from 3.9 km (2750 /2999 m) to 10.3 km (300 /499 m) with a mean of 5.6 km.

Transect subdivision

Table 1. Observed and estimated species richness values of elevational belts studied along the Carrasco transect on the east Andean slope in central Bolivia. Only those belts included in the analysis of the relationship between species richness and elevation are shown. Elevation in m.

through the habitat where feasible, SKH continuously recorded all visual and acoustical observations of birds (including numbers of individuals per species) within 50 m of the observer. The observer’s movement rate largely depended on the level of bird activity. When spending longer periods in one spot and during very occasional resampling of an area (the latter occurred to approximately the same degree in all elevational belts and thus did not introduce a systematic error), repeated counts of obviously territorial individuals were avoided. Tape recordings were made extensively to supplement observations and to identify unknown voices, and were integrated into the master list of temporally consecutive observations. Fjeldsa˚ (1999) quantitatively compared this approach with standardized point counts. Its main advantages compared to any timed species-count method (e.g. point counts) are time efficiency and relative observer independence (Fjeldsa˚ 1999, Herzog et al. 2002).

ECOGRAPHY 28:2 (2005)

175

Standardisation of survey effort for species richness estimation

ECOGRAPHY 28:2 (2005)

Cumulative species richness

32.8%

125 100 75

67.2%

50 25 0

100

B Cumulative species richness

Survey effort was not exhaustive and varied between elevational belts, precluding the use of raw species counts in our analyses. Therefore, we used three methods to standardize species richness values for survey effort (Table 1). First, we used a modified version of the ‘‘m-species-list method’’ (MacKinnon and Phillipps 1993, Poulsen et al. 1997) following the recommendations of Herzog et al. (2002) (method 1). We divided the master list of temporally consecutive bird observations in each 250-m belt into lists of 10 species: the first list consists of the first 10 species observed, the second list includes the following 10 species and may contain species already found on the first list, and so on. We then plotted cumulative species number as a function of list number, treating each 10-species list as a separate sample. By randomising sample accumulation order 50 times using EstimateS 5.0.1 (Colwell 1997), we obtained samplebased rarefaction curves and estimated total species richness in each belt with the MMMean statistic (Fig. 2; Raaijmakers 1987, Keating and Quinn 1998). For species-rich neotropical bird data sets, MMMean was the least biased of nine estimators evaluated by Herzog et al. (2002), but a drawback of MMMean is that no statistically sensible variance estimator exists (Colwell and Coddington 1994, Colwell pers. comm.). Ideally, the MMMean curve for each elevational belt would quickly reach an asymptote after 10 /15 lists. As this was not the case, we standardized all data sets for survey effort following the procedure in Herzog et al. (2002) (Fig. 2). Gotelli and Colwell (2001) suggested that x-axes of sample-based rarefaction curves should be rescaled from samples to individuals because datasets may differ systematically in the number of individuals per sample. Thus, our second standardisation method (method 2) is derived from individual-based rarefaction. Accumulation order of individuals was randomised 50 times using EstimateS 5.0.1 (Colwell 1997), and we used the observed species richness after 337 individuals (the lowest number of individuals recorded in any elevational belt; Table 1) as the standardized cut-off point. In the third method, we determined MMMean values of estimated species richness after 337 individuals from the same individual-based rarefaction curves using EstimateS 5.0.1 (Colwell 1997) (method 3). Individual-based rarefaction probably is the least biased method for comparing species richness via species-accumulation curves (Gotelli and Colwell 2001), but it is not without problems. In cases with strongly differing sampling intensities, such as here (Table 1), it implies a loss of much valuable data. It also is sensitive to biases in the quantification of the number of individuals per species. Such biases are

100% 150

100%

32.8%

67.2% 20

Number of 10-species lists

Fig. 2. Sample-based rarefaction curves for bird data sets from two elevational belts (A: 500 /749 m; B: 2000 /2249 m) in Carrasco National Park, Bolivia. Observed (Sobs; circles) and estimated species richness using the MMMean statistic (squares) are expressed as a function of the number of 10species lists. Sample accumulation order of all curves was randomized 50 times, and each point represents the mean of the resulting 50 values. To control for the confounding effects of survey effort, we determined a standardized cut-off point from the relation between the Sobs and the MMMean curve: for each data set, every Sobs value was expressed as the proportion of the respective MMMean value (Herzog et al. 2002). Survey effort was lowest in the 500 /749-m belt (A), where Sobs comprised 67.2% of the estimated richness (149 species) at maximum sample size (27 10-species lists). The equivalent cut-off point was determined for all other data sets as illustrated for the 2000 /2249-m belt (B), where a standardized estimate of 71 species was obtained.

inevitable when studying birds in tropical forests, where many birds are only seen briefly and where many species move in large mixed-species flocks. Also, the presence (or absence) of a few large single-species flocks of some species, e.g. swifts or parrots, can strongly influence total individual numbers and thereby the slope of the accumulation curve, artificially inflating survey effort. Finally, whereas species richness values obtained by individual-based rarefaction may accurately reflect relative differences in species richness between sites, they almost invariably underestimate true species richness and usually are lower than the total observed species numbers. 213

Species richness estimation based on samples such as m-species lists is less sensitive to these problems by including all or most of the collected data, by counting a species only once when two or more individuals are registered together, and by providing species richness values somewhat above the observed values. However, this method also has three important weaknesses (Gotelli and Colwell 2001). First, different estimation methods provide different results, and although there is some empirical data to suggest which statistic might be suitable for a given data set, no general test exists to choose the least biased estimator (of course this also applies to individual-based richness estimation). In general, the response of estimators to differences in species abundance distributions, species richness, etc. is poorly understood. Second, the mean number of individuals per list tends to be higher in species-poor habitats, simply because more birds have to be recorded to accumulate 10 (or five, or 20) species. This leads to relatively steeper accumulation curves in species-poor habitats, and hence an overestimation of species richness relative to species-rich habitats. Finally, and perhaps most importantly, the arbitrary decision on how many species to include in each list will influence the estimated values. High species numbers per list will reduce the number of lists and lead to relatively steeper accumulation curves and higher estimates (Herzog et al. 2002). All three error sources will be stronger in cases with more pronounced differences in actual species richness between sites. Given the potential biases of both approaches for obtaining comparable, standardised species richness values, here we chose to apply both sample- and individual-based rarefaction. As the elevational richness patterns obtained by both approaches are highly correlated (Spearman rank correlation, values of standardization method 1 vs method 2: r /0.98, pB/0.0001; method 1 vs 3: r /0.99, pB/0.0001; method 2 vs 3: r /1.00, p / 0), we conclude that our raw data are fairly robust to standardization, and that both standardization approaches accurately reflect the richness pattern captured by our field data. On the other hand, since the raw species counts show lower correlation values with the standardized data (Spearman rank correlation, raw data vs method 1: r /0.85; raw data vs method 2: r /0.88; raw data vs method 3: r /0.87; pB/0.001 in all cases), we conclude that standardization is appropriate.

Explanatory variables Area / due to the conical shape of mountains, land surface area decreases steadily with increasing elevation (Graves 1988, Rahbek 1997). The effect of the area of elevational belts on species richness at the regional scale was illustrated by Rahbek (1997), but the procedures 214

used therein to control for area are not applicable to our local-scale data set. Rahbek (1997) also used belts of greater amplitudes than analysed here. As a proxy for area we measured the horizontal width of each 250-m belt in central Bolivia on topographic maps (scale 1:50 000) issued by the Inst. GeograÂ´fico Militar, La Paz, Bolivia. For the lowland belt we measured only the area covered by evergreen terra firme forest as surveys were restricted to this habitat, excluding open savannah further east. We took five measurements for each belt (one in the study area, two to the north, and two to the south at 10 km intervals), and the mean was calculated and rounded to the closest 100 m (Table 2). As species richness does not increase linearly with area, we multiplied the mean horizontal width of each belt by an areadependent factor, assuming a slope of z /0.13 in a double-log species-area plot, which corresponds to the mean z value obtained for birds in the tropical Andes by Rahbek (1997: Fig. 2c). Elevation / this represents the model of a monotonic decline of species richness with elevation (Stevens 1992). We used the mean elevation of each belt, except for the lowland belt, where we used the actual elevation (Table 2). Ecosystem productivity / Kessler (2001b) determined potential evapotranspiration (PET) calculated after Thornthwaite and Mather (1957) and mean canopy tree height in mature forest as indices of ecosystem productivity (Rosenzweig 1968, Lieth 1975) for our study area. Within the boundaries of the present gradient, both PET (Pearson correlation: r2 /1.00, p /0) and mean canopy tree height (Pearson correlation: r2 /0.95, p B/0.0001) closely correlate negatively with elevation. Hence, only elevation is used in the regression analysis. MDE / we used the richness values predicted by the empirical version of the Colwell and Lees (2000: Box 5) constrained range-size null model, shuffling empirical range size distributions by random mid-point Monte Carlo simulations using RangeModel 3.1 (Colwell 2000: Model 4). Upper and lower domain limits were defined as the natural upper limit of humid forest on the eastern slope of the central Andes (ca 4200 m) and the lower limit of Amazonian evergreen forest (sea level), respectively. Only those species whose elevational distribution throughout the neotropics falls entirely within these domain limits were included in the model. Upper and lower elevational range limits were taken from standard references on neotropical birds (Ridgely and Tudor 1989, 1994, FjeldsaË&#x161; and Krabbe 1990, Hoyo et al. 1992 /2002, Stotz et al. 1996, Isler and Isler 1999, Ridgely and Greenfield 2001, Hennessey et al. 2003). Unlike Lees et al. (1999), who defined their domain as the present day extent of forest cover on Madagascar, we did not use the present elevation of the closed upper timberline (ca 3400 m) as the upper domain limit because remnant forest patches of varying sizes still ECOGRAPHY 28:2 (2005)

Horizontal width in km determined from topographical maps (mean of five measurements rounded to the closest decimal). Mean elevation of each belt, representing both elevation per se as well as PET and mean canopy tree height as indices of ecosystem productivity. Values predicted by the empirical version of the Colwell and Lees (2000: Box 5) and Colwell (2000: Model 4) constrained range-size MDE null model for 548 species. e Values predicted by Rahbek’s (1997) regional hump-shaped model for South American birds.

0.4 3125 268 351 0.5 2875 318 407 0.5 2625 353 463 0.5 2375 376 512 0.6 2125 387 561 0.6 1875 384 606 0.7 1625 364 638 0.7 1375 334 666 0.8 1125 288 682 1.6 625 174 662 3.0 400 116 630

3000 /3249 2750 /2999 2500 /2749 2250 /2499 2000 /2249 1750 /1999 1500 /1749 1250 /1499 1000 /1294 500 /749 300 /499 220

15.0 220 70 589 Area Elevationb MDEc Species poole

Table 2. Values of the explanatory variables for each elevational belt included in the analysis of the relationship between species richness and elevation. Elevation in m.

ECOGRAPHY 28:2 (2005)

exist up to elevations of 4200 m in the study area, in part only a few kilometers from our actual study sites (see, e.g. photos in Kessler 2000b). While these patches were not accessible to us, they certainly provide suitable habitat for the local forest-based avifauna, and the upper elevational distribution of the bird species is consequently not limited to the forest line at 3400 m along the study road. Extensive ornithological surveys in Carrasco National Park and adjacent lowland forests in recent years by a number of field workers (especially R. MacLeod, J. Fjeldsa˚, and collaborators) have resulted in a nearly complete inventory of the area’s avifauna (Herzog et al. unpubl.). With the exception of species depending primarily on aquatic habitats and lowland species not found in terra firme forest, all recorded species that inhabit humid forest were included in the null model computations. An additional 19 species that have gone unrecorded so far but that are expected to occur in the area also were included, resulting in a total of 548 species. Monte Carlo simulations were replicated 20 times and mean richness values were computed for 60 domain divisions (the default of RangeModel 3.1). Predicted values for each 250-m belt were taken from the resulting curve (Table 2). As some of the MDE debate has focused on whether to use theoretical or empirical range size frequency distributions for MDE predictions (Colwell et al. 2004), we also computed the bivariate uniform model of Colwell and Lees (2000: Box 2), which is equivalent to the Colwell and Hurtt (1994) null model 2 of bounded random geographical ranges, for 548 species using RangeModel 3.1 (Colwell 2000: Model 1). This theoretical model takes no account of the empirical range size frequency distribution. As above, Monte Carlo simulations were replicated 20 times and mean richness values were computed for 60 domain divisions. As both null models were highly correlated (Pearson correlation: r2 / 0.99, pB/0.0001; however, the constrained range size model predicted higher absolute values than the bivariate uniform model) only the constrained range size model is used in the analysis. Regional species pool / the regional species pool has been proposed as a significant determinant of local species richness as it places an upper limit on the number of species potentially able to colonise local habitats (Cornell and Lawton 1992, Caley and Schluter 1997). We considered Rahbek’s (1997) regional hump-shaped model for South American birds as the most appropriate regional source pool model and used the richness values predicted by this model (Fig. 4 in Rahbek 1997; values taken from the fitted curve) for each of our elevational belts (Table 2). It may be argued that this model is incorrect to some degree as it is based on partly interpolated species-distribution data (see Grytnes and Vetaas 2002). However, at the broad spatial scale 215

considered by Rahbek (1997) we regard the data on South American bird distributions as sufficiently complete so as to render the influence of interpolation on the species-elevation curve negligible.

Multiple regression analysis We first performed bivariate linear regressions of the empirical species richness pattern (separately for each of the three standardization methods) against each of the four explanatory variables individually, followed by multiple regression analysis. Due to the spatial proximity of most elevational belts, estimates of species richness violate statistical assumptions of independence, and the present analysis therefore focuses on the proportion of variance explained rather than on probability values. All regression and correlation analyses were performed using STATISTICA for Windows (ver. 5.1, Anon. 1997).

Monte Carlo simulations: range midpoints and turnover To asses whether species are randomly distributed along the elevational gradient or show clear zonations, we conducted Monte Carlo simulations with a program written by KB in Visual Basic within EXCEL. Observed elevational ranges of all recorded species were randomly placed along the elevational gradient treating the gradientâ&#x20AC;&#x2122;s end points as hard boundaries not to be crossed by any speciesâ&#x20AC;&#x2122; range. Total species number per belt was constrained in two ways, according to the MMMean-standardized observed richness pattern, and to the richness pattern predicted by the MDE constrained range size model. Early simulation trials showed that not all species could be accommodated in the simulations using the observed standardized species richness. This was a result of tight species packing in the empirical data, whereas in the randomised data many gaps between species could not be filled because they were narrower than the species ranges remaining to be placed. Therefore, the observed standardized data was multiplied by the factor 1.5. The MDE model values were high enough that this correction was not necessary. Randomisations were repeated 1000 times, and we calculated the mean and 95% confidence intervals for three parameters (midpoints per belt, number of upper/ lower elevational limits per belt, species turnover between adjacent belts) explained below. These values were compared to the observed data, assuming that empirical and predicted values were significantly different if the observed value fell outside the 95% confidence intervals of the predicted value. Midpoints: as hump-shaped patterns that may correlate to MDE predictions can arise from a variety of 216

different processes, we compared not only the observed and predicted species richness patterns but also the empirical and predicted range midpoint distributions, counting the number of midpoints per belt (Grytnes 2003). Upper and lower limits: to asses whether species turnover along the gradient is homogeneous, we separately compared the number of upper and lower elevational limits per belt of empirical vs simulated data. Turnover: as a further test for species zonation, species turnover along the gradient was measured with the Wilson-Shmida index (Wilson and Shmida 1984). The index is calculated for pairs of adjacent elevational belts as follows: b (b c)=(2a b c) where a is the number of species recorded in both belts, and b and c the number of species lost and gained, respectively. The more dissimilar two belts are, the higher is the index, reaching a maximum of 1 at total dissimilarity. The Wilson-Shmida index produces results comparable to those of other b-diversity indices (Davis et al. 1999, Koleff et al. 2003).

Results We recorded a total of 449 bird species, 38 of which were found only in the El Palmar-Villa Tunari area. Species richness peaked around 1000 m and slightly decreased towards the lowlands (Fig. 3, Table 1). Above 1250 m species richness decreased quickly to a minimum at 2500 m, followed by a slight increase towards the highest elevation. In general, however, variation above 1750 m was slight and richness remained surprisingly constant, generating a high-elevation plateau (Fig. 3, Table 1). Examining the congruence between empirical and model data (Fig. 3, Table 3), elevation (also representing ecosystem productivity) was the best predictor variable explaining 78 /85% of the variance in the empirical species richness pattern. The regional species pool was a moderate predictor of the empirical pattern, explaining 59 /63% of its variance. At low elevations it was largely in accordance with the empirical curve, but above 1250 m observed species richness decreased much faster than predicted by this model (Fig. 3D). Area had little explanatory power, concurring with the empirical pattern only in the lowlands (Fig. 3A). MDE predictions were negatively correlated to and strongly contrasted with the observed species richness pattern. The former reached its peak at elevations where the latter approached its minimum (Fig. 3C), and the maximum richness predicted by the MDE model is over twice as high as the empirical richness peak (Table 1). The best multiple regression model was elevation combined with area and MDE, explaining 85 /90% of ECOGRAPHY 28:2 (2005)

100

Relative species richness

Fig. 3. Comparison of the empirical richness pattern (circles; standardized for survey effort by MMMean estimates for samples) and the patterns predicted by the four explanatory models examined (squares): (A) area, (B) elevation, (C) mid-domain effect, and (D) regional species pool. Empirical data were not standardized for area since area was included as an explanatory model. Curves were fitted by distance-weighted least-squares smoothing.

100

30 20

20 10

500

1000

1500

2000

2500

100

3000

20 0

500

1000

1500

2000

2500

3000

500

1000

1500

2000

2500

3000

100

500

1000

1500

2000

2500

3000

Elevation (m)

the variance in the empirical species richness pattern (Table 3). The multiple regression of richness vs. regional species pool and area combined also obtained high determination coefficients of 0.72 /0.79 (Table 3). The proportion of observed range midpoints was significantly higher at 250 /750 m and 3000 /3250 m than predicted by the Monte Carlo simulations (Fig. 4). Thus, the empirical richness peak at ca 1000 m results at least partly from a large number of species restricted to low elevations. The significant surplus of elevational midpoints at 3000 /3250 m might be a sampling artefact as species mainly occurring at higher elevation and barely entering our study gradient, are falsely considered as having their mid-points in the highest survey belt.

Further, the results of the analyses of range midpoints may be biased by two shortcomings in our data set, i.e. the sampling gap at 750 /999 m and the fact that the area from 300 to 750 m was surveyed in a separate location. Both shortcomings may potentially inflate the number of midpoints at elevations of 300 /750 m because species actually reaching up to 750 /999 m may be misinterpreted as having their upper elevational limits at lower elevations, and because species found only at the separate survey location could not be recorded above 750 m. However, we doubt that these potential biases are a major problem, for two reasons. First, the survey data from the El Palmar-Villa Tunari area were included in determining the elevational ranges of species. As a result,

Table 3. Linear determination coefficients (r2) between observed species richness standardized by three methods and four explanatory model response variables. The upper four lines show the individual regression values, the lower six lines multiple regressions combining two and three parameters (starting with elevation as a fixed variable because it has the highest single explanatory power). Since Rahbekâ&#x20AC;&#x2122;s (1997) regional species pool model is corrected for area, we also performed a multiple regression combing species pool and area. Note the consistency in results between the three richness estimation methods. Sample-based standardized MMMean estimate Areaa Elevation (E)b MDEc Species poole E and area E and MDE E and species pool E and area and MDE E and area and species pool Species pool and area

0.37 0.78 0.43 0.59 0.81 0.78 0.79 0.85 0.82 0.72

Sobs after 337 individuals 0.43 0.85 0.46 0.63 0.87 0.85 0.86 0.90 0.88 0.79

Individual-based standardized MMMean estimate 0.43 0.84 0.47 0.60 0.85 0.84 0.84 0.88 0.86 0.77

Horizontal width in km determined from topographical maps (mean of five measurements rounded to the closest decimal). Mean elevation of each belt, representing both elevation per se as well as PET and mean canopy tree height as indices of ecosystem productivity. c Empirical version of the Colwell and Lees (2000: Box 5) and Colwell (2000: Model 4) constrained range-size MDE null model for 548 species. e Rahbekâ&#x20AC;&#x2122;s (1997) regional hump-shaped model for South American birds. b

ECOGRAPHY 28:2 (2005)

217

100

Number of species

Percentage of species

0 0

500

1000

1500

2000

2500

3000

Elevation (m)

empirical range midpoints are affected only by an elevational gap of 50 m (900 /950 m), and it is unlikely that such a negligible a gap will inflate the number of range midpoints below it. Second, with regard to the separate location of the area from 300 to 750 m, of the 127 species that had their midpoints within this elevational range only 16 (13%) were exclusive to that area. For these 16 species, the mean upper elevational limit in Bolivia (based on Hennessey et al. 2003) is 1440 m. This implies that their elevational midpoints would be located at ca 700 m elevation. As a result, even if surveys at the separate site had extended to higher elevation and recorded the full elevational ranges of these species, their range midpoints would still be located well below the elevation predicted by the MDE model. Regarding species zonation, significantly more lower limits than expected by chance occurred at 220 /500 m, 1000 /1250 m, and 3000 /3250 m, and significantly fewer lower limits at 500 /750 m (Fig. 5A). Upper elevational limits showed significantly higher values at 500 /750 and 1500 /1750 m and significantly lower values at 250 /500 m, 1750 /2000 m, and above 2500 m (Fig. 5B). Turnover was significantly higher than expected from the Monte Carlo simulations at 750 /1250, 1750 and 2500 m, and lower at 200 /500, 2000, and 3000 m (Fig. 6). Taken together, these analyses show that the change of species composition along the elevational gradient is not smooth. Rather, areas of rapid change and others of limited turnover exist. However, the location of peaks of upper and lower elevational species limits and of peaks in species turnover are not concordant and it is difficult to point out distinct elevational zones. 218

1000

1500

2000

2500

3000

B 50

Number of species

Fig. 4. Proportion of empirical range midpoints (open circles) and those predicted by Monte Carlo simulations restricting species richness per elevational belt according to the empirical richness pattern (closed circles) and MDE model predictions (squares). Vertical bars indicate 95% confidence intervals of the predicted values based on 1000 randomizations.

500

0 0

500

1000

1500

2000

2500

3000

Elevation (m)

Fig. 5. Number of lower (A) and upper (B) elevational range limits observed (open circles) and predicted by Monte Carlo simulations restricting species richness per elevational belt according to the empirical richness pattern (closed circles). Vertical bars indicate 95% confidence intervals of the predicted values based on 1000 randomizations.

Discussion The elevational pattern of bird species richness at the local scale documented here shows three distinct zones: a slight increase from the lowlands to a maximum at ca 1000 m, a sharp decline at 1250 /1750 m, and roughly constant values up to the highest survey elevation at 3250 m. Of course, at even higher elevations at and above timberline species richness will decrease. Among the four models examined, this pattern best fits the elevation model, although there are noticeable discrepancies. Neither the slightly lower values in the lowlands, nor the sharp decline at mid-elevations, nor the richness plateau at high elevations are adequately predicted by this model. In the multiple regression model, these divergences are partly accounted for by the additional factors included. Area predicts fairly constant and high values at higher elevations, thus modelling the highelevation plateau. The MDE model, by contrast, predicts a mid-elevation peak and, although negatively correlated with the observed pattern when compared directly, in the multiple regression model, it explains the ECOGRAPHY 28:2 (2005)

0.4

Tur nover

0.3

0.2

0.1

0.0 0

500

1000

1500

2000

2500

3000

Elevation (m)

Fig. 6. Species turnover between adjacent elevational belts calculated with the Wilson-Shmida index (Wilson and Shmida 1984) for the empirical data (open circles) and predicted by Monte Carlo simulations restricting species richness per elevational belt according to the empirical richness pattern (closed circles). Vertical bars indicate 95% confidence intervals of the predicted values based on 1000 randomizations.

peak in the foothills and the decrease of species richness towards the lowlands. The rather good correlation between the observed richness pattern and elevation indicates that local ecosystem properties may be involved in shaping the pattern. However, for two reasons we are unable to pinpoint which local factors might be responsible. First, most local factors co-vary closely along the gradient and cannot be disentangled. Potential evapotranspiration and canopy height decline almost linearly with elevation (see Methods). Other factors not quantified by us, e.g. insect abundance and habitat complexity, very likely decline in a similar fashion with elevation. Which of these factors or factor combinations are ultimately the most important can only be determined by comparing different transects where these variables show different distributions. Second, there is no consensus about the general relationship between richness and productivity. Both monotonic and hump-shaped relationships have been documented (Waide et al. 1999, Mittelbach et al. 2001). As yet, no comprehensive theoretical framework exists for the mechanisms involved in productivityrichness relationships (Currie et al. 1999, Ricklefs 2004). Regardless of the causes for the correlation between the empirical pattern and elevation, we are still faced with explaining the three deviations outlined above, i.e. the foothill hump, the sharp mid-elevation decline, and the high-elevation plateau. Among the models considered here, the foothill peak is best predicted by the regional species pool model. Rahbek (1997) concluded that the regional skewed mid-elevation peak is best explained by geometric boundary constraints, i.e. MDE, whereas the details of the pattern are governed by other factors. However, our Monte Carlo simulations suggest that the contribution of MDE to the multiple ECOGRAPHY 28:2 (2005)

regression model is likely spurious. If this model were to explain the foothill richness peak, we would also expect a peak of elevational midpoints at the same elevation. This simply is the result of the nature of the mid-domain null model. Species are randomly placed along the gradient, accumulating at mid-elevations because species with large elevational amplitudes must occur in the middle of the gradient (Colwell and Lees 2000, Colwell et al. 2004). As shown by Laurie and Silander (2002), Grytnes (2003), and our simulations (Fig. 4), this leads to a slight mid-domain peak of range midpoints. By contrast, in our empirical data midpoints are accumulated at low elevations. In combination with the zonation simulations (Figs 5 and 6), this can be interpreted as reflecting a prominent low-elevation avifauna. This avifauna can be detected by the high number of lower elevational limits at 220 /500 m, the surplus of midpoints at 250 /750 m, and the high number of upper elevational limits at 500 / 1250 m. Thus, the peak of species richness at 1000 m apparently results from an overlap of a distinct lowland with a distinct highland avifauna, and coincides with the significant peak of species turnover at 750 /1250 m. When assigning all recorded species to mutually exclusive categories of lowland species (present below 300 m) and highland species restricted to elevations above 300 m, this overlap and richness accumulation becomes apparent (Fig. 7). Although this seems trivial at first sight, there is no a priori reason to assume that the overlap between the two regional species pools leads to a hump-shaped richness distribution. It is equally conceivable that the decline of lowland species is sufficiently strong and/or the increase of highland species so weak that the resulting overall richness curve peaks in the lowlands. The sharp decline of species richness at 1250 /1750 m is largely caused by the rapid loss of lowland species. This decline is evidenced in the large number of upper elevational limits at this elevation (Fig. 5B) and causes high species turnover at 1750 m (Fig. 6). The highelevation plateau at 1750 /3250 m is puzzling, although we are not the first to document this pattern. Local bird species richness in Manu National Park of south-central Peru (Patterson et al. 1998: Fig. 1a) shows a similar plateau with almost identical richness values, but this phenomenon is not discussed. For geometrid moths in southeast Ecuador, Brehm et al. (2003) also found roughly constant values at 1800 /2700 m. A highelevation richness plateau is not predicted by Rahbekâ&#x20AC;&#x2122;s (1997) hump-shaped model or by the geometric null model, and strongly disagrees with the productivity hypothesis. Currently we have no plausible explanation for it. Although area apparently accounts for the plateau to a certain degree in our multiple regression model, we cannot conceive any causal relationship. Comparing our results with those of other avian transect studies in the Neotropics, we find substantial 219

200

All species Montane species Lowland species

Sp ec i es r i c h n es s

160

120

500

1000

1500

2000

2500

3000

3500

Elevation (m)

Fig. 7. Changes in species composition along an elevational gradient on the east Andean slope in central Bolivia. Montane species were defined as restricted to areas above 300 m and absent from the lowlands, whereas lowland species may or may not occur in montane areas. Elevational ranges of species were interpolated to all elevations between the recorded minimum and maximum of each species. Since interpolation added no species to the lowest and highest elevational belt, values for these belts are negatively biased, whereas a sampling bias accounts for the minimum at 750 /999 m. Note the nearly constant number of montane species between 1750 and 3250 m.

similarities. When removing the effect of increased lowland habitat diversity from their Manu data set, Patterson et al. (1998) found that species richness peaked around 1000 m rather than in the lowlands, i.e. exactly at the same elevation as in our study. Stotz et al. (1996: Fig. 3.3) reported similar cases. Whereas absolute richness values of these studies and ours coincide above about 2000 m, our values are much lower on the remainder of the transect. The avifauna of lowland terra firme forest in our study area is depauperate due to its semi-isolated location at the end of a narrow (10 /30-km wide), ca 500km long band of lowland forest wedged in between the Andes and open savannah habitats. Consequently, many bird species typical of Amazonian forest are lacking despite ecological conditions that in principle appear suitable. Blake and Loiselle (2000) found a diversity peak at 500 m in Costa Rica based on mist net captures, but point count data from the same transect indicated highest richness at 50 m. Although Blake and Loiselle (2000) standardized their data for sampling effort, they did not account for the effect of area. A reanalysis controlling for area seems worthwhile and we suspect that a hump-shaped pattern would result regardless of the survey method. Terborgh (1977: Fig. 5) also found a richness maximum at mid-elevation in Peru when controlling his data for sampling effort, but as noted by Rahbek (1995), only Terborgh’s non-standardized graph depicting a monotonic decline of species richness is usually cited even in recent papers (Heaney 2001). In conclusion, when excluding birds of seasonally flooded forests, a maximum of bird species richness in the foothills appears to be a general pattern in the Neotropics. 220

Taking into account other taxonomic groups along the same gradient can be an effective way to determine general species richness patterns (Heaney 2001). Two other groups of organisms have been studied in Carrasco at largely the same sites as birds. Ko¨hler (2000) surveyed amphibians at 500 /2500 m, and species richness peaked at 1500 m. However, amphibian species richness is about four to five times higher in adjacent lowland forests than at 1500 m (Ko¨hler and Reichle pers. comm.). This probably reflects the influence of temperature on the diversity of ectothermic organisms and the absence of lentic waters in montane forests required for the reproduction of many amphibians (Ko¨hler 2000). By contrast, alpha diversity of most of the six plant groups sampled by Kessler (2001a, b) on the Carrasco transect did show a mid-elevation peak (except for a monotonic decline in Melastomataceae), but individual maxima varied from elevations of 500 to 3000 m. Thus, while most taxa appear to show a humped richness-elevation relationship, the actual elevation of the peaks varies to such a degree that no single explanation seems plausible. In conclusion, the hump-shaped relationship between elevation and species richness of neotropical birds appears to be caused by the overlap between two large regional species pools, i.e. lowland and highland. Although at a first glance the resulting pattern may appear to be driven by MDE, at least in our study this is not the case. As outlined by Grytnes (2003) close examination of the elevational distribution of empirical range midpoints in indispensable for correctly interpreting any correlation between observed richness and MDE predictions. We suspect that some of the elevational gradient studies that have supported MDE models simply on the basis of correlations of species numbers without considering range midpoints would have to be re-assessed. It also is important in this context to distinguish latitudinal from elevational patterns. Elevational gradients typically cover much shorter spatial distances, and local population-level processes, such as source-sink dynamics (Wiens and Rotenberry 1981, Pulliam 1988, Grytnes 2003), can therefore have a stronger effect on richness patterns. For example, it would be interesting to examine the reproductive output of bird populations along elevational gradients to establish if high species richness at the overlap between the lowland and highland faunas corresponds to an accumulation of sink populations. The high-elevation plateau documented by us further exemplifies that despite several decades of ecological research, elevational gradients are still not well understood and other unexpected patterns may remain to be discovered. Acknowledgements / We thank J. Balderrama, A. Green, B. Rios, A. Rojas, R. Soria, and L. Tangara for help in the field and S. Mayer for assistance with the identification of tape recordings. We are grateful to J. Aparicio and C. Quiroga, Coleccioń Boliviana de Fauna, La Paz, for logistical support, to the Direccioń Nacional de Conservacioń de la Biodiversidad, ECOGRAPHY 28:2 (2005)

La Paz, for work permits, to I. Da´valos for access to Carrasco N.P., and to the Univ. Mayor de San Simo´n, Cochabamba, for access to Valle del Sacta. SKH would like to thank F. Bairlein for guidance and support. We also thank R. K. Colwell, N. Gotelli, C. Rahbek, and T. Romdal for helpful discussions and for providing unpublished manuscripts. F. Bairlein, R. K. Colwell, J. A. Grytnes, and C. Rahbek made valuable comments on earlier drafts of this paper. Financial support was provided to SKH by the Gesellschaft fu¨r Tropenornithologie and to MK by the Deutsche Forschungsgemeinschaft and the DIVA project under the Danish Environmental Programme.

References Anon. 1997. STATISTICA for Windows. / Statsoft, Tulsa, OK. Begon, M., Harper, J. L. and Townsend, C. R. 1990. Ecology: individuals, populations and communities. / Blackwell. Blake, J. G. and Loiselle, B. A. 2000. Diversity of birds along an elevational gradient in the Cordillera Central, Costa Rica. / Auk 117: 663 /686. Bokma, F. and Mo¨nkko¨nen, M. 2000. The mid-domain effect and the longitudinal dimension of continents. / Trends Ecol. Evol. 15: 288 /289. Bokma, F., Bokma, J. and Mo¨nkko¨nen, M. 2001. Random processes and geographic species richness patterns: why so few species in the north. / Ecography 24: 43 /49. Brehm, G., Suessenbach, D. and Fiedler, K. 2003. Unique elevational diversity patterns of geometrid moths in an Andean montane rainforest. / Ecography 26: 456 /466. Brown, J. H. and Gibson, A. C. 1983. Biogeography. / Mosby. Brown, J. H. 2001. Mammals on mountainsides: elevational patterns of diversity. / Global Ecol. Biogeogr. 10: 101 /109. Caley, J. and Schluter, D. 1997. The relationship between local and regional diversity. / Ecology 78: 70 /80. Colwell, R. K. 1997. EstimateS: Statistical estimation of species richness and shared species from samples. Version 5. / User’s guide and application published at: B/http:// viceroy.eeb.uconn.edu/EstimateS /. Colwell, R. K. 2000. RangeModel: a Monte Carlo simulation tool for assessing geometric constraints on species richness. Version 3. / User’s guide and application published at: B/http://viceroy.eeb.uconn.edu/RangeModel /. Colwell, R. K. and Coddington, J. A. 1994. Estimating terrestrial biodiversity through extrapolation. / Philos. Trans. R. Soc. Lond. B 345: 101 /118. Colwell, R. K. and Hurtt, G. C. 1994. Nonbiological gradients in species richness and a spurious Rapoport effect. / Am. Nat. 144: 570 /595. Colwell, R. K. and Lees, D. C. 2000. The mid-domain effect: geometric constraints on the geography of species richness. / Trends Ecol. Evol. 15: 70 /76. Colwell, R. K., Rahbek, C. and Gotelli, N. J. 2004. The middomain effect and species richness patterns: what have we learned so far? / Am. Nat. 163: E1 /E23. Cornell, H. V. and Lawton, J. H. 1992. Species interactions, local and regional processes, and limits to the richness of ecological communities: a theoretical perspective. / J. Anim. Ecol. 61: 1 /12. Currie, D. J., Francis, A. P. and Kerr, J. T. 1999. Some general propositions about the study of spatial patterns of species richness. / EcoScience 6: 392 /399. Davis, A. L. V., Scholtz, C. H. and Chown, S. L. 1999. Species turnover, community boundaries and biogeographical composition of dung beetle assemblages across an altitudinal gradient in South Africa. / J. Biogeogr. 26: 1039 /1055. Diniz-Filho, J. A. F. et al. 2002. Null models and spatial patterns of species richness in South American birds of prey. / Ecol. Lett. 5: 47 /55. Ergueta S. P. and Go´mez C. H. (eds) 1997. Directorio de Areas Protegidas de Bolivia. / CDC, Bolivia. ECOGRAPHY 28:2 (2005)

Fjeldsa˚, J. 1999. The impact of human forest disturbance on the endemic avifauna of the Udzungwa Mountains, Tanzania. / Bird Conserv. Int. 9: 47 /62. Fjeldsa˚, J. and Krabbe, N. 1990. Birds of the high Andes. / Apollo Books. Fjeldsa˚, J., Lambin, E. and Mertens, B. 1999. Correlation between endemism and local ecoclimatic stability documented by comparing Andean bird distributions and remotely sensed land surface data. / Ecography 22: 63 /78. Gotelli, N. J. and Colwell, R. K. 2001. Quantifying biodiversity: procedures and pitfalls in the measurement and comparison of species richness. / Ecol. Lett. 4: 379 /391. Graves, G. L. 1988. Linearity of geographic range and its possible effect on the population structure of Andean birds. / Auk 105: 47 /52. Grytnes, J. A. 2003. Ecological interpretations of the middomain effect. / Ecol. Lett. 6: 883 /888. Grytnes, J. A. and Vetaas, O. R. 2002. Species richness and altitude: a comparison between null models and interpolated plant species richness along the Himalayan altitudinal gradient, Nepal. / Am. Nat. 159: 294 /304. Hawkins, B. A. and Diniz-Filho, J. A. F. 2002. The mid-domain effect cannot explain the diversity gradient of Nearctic birds. / Global Ecol. Biogeogr. 11: 419 /426. Heaney, L. R. 2001. Small mammal diversity along elevational gradients in the Philippines: an assessment of patterns and hypotheses. / Global Ecol. Biogeogr. 10: 15 /39. Henkel, R. 1995. Coca (Erythroxylum coca ) cultivation, cocaine production, and biodiversity in the Chapare region of Bolivia. / In: Churchill, S. P. et al. (eds), Biodiversity and conservation of neotropical montane forests. The New York Botanical Garden, NY, pp. 551 /560. Hennessey, A. B., Herzog, S. K. and Sagot, F. 2003. Lista anotatda de las aves de Bolivia. Quinta edicioń. / Asociacioń Armonıá/BirdLife International. Herzog, S. K., Kessler, M. and Cahill, T. M. 2002. Estimating species richness of tropical bird communities from rapid assessment data. / Auk 119: 749 /769. Hoyo, J. del, Elliott, A. and Sargatal, J. (eds) 1992 /2002. Handbook of the birds of the world, Vol. 1-7. / Lynx Edicions. Ibisch, P. L. 1996. Neotropische Epiphytendiversita¨t / das Beispiel Bolivien. / Archiv naturwissenschaftlicher Dissertationen 1, Martina Galunder-Verlag, Germany. Isler, M. I. and Isler, P. R. 1999. Tanagers. / Christopher Helm. Jetz, W. and Rahbek, C. 2001. Geometric constraints explain much of the species richness pattern in African birds. / Proc. Nat. Acad. Sci. USA 98: 5661 /5666. Keating, K. A. and Quinn, J. F. 1998. Estimating species richness: the Michaelis-Menten model revisited. / Oikos 81: 411 /416. Kessler, M. 1999. Plants species richness and endemism during natural landslide succession in a perhumid montane forest in the Bolivian Andes. / Ecotropica 5: 123 /136. Kessler, M. 2000a. Altitudinal zonation of Andean cryptogam communities. / J. Biogeogr. 27: 275 /282. Kessler, M. 2000b. Observations on a human-induced fire event at a humid timberline in the Bolivian Andes. / Ecotropica 6: 89 /93. Kessler, M. 2000c. Elevational gradients in species richness and endemism of selected plant groups in the central Bolivian Andes. / Plant Ecol. 149: 181 /193. Kessler, M. 2001a. Pteridophyte species richness in Andean forests in Bolivia. / Biodiv. Conserv. 10: 1473 /1495. Kessler, M. 2001b. Patterns of diversity and range size of selected plant groups along an elevational transect in the Bolivian Andes. / Biodiv. Conserv. 10: 1897 /1920. Kessler, M. and Herzog, S. K. 1998. Conservation status in Bolivia of timberline habitats, elfin forest and their birds. / Cotinga 10: 50 /54. Ko¨hler, J. 2000. Amphibian diversity in Bolivia: a study with special reference to montane forest regions. / Bonner Zool. Monogr. 48.

221

Koleff, P. and Gaston, K. J. 2001. Latitudinal gradients in diversity: real patterns and random models. / Ecography 24: 341 /351. Koleff, P., Lennon, J. J. and Gaston, K. J. 2003. Are there latitudinal gradients in species turnover? / Global Ecol. Biogeogr. 12: 483 /498. Laurie, H and Silander, J. A. 2002. Geometric constraints and spatial pattern of species richness: critique of range-based models. / Div. Distribut. 8: 351 /364. Lees, D. C., Kremen, C. and Andriamampianina, L. 1999. A null model for species richness gradients: bounded range overlap of butterflies and other rainforest endemics in Madagascar. / Biol. J. Linn. Soc. 67: 529 /584. Lieberman, D. et al. 1996. Tropical forest structure and composition on a large-scale altitudinal gradient in Costa Rica. / J. Ecol. 84: 137 /152. Lieth, H. 1975. Modeling the primary productivity of the world. / In: Lieth, H. and Whittaker, R. H. (eds), Primary productivity of the biosphere. Springer, pp. 237 /263. MacKinnon, S. and Phillipps, K. 1993. A field guide to the birds of Borneo, Sumatra, Java and Bali. / Oxford Univ. Press. McCain, C. M. 2003. North American desert rodents: a test of the mid-domain effect in species richness. / J. Mammal. 84: 967 /980. McCain, C. M. 2004. The mid-domain effect applied to elevational gradients: species richness of small mammals in Costa Rica. / J. Biogeogr. 31: 19 /31. Md. Nor, S. 2001. Elevational diversity patterns of small mammals on Mount Kinabalu, Sabah, Malaysia. / Global Ecol. Biogeogr. 10: 41 /62. Mittelbach, G. G. et al. 2001. What is the observed relationship between species richness and productivity? / Ecology 82: 2381 /2396. Montes de Oca, I. 1997. Geografıá y recursos naturales de Bolivia. / EDOBOL, Bolivia. Navarro, G. 1997. Contribucioń a la clasificacioń ecolo´gica y florı´stica de los bosques de Bolivia. / Revista Boliviana de Ecologıá y Conservacioń Ambiental 2: 3 /38. Patterson, B. D. et al. 1998. Contrasting patterns of elevational zonation for birds and mammals in the Andes of southeastern Peru. / J. Biogeogr. 25: 593 /607. Poulsen, B. O. et al. 1997. A rapid assessment of Bolivian and Ecuadorian montane avifaunas using 20-species lists: efficiency, biases and data gathered. / Bird Conserv. Int. 7: 53 / 67. Pulliam, H. R. 1988. Sources, sinks and population regulation. / Am. Nat. 132: 652 /661. Raaijmakers, J. G. W. 1987. Statistical analysis of the MichaelisMenten equation. / Biometrics 43: 793 /803. Rahbek, C. 1995. The elevational gradient of species richness: a uniform pattern? / Ecography 18: 200 /205.

222

Rahbek, C. 1997. The relationship among area, elevation, and regional species richness in neotropical birds. / Am. Nat. 149: 875 /902. Ribera, M. O. et al. 1996. Vegetacio´n de Bolivia. / In: Mihotek, K. (ed.), Comunidades, territorios indı´genas y biodiversidad en Bolivia. CIMAR-UAGRM, Bolivia, pp. 169 /222. Ricklefs, R. E. 2004. A comprehensive framework for global patterns in biodiversity. / Ecol. Lett. 7: 1 /15. Ridgely, R. S. and Tudor, G. 1989. The birds of South America. Vol. 1. / Oxford Univ. Press. Ridgely, R. S. and Tudor, G. 1994. The birds of South America. Vol. 2. / Oxford Univ. Press. Ridgely, R. S. and Greenfield, P. J. 2001. The birds of Ecuador. Status, distribution and taxonomy. / Christopher Helm. Rohde, K. 1992. Latitudinal gradients in species diversity: the search for the primary cause. / Oikos 65: 514 /527. Rosenzweig, M. L. 1968. Net primary productivity of terrestrial communities: prediction from climatic data. / Am. Nat. 102: 67 /74. Sanders, N. J. 2002. Elevational gradients in ant species richness: area, geometry, and Rapoport’s rule. / Ecography 25: 25 /32. Stevens, G. C. 1992. The elevational gradient in altitudinal range: an extension of Rapoport’s latitudinal rule to altitude. / Am. Nat. 140: 893 /911. Stotz, D. F. et al. 1996. Neotropical birds: ecology and conservation. / Univ. of Chicago Press. Terborgh, J. 1977. Bird species diversity on an Andean elevational gradient. / Ecology 58: 1007 /1019. Thornthwaite, C. B. and Mather, J. R. 1957. Instructions and tables for computing potential evapotranspiration and the water balance. / Drexel Inst. of Technology, Laboratory of Climatology, Publ. on Climatology 10: 181 /311. Waide, R. B. et al. 1999. The relationship between productivity and species richness. / Annu. Rev. Ecol. Syst. 30: 257 /300. Whittaker, R. H. 1972. Evolution and measurement of species diversity. / Taxon 21: 213 /251. Wiens, J. A. and Rotenberry, J. T. 1981. Censusing and the evaluation of avian habitat occupancy. / Stud. Avian Biol. 6: 522 /532. Willig, M. R. and Lyons, S. K. 1998. An analytical model of latitudinal gradients of species richness with an empirical test for marsupials and bats in the New World. / Oikos 81: 93 /98. Wilson, M. V. and Shmida, A. 1984. Measuring beta diversity with presence-absence data. / J. Ecol. 72: 1055 /1064. Zapata, F. A., Gaston, K. J. and Chown, S. L. 2003. Middomain models of species richness gradients: assumptions, methods and evidence. / J. Anim. Ecol. 72: 677 /690.

ECOGRAPHY 28:2 (2005)

ECOGRAPHY 26: 411–420, 2003

Area, altitude and aquatic plant diversity J. Iwan Jones, Wei Li and Stephen C. Maberly

Jones, J. I., Li, W. and Maberly, S. C. 2003. Area, altitude and aquatic plant diversity. – Ecography 26: 411– 420. Several explanations have been given for the decline in species richness with altitude. However, separating the influences of altitude, area, and isolation is difficult because of the conical shape of mountains. We used species lists of aquatic plants from \ 300 lakes in a small geographical area to investigate the influence of altitude on species richness. Altitude and/or surface area were better predictors of species richness than any measure of water chemistry. The surface area and depth of individual lakes were not related to altitude, neither was the degree of isolation from other waterbodies. Although species range size increased with altitude, range sizes of all but the rarer species (in the data set) encompassed the lowest altitudes, indicating species loss rather than turnover and no influence of the Rapoport rescue effect. Nevertheless we found a decline in species richness with altitude, additive to the effect of area. Species were ascribed to attribute groups according to an a priori classification based on morphological and life-history traits. The number of attribute groups and number of species within each group increased with area, suggesting both increased diversity and coexistence within niches. With altitude, the number of attribute groups declined, but the number of species per group increased, consistent with a loss of richness and reduced competition. The species remaining at high altitudes were characterised by stress tolerant traits, associated with sites of low productivity. Our results suggest an absolute effect of altitude on species richness, irrespective of other influences and consistent with a decline in productivity. J. I. Jones ( j.i.jones@qmul.ac.uk), School of Biological Sciences, Queen Mary, Uni6. London, London, U.K. E1 4NS. – W. Li, Lab. of Aquatic Plant Biology, Wuhan Inst. of Botany, The Chinese Academy of Sciences, Wuhan, 430074, People’s Republic of China. – S. C. Maberly, CEH Windermere, The Ferry House, Ambleside, Cumbria, U.K. LA22 0LP.

The relationship between area and species richness is probably one of the few general laws of ecology (Lawton 1999), with a weight of evidence and theoretical background supporting it (MacArthur and Wilson 1967, Rosenzweig 1995). In contrast, the decline in species richness with latitude, although widely applicable (Lawton 1999) and long recognised (Wallace 1878), still does not have a universally accepted mechanism at its root (Stevens 1989, Begon et al. 1990, Rohde 1992, Rosenzweig 1995, Willig and Lyons 1998, Chown and Gaston 2000, Lambers et al. 2002). Although studies of the influence of latitude on species richness abound (Rosenzweig 1995, Chown and Gaston 2000), fewer have contemplated the influence of

altitude (Rahbek 1995, Ko¨rner 2000). Nevertheless, a natural comparison exists between latitude and altitude, both representing a gradient of worsening climate. As with latitude, there are, however, several confounding factors. The conical shape of mountains means that the influence of altitude is compounded by one of area. The area contained within altitudinal zones is progressively reduced towards the summit and the species characteristic of higher altitudes influenced by a restricted area of habitat available to them. Studies of variation in species richness with altitude which have acknowledged this problem have sought to overcome it by including the area of arbitrary altitudinal bands (i.e. not necessar-

411

ily corresponding to species distributions, e.g. every 50 m) as a covariable in their models, with some success (Rahbek 1997, Odland and Birks 1999, Grytnes and Vetaas 2002). Nevertheless, due to covariation (the tops of mountains are always smaller than the lower parts), altitude and area are not entirely separable, and these models can only seek to identify any effect of area additional to that of altitude. The potential hard boundaries presented to species distributions at the upper (top of a mountain) or lower (bottom of a valley or sea level) extremes of the altitudinal gradient represent geometric constraints to random distributions of species. Incorporation of such hard boundaries in random simulation models leads to a humped response of species richness to altitude as a null model (Willig and Lyons 1998, Grytnes and Vetaas 2002). This humped response is accentuated if species richness is calculated from regional descriptions of species distributions and species assumed to be present between their highest and lowest sightings (e.g. Stevens 1992, Rahbek 1997, Fleishman et al. 1998). Calculated this way, species richness near the altitudinal extremes consists only of actual sightings, whereas at other altitudes species richness is inflated by interpolations in addition to actual observations (Grytnes and Vetaas 2002). Further complications arise from an increasing degree of isolation with altitude, with populations of those species restricted to mountaintop ‘‘islands’’ separated from one another by an ocean of valleys, which is not true for the converse. The species of lakes suffer from isolation also, surrounded by terrestrial habitats largely inhospitable to aquatic organisms. Lakes are easily separable from the surrounding terrestrial matrix and their area estimated, and because of this discrete nature they have frequently been used to test theories of island biogeography (e.g. Browne 1981, Tonn and Magnuson 1982, Fryer 1985). However, a number of studies have stressed the importance of local environmental conditions in determining species richness and composition in lakes, particularly those related to water chemistry and productivity (Spence 1967, Seddon 1972, Rørslett 1991, Vestergaard and Sand-Jensen 2000, Jeppesen et al. 2000, Heegaard et al. 2001). Here we present a study of species richness of aquatic plants in lakes that cover a range of altitude and surface area, yet within a restricted geographical region. We have used lakes in order to isolate the effects of area, range size and isolation, from altitude, and species richness directly from point samples to avoid those errors that result from the use of distributions interpolated from regional descriptions of species. We have also incorporated the environmental characteristics of each lake (where possible) to include local influences on species richness. 412

Methods Between 1973 and 1980 the amateur naturalist R. Stokoe conducted an extensive survey of aquatic plants in 316 lakes, ponds and tarns in Cumbria. At each water body he compiled a list of plant species present, based on one or several visits. Specialists verified voucher specimens at the time. Upon his untimely death his data were donated to and published by the Freshwater Biological Association (Stokoe 1983), and it is these data which comprise the species lists used here. We used a more strict definition of aquatic plant species than that of Stokoe (1983), who included all plants found in damp and wet habitats surrounding the water bodies, namely those species which are described in the texts of Moore (1986), Cook (1990) or Preston and Croft (1997). We adopted this protocol to provide an objective boundary between aquatic and terrestrial habitats, which in reality is transitional and fluctuates temporally. Applying these criteria reduced the recorded number of species from 233 to 130. Many of those excluded were terrestrial and only occurred at a single site. We disregarded all taxonomy below the species level. Records that were not determined below genus were included where they could be attributed exclusively to an aquatic species. We followed the taxonomy of the above texts. Although the species used in our analysis cover a wide taxonomic range (green algae, bryophytes, pteridophytes and angiosperms), they are all macrophytes (large photosynthetic organisms typically rooted in a permanently submerged substrate) constrained by the selective pressures of living in an aquatic environment. Data describing the environmental conditions in the lakes were derived from various sources (Table 1). Where repeat measurements were available, we used Table 1. Sources used to derive data describing environmental conditions in 316 Cumbrian water bodies. Sources Altitude Area Depth Inflows Outflow Dam Distance to nearest standing waterbody Major ions Alkalinity pH Nitrate Phosphorus Eastings Northings

1, 2, 2, 6, 6, 6, 6 2, 2, 2, 2, 2, 6 6

2 ,6, 7 3, 6 3, 5, 7, 8 7 7 7 8 4, 8 8 8 8

1. Stokoe (1983). 2. Carrick and Suttcliffe (1982). 3. Smyly (1958). 4. Knudson (1954). 5. Talling (1999). 6. Ordnance Survey 1:25 000 maps. 7. The original Stokoe record cards held at the FBA library, The Ferry House, Windermere. 8. unpubl. ECOGRAPHY 26:4 (2003)

means encompassing those most proximal to the date of Stokoe’s survey in preference. Where they were unavailable from other sources we calculated the surface area of water bodies using 1:25000 Ordnance Survey maps. To assess the proximity of the nearest standing water body, we measured the distance from each lake to the nearest standing water shown on 1:25000 Ordnance Survey maps. The most complete sets of data were those describing morphometry and geographical setting. We calculated species altitudinal ranges as the highest and lowest sites where the species occurred, and the mid-point as the average of these values. To determine the effect of temporal incongruence and missing data on the results, we also conducted analyses using only those sites where a complete set of water chemistry data from the 1970s was available, which restricted the number of sites to 67 and species to 86. Initially, we tested for any covariation among the environmental variables using correlation, or logistic regression in the case of presence absence data. Subsequently, we used linear regression with stepwise selection to determine any relationships between species richness (as number of species) and the available environmental variables using SAS (Anon. 1989). To remove heteroscedasticity we log transformed the data where appropriate, after examination of the residuals. For variables without complete sets of data, we repeated analyses excluding the variable to determine the influence of missing data on the results. We investigated the relationship between the occurrence of individual species and environmental conditions using canonical correspondence analysis (CCA), with rare species downweighted using Canoco (ter Braak 1987). Initial tests indicated that the first axis was \ 2 standard deviations in length indicating that unimodal methods were more appropriate than linear ones. In order to determine how different kinds of species responded to changing conditions, we classified plant species, a priori, into the 20 attribute groups (groups of species sharing similar suites of morphological and life history traits) of Willby et al. (2000), together with a further 2 nominal groups, bryophytes and charophytes. We allocated the two hybrids whose parent species spanned attribute groups to the attribute group of the most morphologically similar parent species. For each water body the number of groups, mean number of species per group, and occurrence of groups were used in analyses.

Results The varied geology of Cumbria and the comprehensive survey of Stokoe resulted in the water bodies used ECOGRAPHY 26:4 (2003)

covering a wide range of environmental conditions from coastal lagoons to mountain corries, farm ponds to glacial lakes. Altitude of the water bodies ranged from 2 to 837 m a.s.l., area from 100 m2 to 14.77 km2, alkalinity from − 16 to 4405 meqv L − 1. Naturally there was considerable covariation in the measures of environmental conditions (Fig. 1), with ion poor lakes predominant at higher altitudes and ion rich lakes at the very lowest (caused by the influence of the sea). Many of the ions (except nitrate, sulphate and sodium) were correlated with alkalinity, and each other (Fig. 1). Area was not correlated with altitude (Fig. 2, p = 0.15): although the largest lakes were found at altitudes B 250 m a.s.l., small and moderate sized water bodies were found at all altitudes. Variance in the size of the water bodies was greatest at lower altitudes. Area was correlated with depth (p B 0.0001), presence (p = 0.012) and number of inflows (pB0.0001), presence of an outflow (p B 0.0001) and nitrate concentration (p = 0.024) (Figs 1 and 3). Altitude was not correlated with presence (p =0.47) or number of inflows (p= 0.20), the presence of an outflow (p= 0.68) or the distance to the nearest standing water shown on a 1:25000 map (p = 0.41) (Fig. 3). Of all the variables tested altitude and area were the most significant predictors of species richness, and their effect was additive (no interaction as indicated by GLM). Species richness declined with altitude, and increased with area (Fig. 2). When both these variables were combined the predictive power of the model increased significantly (log species richness + 1=0.1739 0.018(Log area) −0.0010198.8e − 5(altitude) + 0.5489 0.077, F =126.4, p B 0.0001, adj R2 = 0.451), indicating that the distribution of lake area with altitude did not

Fig. 1. Biplot showing correlations between the environmental variables used in the 316 sites. The angle between variables indicates the extent of correlation: acute = correlated, 90°= not correlated, 180°=inversely correlated.

413

sively steeper with increasing numbers of observations, and thus an increasing accuracy with which range size could be estimated. With an increasing number of observations, the slope of this relationship approached 2 (Fig. 5); i.e. where the range of a species could be accurately estimated it included the lowest altitudes. There were many upper elevational limits to species’ distributions but not lower elevational limits. The number of attribute groups per water body increased with area and declined with altitude (Fig. 6a, b), whilst the mean number of species per group increased in both cases (Fig. 6c, d). Other variables were significantly correlated with the number of attribute groups per water body and the mean number of species per group (e.g. number of inflows, alkalinity, chloride concentration) but again these did not have a significant effect when the better predictors, altitude or area, were included in the model. Naturally, attribute groups responded to the environmental variables in a similar fashion to their constituent species (Fig. 7). Notably group 20 (characterised by species with traits of stress tolerance) were associated with higher altitude (Fig. 7), although the constituent species were found throughout the range.

Fig. 2. Relationships between a) altitude and area (p = 0.15), b) species richness and area (F = 71.8, p B0.0001), and c) species richness and altitude (F = 83.97, p B0.0001). Including both altitude and area produced the model, log species richness + 1 = 0.173 9 0.018(Log area) −0.001019 8.8e − 5(altitude) + 0.548 9 0.077, F =126.4, p B0.0001, adj R2 = 0.451.

influence these relationships. Species richness was not correlated with total phosphorus or nitrogen, either independently or in combination with altitude and/or area. Although species richness was also correlated with other variables (e.g. number of inflows, alkalinity, chloride concentration) these variables covaried with altitude or area (Fig. 1) and were less good at predicting richness than either of these factors, singly or in combination. The size of the range of altitudes over which species occurred increased with increasing altitude (Fig. 4, p B 0.0001). However, there did not appear to be any zonation of species (Fig. 5). Although rare species (those occurring in only 1 or 2 sites) were more frequent at low altitudes (Fig. 4), species of high altitudes were found throughout the gradient of altitude (Figs 4 and 5). Estimated altitudinal range size was influenced by the frequency of occurrence of species within the data set (Fig. 4). The slope of the relationship between the size of species’ altitudinal ranges and the altitudes at which their mid-points occurred became progres414

Discussion The influence of water chemistry on the distribution of aquatic plants has long been known (Iversen 1929), with alkalinity and nutrient availability variously described as major predictors of species distributions (Seddon 1972, Kadono 1982, Tiovonen and Huttunen 1995, Vestergaard and Sand-Jensen 2000, Heegaard et al. 2001). As with all correlative studies of species distributions, these descriptions suffer from an inability to separate the influence of naturally associated variables (e.g. calcium and alkalinity) without considerable effort in site selection. Here we used sites from a small but varied geographical area, rich in water bodies, to investigate environmental influences on aquatic plant distribution. Together with many other studies of species distributions this is a retrospective analysis of a large data set collected for other purposes, and fraught with the difficulties of covariation. Although it was not possible to attribute sole influence to many of the water chemistry variables, notably alkalinity and phosphorus, the influence of surface area was separable from these variables and from altitude (Fig. 1). The relationship between species richness and area is well founded in island biogeography (MacArthur and Wilson 1967). Aquatic habitats lend themselves well to such studies because of their easily prescribed boundaries, and several descriptions in support of the species-area relationship have come from such habitats (e.g. Browne 1981, Tonn and Magnuson 1982). Here ECOGRAPHY 26:4 (2003)

Fig. 3. Influence of altitude and area on isolation from other waterbodies, measured as a) presence of at least one inflow (altitude, p =0.47, area, p = 0.012), b) presence of an outflow (altitude, p = 0.68, area, p B0.0001), c) number of inflows (altitude, p =0.20, area, p B 0.0001), d) distance to the nearest standing waterbody visible on a 1:25000 map (altitude, p =0.41, area, p =0.21). ECOGRAPHY 26:4 (2003)

415

Fig. 4. Relationship between the size of the range of altitudes over which species occurred and the altitudes at which their mid-points occurred. Species are categorized according to frequency of occurrence within the data set (number of sites where that species was observed). According to ANCOVA, altitude F1,123 = 760, p B 0.0001, frequency of occurrence F3,123 = 79.5, p B 0.0001, altitude Ă&#x2014; occurrence F3,123 =30.1, p B 0.0001. Influence of occurrence on mid-point of range F3,127 = 8.67, p B 0.0001.

we found an increase in species richness with area, which was not attributable to an increase of elodeid species in alkaline lakes (Vestergaard and Sand-Jensen 2000). The slope of the relationship between species richness and area (z = 0.1739 0.018) was comparable to that found by other workers for aquatic plants (Vestergaard and Sand-Jensen 2000) and other aquatic organisms (Browne 1981). Even though our lakes spanned a range from oligotrophic to hyper-eutrophic (Anon. 1982) neither total phosphorus nor nitrate had any effect on species richness. The number of species did increase logarithmically with alkalinity and several other ions (e.g. chloride, potassium). Nevertheless altitude was a better predictor of species richness than any of these other variables. Over the altitudinal gradient used here, the largest altitudinal range in England, we found a linear decline in species richness. The influence of altitude was additive to that of area (from GLM), with these two variables explaining 45% of the variation in species richness.

Fig. 5. Highest and lowest altitude at which species occurred, ranked according to the highest altitude of occurrence.

416

ECOGRAPHY 26:4 (2003)

Fig. 6. Relationship between number of attribute groups per lake and either a) area (F= 87.85, p B0.0001) or b) altitude (F = 80.5, p B 0.0001), and relationship between mean number of species per attribute group and either c) area (F =19.7, p B 0.0001) or d) altitude (F = 14.2, p = 0.0002). Species were ascribed to attribute groups according to an a priori classification based on morphological and life-history traits (Willby et al. 2000) together with a further 2 nominal groups, bryophytes and charophytes, giving a total of 21 groups.

The likelihood of an inflow or outflow being present, and the number of inflows increased with area, indicating that isolation and area were inversely correlated and that larger lakes were more likely to receive an influx of colonists from other waterbodies. Depth was highly correlated with area also. Zonation of aquatic plants with depth is well known, with species having a characteristic position along this gradient (Spence 1982). An increase in depth will result in an increase in the number of niches available along this gradient, and hence species richness. In addition to using species per se we allocated the species to the attribute groups of Willby et al. (2000) classified a priori according to morphological and reproductive characteristics. The number of attribute groups bore a strong relationship with area, with both the number of groups and the average number of species within each group increasing with area. Species within attribute groups share similar sets of traits and are thus adapted to similar environmental conditions. Thus, an increase in area not only results in more different kinds of species, but more similar species also. If we accept that these attribute groups are likely to correspond with niches (Grime et al. 1988), then an increase in area results in an increase in the number of niches and also the density of species within each niche. ECOGRAPHY 26:4 (2003)

The former could be explained by an increase in habitat diversity (microhabitats) or a more complete exploitation of those available, whilst the latter could be a consequence of increased coexistence within niches (niche packing) or a decrease in niche size. Although the largest lakes were found B 250 m a.s.l., there was no effect of altitude on area (Fig. 2); moderate and small sized water bodies were found at all altitudes. Depth, which has the potential of increasing species coexistence by increasing the number of niches within a water body, was not correlated with altitude either. Typically, the influence of altitude is not separable from that of area. Mountains are conical, and the species of higher altitudes are influenced by a restricted area of habitat. Whereas other workers have dealt with this by incorporating the area of altitudinal zones in their models (Rahbek 1997, Odland and Birks 1999, Grytnes and Vetaas 2002), here we separated this effect of declining habitat area with altitude by using species from a prescribed habitat found at a range of altitudes, and the areas of these patches of habitat. Furthermore, altitudinal zonation of species was not apparent in our data. There were many upper elevational limits to speciesâ&#x20AC;&#x2122; distributions but not lower elevational limits (Figs 4 and 5). As a consequence, those species that 417

Fig. 7. Biplot obtained through canonical correspondence analysis of the distributions of attribute groups and environmental variables. Attribute groups correspond to those described in Willby et al. (2000) where details of the constituent species are given, together with a further 2 nominal groups, bryophytes (group 21) and charophytes (group 22). Group 1 = Alisma plantago-aquatica type, 2 = Glyceria fluitans type, 3 = Nymphaea alba type, 4 =Potamogeton polygonifolius type, 5 =Apium inundatum type, 6 =Persicaria amphibia type, 7 = Callitriche hamulata type, 8 =Ranunculus peltatus type, 9 = Lythrum portula type, 10 = Ranunculus omiophyllus type, 11 = Ranunculus flammula type, 12 =Potamogeton berchtoldii type, 13 = Elodea canadensis type, 14 =Utricularia minor type, 15 = Potamogeton crispus type, 16 =Myriophyllum alterniflorum type, 17 =Zannichellia palustris type, 18 = Lemna minor type, 20 =Juncus bulbosus type. No representative species of group 19 were found.

were present at higher altitudes had a larger total area of habitat available to them than those restricted to lower altitudes, the converse of what is usually found for montane vegetation. Thus, the decline in species richness with altitude is not attributable to a decline in the total area of habitat available, and not caused by differences in the species susceptibility to metapopulation sizes. The distribution of the species with altitude also has consequences with respect to the Rapoport effect (Stevens 1992). (The Rapoport rescue effect maintains that where ranges are narrow there is an increased likelihood of transient populations of species not native to that range being sustained by repeated colonisation, thus inflating species richness.) Although the size of speciesâ&#x20AC;&#x2122; altitudinal range did increase with altitude, species were found throughout the altitudinal gradient with no zonation or turnover of species and hence no potential for the rescue effect inflating the numbers of species at lower altitudes. The Rapoport effect on species richness (Stevens 1989, 1992), therefore, does not 418

contribute to the decline in species with altitude seen in this data set. Unlike other montane vegetation, the degree of isolation was not related to altitude. Neither the presence of an inflow nor an outflow were correlated with altitude, indicating that water bodies at lower altitudes were not more likely to receive a supply of colonists from upstream or downstream (Johansson and Nilsson 1993). The degree of isolation in the varied landscape of Cumbria is more dependent on geographical location than altitude. Measures of isolation had no effect on species richness additional to those of area and altitude, suggesting that metapopulation effects on species richness were not apparent. Although altitude is correlated inversely with human population density and hence with the degree of human influence (as vectors for plant propagules), previous workers have found that population density only influences the number of alien species and not species richness (Roy et al. 1999). In Cumbria, there is the additional complication of tourists who visit the area with the express purpose of climbing to high altitude. Using this methodology to separate the effects of altitude, area and isolation, we found a linear decrease in species richness which can be ascribed solely to increasing altitude, probably due to an indirect effect on temperature and duration of growing season. Mean temperatures decline at a rate of ca 0.65Â°C 100 m â&#x2C6;&#x2019; 1, with a resultant reduction in the length of the growing season. Precipitation and wind speed also increase with altitude (Woodward 1993), but these are unlikely to have a substantial direct effect on the growth of submerged plants. Together with species richness, the number of attribute groups declined with altitude (Fig. 6b); not only were there fewer species but fewer kinds of species, with group 20, the stress tolerators, predominant at higher altitudes. This was not simply a consequence of sampling effort (fewer species =fewer groups) as the mean number of species per group increased with altitude (Fig. 6d). An increase in species per group with altitude implies increased coexistence (within niches) and reduced competition. Such an effect is typical of low productivity environments (Begon et al. 1990), and is consistent with stress tolerant species being found at altitude, competition being less between such species (Grime 1979). The response of the species within this attribute group to other co-varying environmental variables cannot be discounted, many are typical of oligotrophic ion-poor conditions which predominate at high altitude, yet the distribution of these species encompassed all altitudes, including the lowest and most ion-rich conditions. This group of species is also predominant at higher latitudes of both Britain (Preston and Croft 1997) and Finland (Virola et al. 2001), a pattern usually linked with geology, a lower human population density and more oligotrophic conditions (Palmer et al. 1992, Preston and Croft 1997). Although ECOGRAPHY 26:4 (2003)

it is not possible to ascribe these patterns solely to the effect of altitude or latitude, because of colinearity with many variables, the species of this attribute group display many traits associated with low productivity and poor growth conditions (e.g. slow growing, evergreen, low stature) consistent with the short growing season and low temperatures found at high altitude and latitude. Although we cannot comment on the evolutionary history, the species used in our analysis cover a wide taxonomic range (green algae, bryophytes, pteridophytes and angiosperms) of varying antiquity, and the area covered by Stokoe’s survey is one shaped by past glaciations. Furthermore, we have separated the effects of area, isolation and range size from that of altitude, and still found a decline in species richness. This decline is attributable solely to altitude, consistent with an effect of declining productivity. Contrary to Rosenzweig’s hypothesis (Rosenzweig 1995), that the latitudinal gradient of species richness is one of area, a consequence of the tropics occupying a larger contiguous area than other climatic regions, our data indicate an effect of climate on species richness, irrespective of area. Acknowledgements – We dedicate this paper to the late Ralph Stokoe and family, without whose efforts this work would not have been possible. We are indebted to Freshwater Biological Association, for their kind permission to use the data. Thanks also to Nigel Willby for assistance with the attribute groups, and to Robert Ricklefs for his useful comments on the manuscript. J. I. Jones is supported by NERC Fellowship GT5/98/21/CB, and W. Li’s visit to the U.K. funded by the Royal Society and the Chinese Academy of Sciences (the Hundred Talents Program).

References Anon. 1982. Eutrophication of waters: monitoring, assessment and control. – Tech. Rep., Environmental Directorate, Organisation for Economic Co-operation and Development. Anon. 1989. SAS/STAT. – SAS Inst. Begon, M., Harper, J. L. and Townsend, C. R. 1990. Ecology: individuals, populations, and communities. – Blackwell. Browne, R. A. 1981. Lakes as islands: biogeographic distribution, turnover rates, and species composition in the lakes of central New York. – J. Biogeogr. 8: 75 – 83. Carrick, T. R. and Suttcliffe, D. W. 1982. Concentrations of major ions in the lakes and tarns of the Lake District (1953 – 1978). – Occas. Publ. No. 6, Freshwater Biol. Assoc. Chown, S. L. and Gaston, K. J. 2000. Area, cradles and museums: the latitudenal gradient in species richness. – Trends Ecol. Evol. 15: 311 –315. Cook, C. D. K. 1990. Aquatic plant book. – SPB Academic Publishing. Fleishman, E., Austin, G. T. and Weiss, A. D. 1998. An empirical test of Rapoport’s rule: elevational gradients in montane butterfly communities. – Ecology 79: 2482 – 2493. Fryer, G. 1985. Crustacean diversity in relation to the size of water bodies: some facts and problems. – Freshwater Biol. 15: 347 – 361. ECOGRAPHY 26:4 (2003)

Grime, J. P. 1979. Plant strategies and vegetation processes. – Wiley. Grime, J. P., Hodgson, J. G. and Hunt, R. 1988. Comparative plant ecology: a functional approach to common British species. – Unwin Hyman. Grytnes, J. A. and Vetaas, O. R. 2002. Species richness and altitude: a comparison between null models and interpolated plant species richness along the Himalayan altitudinal gradient, Nepal. – Am. Nat. 159: 294 – 304. Heegaard, E. et al. 2001. Species – environmental relationships of aquatic macrophtes in Northern Ireland. – Aquat. Bot. 70: 175 – 223. Iversen, J. 1929. Studien u¨ ber pH-verha¨ ltnisse da¨ nischer gewa¨ sser und ihren einfluss auf die hydrophyten-vegetation. – Bot. Tidsskr. 40: 277 – 333. Jeppesen, E. et al. 2000. Trophic structure, species richness and biodiversity in Danish lakes: changes along a phosphorus gradient. – Freshwater Biol. 45: 201 – 218. Johansson, M. E. and Nilsson, C. 1993. Hydrochary, population-dynamics and distribution of the clonal aquatic plant Ranunculus lingua. – J. Ecol. 81: 81 – 91. Kadono, Y. 1982. Distribution of Japanese Potamogeton. – Bot. Mag. Tokyo 95: 63 – 76. Knudson, B. M. 1954. The ecology of the genus Tabellaria in the English Lake District. – J. Ecol. 42: 345 – 358. Ko¨ rner, C. 2000. Why are there global gradients in species richness? Mountains might hold the answer. – Trends Ecol. Evol. 15: 513 – 514. Lambers, J. H. R., Clark, J. S. and Beckage, B. 2002. Densitydependent mortality and the latitudinal gradient in species diversity. – Nature 417: 732 – 735. Lawton, J. H. 1999. Are there general laws in ecology? – Oikos 84: 177 – 192. MacArthur, R. H. and Wilson, E. O. 1967. The theory of island biogeography. – Princeton Univ. Press. Moore, J. A. 1986. Charophytes of Great Britain and Ireland. – Bot. Soc. of the British Isles. Odland, A. and Birks, H. J. B. 1999. The altitudinal gradient of vascular plant richness in Aurland, western Norway. – Ecography 22: 548 – 566. Palmer, M. A., Bell, S. L. and Butterfield, I. 1992. A botanical classification of standing waters in Britain: applications for conservation and monitoring. – Aquat. Conserv. Mar. Freshwater Ecosyst. 2: 125 – 143. Preston, C. D. and Croft, J. M. 1997. Aquatic plants in Britain and Ireland. – Harley Books. Rahbek, C. 1995. The elevational gradient of species richness: a uniform pattern? – Ecography 18: 200 – 205. Rahbek, C. 1997. The relationship among area, elevation, and regional species richness in neotropical birds. – Am. Nat. 149: 875 – 902. Rohde, K. 1992. Latitiudinal gradients in species diversity: the search for the primary cause. – Oikos 65: 514 – 527. Rørslett, B. 1991. Principal determinants of aquatic macrophyte richness in northern European lakes. – Aquat. Bot. 39: 173 – 193. Rosenzweig, M. L. 1995. Species diversity in space and time. – Cambridge Univ. Press. Roy, D. B., Hill, M. O. and Rothery, P. 1999. Effects of urban land cover on the local species pool in Britain. – Ecography 22: 507 – 515. Seddon, B. 1972. Aquatic macrophytes as limnological indicators. – Freshwater Biol. 2: 107 – 130. Smyly, W. J. P. 1958. The Cladocera and Copepoda (Crustacea) of the tarns of the English Lake District. – J. Anim. Ecol. 27: 87 – 103. Spence, D. H. N. 1967. Factors controlling the distribution of freshwater macrophytes with particular reference to the lochs of Scotland. – J. Ecol. 55: 147 – 170. Spence, D. H. N. 1982. The zonation of plants in freshwater lakes. – Adv. Ecol. Res. 12: 37 – 125. Stevens, G. H. 1989. The latitudinal gradient in geographical range: how so many species coexist in the tropics. – Am. Nat. 133: 240 – 256.

419

Stevens, G. H. 1992. The elevational gradient in altitudinal range: an extension of Rapoport’s latitudinal rule to altitude. – Am. Nat. 140: 893 –911. Stokoe, R. 1983. Aquatic macrophytes in the tarns and lakes of Cumbria. – Occas. Publ. No. 18, Freshwater Biol. Assoc. Talling, J. F. 1999. Some English lakes as diverse and active ecosystems: a factual summary and source book. – Freshwater Biol. Assoc. ter Braak, C. J. F. 1987. Canoco – a FORTRAN program for Canonical Community Ordination by [Partial] [Detrended] [Canonical] Correspondence Analysis, Principal Components Analysis and Redundancy Analysis (ver. 2.1). Tiovonen, H. and Huttunen, P. 1995. Aquatic macrophytes and ecological gradients in 57 small lakes in southern Finland. – Aquat. Bot. 51: 197 –221. Tonn, W. M. and Magnuson, J. J. 1982. Patterns in the species composition and richness of fish assemblages in northern Wisconsin lakes. – Ecology 63: 1149 –1166.

420

Vestergaard, O. and Sand-Jensen, K. 2000. Aquatic macrophyte richness in Danish lakes in relation to alkalinity, transparency, and lake area. – Can. J. Fish. Aquat. Sci. 57: 2022 – 2031. Virola, T. et al. 2001. Geographic patterns of species turnover in aquatic plant communities. – Freshwater Biol. 46: 1471 – 1478. Wallace, A. R. 1878. Tropical nature and other essays. – Macmillan. Willby, N. J., Abernethy, V. J. and Demars, B. O. L. 2000. Attribute-based classification of European hydophytes and its relationship to habitat utilization. – Freshwater Biol. 43: 43 – 74. Willig, M. R. and Lyons, S. K. 1998. An analytical model of latitudinal gradients of species richness with an empirical test for marsupials and bats in the New World. – Oikos 81: 93 – 98. Woodward, F. I. 1993. The lowland-to-upland transition – modelling plant responses to environmental change. – Ecol. Appl. 3: 404 – 408.

ECOGRAPHY 26:4 (2003)

ECOGRAPHY 22: 659-673. Copenhagen 1999

Assemblage structure and quantitative habitat relations of small mammals along an ecological gradient in the Colorado Desert of southern California Douglas A. Kelt

Kelt, D. A. 1999. Assemblage structure and quantitative habitat relations of small mammals along an ecological gradient in the Colorado Desert of southern California. Ecography 22: 659-673. Ecological gradients have intrigued ecologists for many years. In southern California the Deep Canyon Transect spans a range of habitats and elevations from Lower Sonoran Desert sand dunes and creosote scrub to Upper Transition coniferous forest, where relict species typical of the Sierra Nevada are found. I sampled a 1050 nci elevational range in this transect to evaluate the ecological distributions of small mammals and to better characterize community structure. Results complement and substantially extend a previous study of this fauna, and provide insights into the habitat associations of species in this complex fauna. Assemblage structure changed greatly between summer and winter, largely due to reduced presence of pocket mice Chuetodipus in winter. Additionally, the distribution of abundanee and species richness was different than reported earlier, suggesting that patterns across this gradient may be temporally variable, and strongly influenced by local dynamics. Most taxa exhibited significantly nonrandom use of a large number of habitat variables but this was not a simple consequence of the elevational gradient. A mid-elevation bulge in species richness was indicated, but likely is not a consequence of mass etTects since a number of animals captured in intermediate regions were reproductively active. D. A. Kelt (dakc'lt{iv,u(dciiis.i'du). Dcpl oJ Wildlijc. Fish, and Consenaiion Univ. of Ccilijornici. Dini.s, CA 95616, USA.

Many species reach their ecological litnits across ecological gradiertts. where they come into contact wit:h taxa from separate biotas. Elevational gradients a:re particularly interesting as these often are rooted at either end by species adapted to very dilTerent environmental conditions (e.g. thermal or precipitation regimes). As a result, numerous studies have been conducted in recent years on patterns of distribution and abundance of various taxa across elevational gradients (e,g,, Whitaker and Neiring 1965. Kikkawa and Williams 1971. Terborgh 1977, Patterson et al, 1989, 1990, Rickart et al, 1991, Kelt et al, 1999), In a recent review of species richness across elevational gradients. Rahbek (1995) concluded that species richness tends to decline with elevation, but often not monotonically.

Biology,

Interestingly, almost half of the studies reviewed (including 36 of 73 studies in tropical regions and eight of 17 studies in non-tropical regions) demonstrated a hump-shaped pattern, with greater species richness at intermediate elevations than at sites located at either extreme. Hypotheses to explain such elevated richness within ecotonal regions generally invoke increased resource availability at intertnediate regions or the overlap of species from faunas at either end of the transition. If environmental conditions are more favorable within ecological transitions then elevated abundance could refiect a higher carrying capacity. Increased habitat heterogeneity should favor elevated species richness, while not necessarily favoring greater abundance. Shmida and Wilson (1985) argued that in

659

some cases intermediate sites are populated by individuals that have emigrated from more optimal habitats. Such "mass effects" would influence richness more than abundance, and have been demonstrated in some systems (see Kunin 1998), The scale of observation may also influence interpretations. Across a substantial ecological transition in southern South America, several authors noted a hump-shaped distribution of species richness for birds (Ralph 1985) and small mammals (Pearson and Pearson 1982), However, closer inspection of the mammal fauna there indicated that the increased richness was caused by increased beta diversity between habitats at intermediate sites, but that at any point across this transition had about equal numbers of species (Kelt 1996). Thus, there is reason to remain skeptical about such generalities until further studies evaluate patterns

at multiple spatial and temporal scales (see Wiens 1989, Levin 1992, Karieva 1994), In southern California the Colorado Desert yields to the Peninsular mountain ranges across a relatively short distance (Burk 1988, Thorne 1988; Fig, 1). In a span of < 15 km the biota changes substantially; faunal and floral elements of the Colorado Desert yield to elements of the San Diegan District, and near the crests of the Peninsular Ranges a subset of Sierra Nevadan elements are found amongst coniferous forest (Fig. 2), Faunal lists taken from Ryan's (1968) study on the mammalian fauna of this transition (Fig. 3) suggest that species richness of both rodent species and all terrestrial mammals are distributed in a bimodal fashion across this elevational gradient. Richness was relatively low in the sand dunes of the central Coachella Valley, increased in the Deep Canyon floodplain, and declined

M-^ 2,000â&#x20AC;&#x201D;3,000 m D 500â&#x20AC;&#x201D;1,000 m ! I Sea levelâ&#x20AC;&#x201D;200 tn below sea level I

Fig. 1. Map of southern California. The Deep Canyoti Transect is indicated by the rectangle that extends froin the western fringe of the Colorado Desert into the Santa Rosa Mountains, part of the Peninsular Ranges that extend south into Baja California, Mexico.

660

ECOGRAPHY 22:6 (t999)

Fig. 2. Idealized cross section of the Deep Canyon Transect, showing the principal habitats encountered. The present study spans the region from Creosote-Palo Verde to Pifion-Juniper habitats. Figure modified slightly from Ryan (1968) with permission of The Palm Springs Desert Museum.

-2500 m Transition Life Zone Coniferous Forest

Chaparral

Upper Sonoran Life Zone

Pihon-Jurtiper

-1000 Lower Sonoran Life Zone

Agave-Oootillo

Rooky Slopes Cholla-Palo Verde

-500

Creosote-Palo Verde Mesquite Sand Dunes 8, Creosote

-'Si'^ail^lil^^.

again in non-floodplain valley floor habitats (Ryan's "cholla/palo verde" habitat). Richness then increased on rocky slopes, had a primary peak in agave/ocotillo habitat (which overlooks the rocky slopes), and declined through pifion-juniper, chaparral, and coniferous forests. Interestingly, however, rodent density (number/acre) was unimodally distributed (Fig, 3), declining both above and below the rocky slopes habitat; although the rocky slopes had only moderate species richness, this habitat had greater abundances than all other habitats sampled. In fact none of these relationships differ from a uniform distribution across elevation (linear regression, all p > 0.40, r-<0,12; no improvement in fit was obtained with a polynomial regression). However, these nascent patterns occur at or near the interface between two biotas, begging questions of local community structure and assembly across ecological frontiers. At intermediate sites along this gradient we might expect changes in species composition and relative abundances, atid consequent shifts in foraging strategies, spatial activity patterns, etc. In order to better understand the factors underlying the observed distributions of the matnmal fauna across an ecological transition. I surveyed small matrtmal communities at ten sites encompassing all major habitats across 1050 m of the Deep Canyon transect, spanning the peaks in both richness and abundance noted by Ryan (1968), Foraging and spatial ecology are not addressed in this report. Rather, I censused ECOGRAPHY 22.6 (19991

small mammal communities with live traps and markrecapture methods, and recorded extensive habitat metrics during both the summer (1996) and winter (1996/97) to evaluate elevational and seasonal shifts in richness, abundance, and habitat use.

Materials and methods Mammals Small mammals were censused with folding Sherman live traps placed singly in 7 x 7 trapping grids (15 m

Fig. 3. Mammal speeies richness (symbols) and total density (bars) reeorded by Ryan (1968) aeross the Deep Canyon elevational transeet. Labels on the abseissa correspond to habitat types used by Ryan. Numbers at the base of vertical bars give the area sampled for density estimates.

661

spacing). Assuming a boundary region of \ inter-trap distance these grids measured ca 1.1 ha in area. Traps were baited with millet seed and were opened in the early evening and checked before dawn. Animals were individually marked with numbered ear tags or by clipping fur, and species, sex, age (based on pelage), reproductive condition, and weight were recorded before releasing the animals at their point of capture. Grids were run for three consecutive nights in summer (June and July) 1996, and winter (December and January) 1996/97,

Habitats sampled Sampling effort was divided among the valley floor (two grids), rocky slopes (two grids), agave/ocotillo (two grids), and both lower and upper pifion-juniper (two grids each) habitats. The lowest end of the Deep Canyon Transect lies within the Coachella Valley, where agricultural and urban development precluded meaningful sampling. Because of a combination of fire hazard and difficulty of access, 1 was unable to sample chaparral and conifer habitats which occur at higher elevations. Brief descriptions of the sites studied follow.

very hot site, and is dominated by barrel cactus and small numbers of creosote. Many brittlebush are here as well, although these are deciduous and in the summer form mere skeletons of plants. Station site (T6S, R5E, SW I Sec. 17320 m) is located on a north facing slope with rocky soils. The dominant plants there are creosote and brittlebush, with scattered agave Agave descrii and ocotillo Fouqueria splendens.

Agave/Ocotillo Two sites (Agave Hill east, T6S, R6E, NW \ Sec. 19: Agave Hill west, T6S, R5E, NW 5 Sec, 19; both sites at 745 m) in this habitat type were located only several hundred meters apart, immediately above the steep walls that form the boundaries of Deep Canyon. These sites are dominated by ocotillo Fouqueria splendens, brittlebush, goldenhead Vigiiieria deltoidea, four o'clock Mirabiiis tenuiloba, galleta grass Hillaria rigida. and clumps of agave. Additionally, there are some barrel cacti as well as jumping O. bigelovii and deerhorn chollas.

Lower Pifion-Juniper (Lower plateau of Zabriskie 1979) Valley Floor - Cholla/Palo Verde habitat Two sites were sampled. These are not intended to be true replicates, as they were dominated by very different plant species, and had very different soils, Palo Verde site (T6S, R5E, NW 5 Sec, 16255 m) was located in a sandy wash dominated by palo verde Cercicliiim fiorideum. cat-claw acacia Acacia greggii, and small amounts of mesquite Prosopsis glandulosa, smokethorn Dalea spinosa, and creosote Larrea tridentata. Small numbers of barrel cactus Ferocactus acanthodes and deerhorn cholla Opwitia acanthocarpa round out the most important plant species found here, Cholla site (T6S, R5E, NW | Sec. 16, 260m) was dominated by a variety of plants including creosote Larrea tridentata and brittlebush Eiuelia farino.sa, indigo bush Dalea schottii, burrobosh Ambrosia dumosa, jojoba Simmondsia chinensis, and beavertail O. basilaris, pencil O. ramasissima, deerhorn O. acanthodes, and golden O. echinocarpa chollas.

Rocky slopes As above, the two sites on rocky slopes were designed to yield information on the diversity of habitats occurring in the rocky slope of Deep Canyon, and are not intended to serve as statistical replicates, Chuckwalla Hill (T6S, R5E, NW \ Sec. 17320 m) is a south facing slope covered with large boulders. As a result it is a 662

Two plots here were dominated by scattered piiion pine Pinus monophylta and California juniper Jimipcrus californicus. Other common plants included sugar bush Rhus ovata, scrub oak Quercus turbinata. crucillo Ziziphus parryi. western bernardia Bernardia incana, desert apricot Primus fremontii. snakeweed Guttierczia sarothrae. Virgin river encelia Fncelia virginen.sis. and buckwheat Eriogonum fasciculatum. Piiion Crest north (T6S, R5E, SW 5 Sec, 26, 1250 m) was located about 200 m from Piiion Crest south (T6S, R5E, SW | Sec. 26, 1250 m).

Upper Pinon-Juniper habitat (Upper plateau of Zabriskie 1979) One site (Observatory site, T6S, R5E, NW \ Sec. 34, 1280 m) was located near the Cecil and Ida Green Piiion Elat Observatory, operated by Univ, of California, San Diego. Dominant plants here were piiion pine and scrub oak, with good representation by California juniper, manzanita Arctostaphylos glauca. sugar bush, antelope bush Purshia tridentata, rabbit bush Crysothamnus teretifolius, and brickelia Brickelia oblongifolia. Other notable plants include golden cholla, pancake cactus O. chlorotica, prickly pear O. phaeacantha. and yuccas Nolina parryi and A', wotfi. The second site (Microwave site, T6S, R5E, SE 5 Sec, 34, 1295 m) was located near a microwave tower at the tXOGRAI'llY 22;(i (IW9)

Pinon Flat Observatory, Species composition here was similar to that at the previous site, with the addition of ribbonwood Adenostoma sparsifoliunu deerhorn cholla, and more grass cover under a slighly denser tree canopy.

Habitat metrics Thirty-one variables, reflecting the general structure and diversity of eaeh trapping point (Table 1), were recorded at each trapping point during both trapping sessions. Variables were evaluated for normality and transformed when possible. Slope, Aspect, # shrub sp. Shrub ht, and Cv-ground did not require transformation, D-tree was arcsin transformed, and # cholla, # holes, D-cholla, D-agave, D-shrub, Soil hardness, Cv-dead shrub, Cv-live shrub, Cv-herb, Cv-rock, Cvgrass, and Shrub area were log transformed. Other variables could not be normalized; these were used for analysis of habitat selectivities, which does not require normally distributed variables, but they were excluded from analyses requiring normally distributed data.

Analyses Local species richness (S) and densities (N = number of animals captured) were evaluated for patterns across elevation. Species diversity was calculated using the Shannon-Weiner index, H' = — Z pi In Pj, where P| is the proportion of the community comprised by species i. Habitat variables were entered into a principal components analysis to reduce the dimensionality of the data set and to increase the sample size/variable ratio. Principal components axes with eigenvectors greater than unity were retained for subsequent analyses. Mammal species distributions were analyzed in PC space by means of multivariate analysis of variance, and two a posteriori tests (Student-Newman-Kuels test and Scheffe's test) were applied to evaluate species differentiation on these axes. Habitat selectivities provide a distribution-free means of evaluating the extent to which species use nonrandom subsets of the available environment (see Patterson et al, 1990, Kelt et al, 1994, 1999), Because the absolute deviation from the available background is highly sensitive to sample size, a randomization routine was used to evaluate the significance of observed departures from random, using a two-tailed significance criterion of 5"A> (2,5% in each tail: see Kelt et al, 1994 for details of the randomization routine).

Results A total of 495 individuals of 11 species were captured ECOGRAPHY 22:6 119991

Table I. Habitat metrics recorded at Deep Canyon. Variables preceded by asterisks could be normalized, and these were used in the principal components analysis. Direct count measurements: Traploc

Trap location: arboreal, by trunk, litter, among herbs, by log, under shrub. * Aspect Coded numerically by cardinal compass points: 1 = N, 2 = NE, etc. *Slope Coded as 1 = < 5°, 2 = 5-20°, 3 = 2045°, 4 = >45°. * # shrub sp Number of shrub species within 3 m of trap marker. #herb sp Number of herbaceous species within 3 m of trap marker. #cacti Number of barrel cacti within 3 m of trap marker. * # cholla Number of cholla cacti within 3 m of trap marker. #agave Number of agave plants within 3 m of trap marker. * # hole Number of obvious burrows within 3 m of trap marker. #tree Number of dead trees within 3 m of trap marker. #sm log Number of small logs (7.5-15 cm diameter) within 3 m of trap marker. #lg log Number of large logs ( > 15 cm diameter) within 3 m of trap marker. Herb ht Mean height of herbaceous vegetation. *Shrub ht Mean height of shrubs. *D-cholla Distance to nearest cholla cactus. *D-agavc Distance to nearest agave. *D-shrub Distance to nearest shrub. Lxw-shrub Length x Width of nearest shrub. (Shrub area was calculated as pi*(0.5*LenShrub*Wid-Shrub). * D-tree Distance to nearest tree. Dbh Diameter at breast height of nearest tree. Litter depth Litter depth in each of four quadrats delineated by the line transects. *Soil hardness Soil hardness in each of four quadrats, using soil penetrometer. Measurements recorded along two 2-m transects (oriented N S & E W): *Cv-live shrub Cover of live shrubs, *Cv-dead shrub Cover of dead shrubs. *Cv-ground Cover by bare ground. *Cv-herb Cover by herbaceous plants. *Cv-rock Cover by rock. *Cv-grass Cover by grass. Cv-moss Cover by moss. Cv-log Cover by logs. Cv-water Cover by water. Estimated by eye: Canopy cover Canopy cover.

(overall \l.l"/u trap success; Table 2), including 308 captures of ten species in the summer (21'%. trap success), and 187 captures of nine species in the winter (14'Xi trap success). These included six species of heteromyid rodents {Cluwtodipus falki.x Merriam, 1889, C, foriiiosus Merriam, 1889, C. spinatus Merriam, 1889, Perognathus kmgimembris (Coues, 1875), Dipodomys agilis Gambel, 1848, D. merriami Mearns, 1890) and tive species of sigmodontine rodents (Neotoma lepida 663

QJ 3^

3 O n " . ' — ^ O ' ^ 0 » 0 ' — ' y^t

___^______O

px'

—— O O

(NOOr'-iOO

O -

Tt O O r^ - ^ - ^ O O ^+ i/~i './^ ^^^ ^^^ '^^ '•^^ ^^^ o^j f ^ l

^"^ f "^ t ' 1 o ^ f' ^ < ;

^^^ V

OO'H S .E

o . bJ: ; — OJ O O

'E-a-5 o c i,

a. '^ -~ O ^ K .tt.C^i,.

— in — •

^ 0-

oooooooooo —

oooooor^oo ; =

j OOOO

OrsOOOO

8 =o

^2d

O O O O O O O — OO

ooooooooo

= 00000000

o o o o o o — r-oo

1^,11 u '-'

—

OOOt

OJ ' — . :

—

\O — O O O O O O

^ DO — ; O O O O

OOOOOOO

1 ^ 1 •= 't O O—O O O O - ^ °

— OO

n-, O O O O (^^ O O ^

: - ?: D.

S S § ? CJ OJ CO JZ nj QJ o id

r" ioi

r-j tu

yon.

•evia

o a

X) QJ aj

5 (^ iJ QJ OJ O-

i.l

X) d

664

ECOGRAPHY 22:6 (1999)

Cholla/ Palo Vertie

Rocky Slopes

Agave/ Ocotillo

Lower P^

Upper P-J

X X ^

X^-:X

\ —

X ^ —^

Cholia/ Paio Verde

Rocky Slopes

Agave/ Ocotillo

Lower V~i

X 30

Fig. 4. Numbers of rodent species captured at Deep Canyon, California, in summer 1996 and winter 1996/97.

/ ( \

~'~~T \

20 -

\ ^ \

15 -

Distribution of species and abundances Species richness across the gradient appeared unimodal in winter (Fig. 5a), but was equally well described by linear and polynomial regression against elevation (Table 2), Summer data also appeared mildly unimodal, but neither regression models yielded significant results (Table 2), When summer and winter data were pooled, however, only a polynomial model approached significance (linear, F = 3,09, p > 0,10, r- = 0,30; polynomial, F = 3,78, 0.05 < p < 0,10, r^ = 0,35), Any incipient peak in richness was largely due to the very low richness in upper piiion-juniper habitat (Table 2). The distribution of population abundances differed greatly between seasons (Fig, 5b), As with species richness, the evidence for a peak at middle elevations was ambiguous. Winter data regressed signifieantly against elevation with both linear and polynomial models (Table 2), In contrast, summer data were poorly fit using a linear model, but approached significance (p < 0,10) with a polynomial model. Summer abundances E C O G R A P H Y 22:6 (1999)

10 5 0

Choiiay Paio Verde

Rocky Siopes

Agave/ Ocotiiio

Lower P^

Upper P-J

pecies Diversity (H'|

Thomas, 1893, Feromyscus boylii (Baird, 1855), P. crinitus (Merriam, 1891), P. eremicus (Baird, 1858), and P. truei (Shufeldt, 1885)), The most abundant species was C, falkix, followed by C. formosus and C spinatus (Fig, 4). Together, these three species constituted just over half of all individuals captured, although their dominance was highly seasonal ((yl"/n in summer vs 'ilVu in winter), reflecting lowered levels of activity during the cooler winter months. However, only two species (Peromyscus eremicus and Neotoma lepida) were more abundant in winter than in summer, suggesting that spring/early summer recruitment may have played a role in the numerical differences observed across these seasons. Two species (P. longimembris and P. boylii) were trapped only in summer, and one species (P. iruei) was captured only in winter.

Upper P,J

—T - Summer -&, Winter

00 %

%„

*o \

Fig. 5. Rodent species richness and density (number captured per census using standardized trapping methods) in summer 1996 and winter 1996/97 across the Deep Canyon elevational transect. Labels on the abscissa are Grid names, while boxes at the top of each panel give the corresponding habitat types used by Ryan. The "X" symbols in the upper panel reflect the total number of species captured across both seasons.

were relatively high throughout much of the transect, with a possible peak in the agave/ocotillo habitat and rapid decline at higher sites. Winter abundances were unimodally distributed with a peak in the rocky slopes. The seasonal difference appears to reflect greatly reduced abundances at most sites in the winter - only sites in the rocky slopes and the upper pifion-juniper habitat (with very low numbers) remained similar in both seasons. The distribution of abundances paralleled that for species richness in winter but not in summer. Abundances were greatest in the agave/ocotillo habitat during the summer, but during the lower overall abundances of winter the peak shifted to the rocky slopes. The seasonal shift in abundances occurred across most species, but was particularly strong for the pocket mice 665

Table 3. Results of regression between elevation and both species and PC axes. Degrees of freedom are 1,8 for summer and 1,7 for winter. E-statistics in bold font indicate analyses that were significant after sequential Bonferoni adjustment. Species

Linear F

C. formosu.s C. fallax C. .spinatu.s P. longimenibris D. njerrianii D. agili.s P. boylii P. eriniius P. ereniieii.s P. iruci N. lepida

47.620 0.425 8.222 1.774 1.747 9.294 1.085 4.204 0.709 1.085 0.176 Principal component axes PC I 28.652 1.059 PC 11 2,030 P C III 5.844 PC IV PC V 0.005 0.594 PC VI PC VII 0.291

2 -

Polynomial ]â&#x20AC;˘-

0.737 0.024 0.326 0.094 0.093 0.354 0.060 0.198 0.040 0.060 0.010

33,209 0.171 7.849 1,874 1.154 11.196 1.144 6.055 1.101 1.144 0.645

0.661 0.010 0.316 0.099 0.064 0.397 0.063 0.263 0.061 0.063 0.037

0.628 0.059 0.107 0.256 0.0003 0.034 0.017

35.217 0.349 3.244 4.600 0.053 0.290 0.198

0.674 0.020 0.160 0.213 0,003 0.017 0,012

(Fig. 4), which ai-e much less active iti wititer, perhaps reflecting seasotial hibertnation or use of torpor. With equal trapping effort, nurnbers of the three principal pocket mice (C. formosus, C. fallax, and C. .spinatu.s) declined by 35-79'%i, Dipodomys agilis and P. criuitus also declined about 3()'yii, and Ncotonui increased by Ab%, but these species only occurred at relatively low numbers such that these proportional changes did not greatly influence comtnunity-wide patterns. Species diversity (H') also was best described with a polynotnial regression for both seasons, but was only marginally significant in summer (Table 2, Fig, 5c). Somewhat unexpectedly, only one speeies (C. fonno.sii.s)

Agave HillE

Agave HillW

Observatory

PCI

Fig. 6. Distribution of sampling grids in the first two principal component axes using summer habitat metrics. Sites were distributed very similarly in winter except that Palo Verde was not sampled. regressed significantly against elevation (Table 3). Three other species (C. spinatu.s, D. agilis, P. crinitiis) regressed significantly before critical values were Bonlerroni adjusted. Low satnple sizes tnay have limited statistical power on these analyses.

Habitat tnetrics Principal components analysis yielded five informative tnetries (A > 1.0; Table 4), The first PC axis (PC I) appears lo be largely an axis of ground cover and soil hardness, being strongly influenced by ground cover variables, with eover by rocks loading positively and that by open ground loading negatively. Soil hardness also loaded positively, as did slope and distance to the nearest tree. PC II refleeted shrub eover and diversity.

Table 4. Principal eigenvectors from a Principal Components Analysis on vegetative and habitat metrics collected at Deep Canvon in summer 1996 and winter 1996 97.

Slope Aspect Shrub sp #chola #hole Shrub ht D-cholla D-agave D-shrub D-tree Soil hardness Cv-dead shrub Cv-live shrub Cv-ground Cv-herb Cv-rock Cv-grass Shrub area

666

PRINl 4.43

PRIN2 2.51

PR1N3 1.75

PRIN4 1.47

PRIN5 1.02

0.389891 -0.088027 0.053391 -0.077257 0.294069 -0.134137 0.187678 0.174159 -0.127821 0.325383 0.415829 -0.000580 0.050200 -0.313307 -0.170808 0.421107 -0.212715 -0.117608

0.084283 -0.212492 0.424451 0.298966 -0.103130 0.009448 - 0 ^75^50 -0.282289 -0.388254 0.109233 0.015029 0.301708 0.395211 -0.063070 0.191485 0.048808 0.161103 -0.053293

0.071150 0.057760 0.194242 -0.376089 0.224001 0.424542 0.360743 0.118710 -0.177537 -0.156708 -0.109747 0.054214 0.314705 -0.142655 0.017972 -0.115740 0.198825 0.440554

-0.159784 -0,270308 0.031630 0.259086 -0.114017 0.399275 -0.241807 0.355634 0.080798 0.273086 0.008148 -0.075169 0.068950 0.227589 -0.276142 0.110567 -0,306130 0.387743

0.105393 -0.351383 -0.109495 0.000629 -0.006937 0.228271 -0.161089 -0.194626 0.477208 0.012904 0.152975 0.164781 -0.308004 -0.384939 0.147318 0.112811 0.369024 0.212949

tiCOGRAPHY 22:6 (1999)

with number of shrub species and cover by both live and dead shrubs loading positively and distance to cholla and to shrubs loading negatively. PC III loaded positively with shrub area and shrub height, as well as with distance to cholla and with live shrub cover, and negatively with cholla cover. Thus, this axis also appears dominated by shrub structural elements, PC IV appeared to reflect a shrub/grass gradient, being strongly influenced by shrub height and area, distance to agave (all positive), and ground cover by grass (negative). F'C V also was influenced by shrubs and ground cover, with strong loadings by distance to shrubs, ground cover by grass (positive loadings), and by bare ground cover, aspect, and live shrub cover (negative). Because interpretation of these patterns rapidly became difficult, discussion is generally limited to the first two PC axes; these axes described just under 40'^;. of the total variance in the data set. 3 , PEBO

PEEI

2 . 1 .

PGLO

1 O

CHFA

i 1 NELE

PECR CHSP —1

-1 -

1 1 DIAG

1> DIME

Total CHFO

-2 •

-3 .

PCI

2 • PEER 1 = 1

CHFA 1 1 (> PETR

O Q.

1 Total

NELE •1 PECR

0 DIME

-in

CHFO

CHSP

-1 DIAG -2 -

-3 .

Trapping sites were arrayed in a counter-clockwise pattern in PC space (Fig, 6), indicating a non-linear but readily interpretable distribution of data in PC space. As a result of this orientation, some sites clustered together that were not located near to each other along the transect, reflecting similar habitat characteristics at spatially disjunct parts of the transect, Chuckwalla Hill, from the rocky slopes, was characterized as extremely rocky and barren of much vegetation, and this site polarized the first PC axis. This site was most closely allied with the Station plot (also from the rocky slopes), which was similar in having a steep aspect and limited vegetation. Two sites from the valley floor (Palo Verde and Cholla) grouped more closely to the highest elevation sites (Observatory and Microwave) than to other low elevation sites (Station and Chuckwalla). This likely reflected the relatively soft soils and the large percentage of cover by open ground at these sites. Additionally, these sites all had a fairly limited diversity of shrub species which tended to be located near to each other. When small mammal captures are superimposed onto principal components space the apparent incongruities in the distribution of sites become clearer, as species are arrayed in an ecologically reasonable sequence. It also is clear that species overlap extensively in PC space (Fig, 7), although segregation is apparent on the first four PC axes (ANOVA, p < 0.0001 for axes 1, 2, and 4, p = 0,0167 for axis 3). A posteriori tests on each PC axis demonstrated habitat segregation by species (Table 5), On axis I, C. spinatus. P. criniius. and A', lepida are distinct from D. agilis, P. triiei, P. boytii, and P. tonginienibris. Varying levels of similarities are exhibited by the remaining species. The clarity of these distinctions diminished on subsequent axes, although segregation of PC space remains apparent at a fairly coarse level. There was a general congruence in species placement in PC space between summer and winter data sets (Fig, 8). Nonetheless, on the first three PC axes, Neotoma, P. criniius. and C. .y^natus converged in the winter, whereas C. Jatla.x and P. eremicus diverged on axes I and II. When viewed in PC Il/III space, D. nierriami and C. fonnosus converged in winter, as did P. crinitus and

Neotonia. Finally, only the first principal component axis regressed significantly with elevation (Table 3), suggesting that it captured elements of environmental variation that reflected changes across the gradient. All other principal component axes were unrelated to elevation.

PCI Fig. 7. Distribution of rodent species in the first two principal component axes, using summer a) and winter b) data. Abbreviations as in Table 2: CHFA, C/kicto(lipiis fiillax: CHFO, C. formosus; CHSP, C. spiiuitu.s- DIAG, Dipoc/onns agiii.v. DIME, D. luciriaiui: NELF, Neotoiiui Icplila: PFBO, Pciomysciis hoyiii: PECR, P. ainiius: PEER, P. cremiciis: PETR, P. tnici: PGLO, Pcrogiuitiiiis longimembris. Total, all species combined. ECOGR.APHY 22:6 (1999)

Habitat selectivities Most taxa were non-randomly distributed on various habitat axes (Fig. 9). Additionally, the number of significant deviations was positively related to number of captures for summer (y = 0.1065x + 11,32, r - = 0,42, p < 0.025, N = 308 captures of 10 species) but not winter 667

Table 5, Results of a posteriori tests of the distribution of species on the ftrst four priticipal comportent axes. Species are arranged vertically, and species sharing letters are not significantly different on a given PC axis. Results for both SNK and Scheffes tests are presented.

CHFA CHFO CHSP DIAG DIME NELE PEBO PECR PEER PGEO PETR

Prin4 1,47

Prin3 1,75

Prin2 2,51

Prinl 4,43 SNK

Scheffe

SNK,

Seheffe

SNK

Scheffe

SNK

Scheffe

bed ab a d cd a d a abc d d

abed abc a d bed ab cd a abed ed d

ab b b b b b a b ab ab ab

ab ab ab b b ab a ab ab ab ab

a a a a

ab ab ab b a ab ab ab ab ab ab

a a a a a a

(p>0,10, N = 187 captures of 9 species): the greater number of captures in sumtrter likely provided for greater resolution than during the winter. Habitat selectivities generally agreed with the limited and generally anecdotal information published on habitat use by these speeies. Perhaps the most useful means of viewing these data is to evaluate divergence among functionally similar speeies.

a a a

a a a a a a

ceous growth in the summer, but with taller shrubs and low cover by logs and herbs in the winter, Chaetodipus fallax was found at sites with low grass cover and close proximity to oeotillo in the summer but taller grasses and greater distanees to oeotillo in 'winter. Finally, sites

Pocket mice All four pocket miee exhibit a large number of significant deviations from the available habitat. Three Chaetodipus were found at sites with more cacti, lower eanopy cover, less grass and shrub cover, greater distance to trees, and shallower litter than expected by chance, Perognathus differed frotn Chaetodipus in selecting sites with greater eanopy cover than random, whereas Chaetodipus species selected sites with significantly less eatiopy cover than random. Additionally. Perognathus was found much closer to trees, at sites with greater litter cover and less shrub cover, and with fewer cacti and more grass, than the three Chaeotodipus. Perognathus and C, falla.x shared some selectivities. such as looser soils, cover by live shrubs, greater proximity to both chollas and agaves, distance to shrubs, and ground cover by roek, Chaetodipus spinatus was found further than expected from shrubs and agaves, and at sites with tall herbs and little cover by logs, Chaetodipus formosus was found with less live shrub cover than other pocket mice in the summer. The most abundant rodent at Deep Canyon. C. falla.x. evidently favored sites with more ground cover of looser soils, but also more herbaceous cover, and less open rock, greater cover by shrubs, fewer agaves and cholla. and fewer apparent holes, than other pocket mice. There were several interesting seasonal shifts in habitat use (Fig, 9), Chaetodipus formo.ms seleeted sites with low shrub height and high cover by logs and herba668

o Q.

PCI PETR o

o \

PGLO \

. PEBO

\pEER

DIAG

\ CHFA

P PECR

Totari \

DIME/

CHSP

CHFO ^NELE

PCI Fig, 8, Shifts in centroid positions in principal components spaee between summer (filled symbols) and winter (open sytnbols). Abbreviations as in Fig, 6, ECOGRAPHY 22;6 (t999)

SUMMER

WINTER

C, formosus N = so, 33

C. faltax N = 101, 24

C. spinatus N = 41, 20 ']

*l^

longimembris mbris U •—j» 1 __• «__,»_ t^=9,o \ B ^ ^ T o i P * " " ^ " ' ' ' - ^ — ^ ^ ~

* * *

* * * *

D. agilis N = 16, 16

D. merriami N = 15, 13

P. boytii N=1,0

P. crinitis N = 16, 18

P. eremicus N = 12, 43

p. truei N =

0,3

Neotoma lepida N = 17, 17

Fig, 9, Habitat selectivities of small mammals across the Deep Canyon elevational transect in summer 1996 and winter 1996,97. Use of measured habitat characteristics are presented as ratios of mean use to the mean value of these metrics across all stations. Asterisks indicate variables for which species differ significantly from the available range of a given characteristic. Ticks on the ordinate axes represent unit deviation. Habitat characteristics are labeled on the ordinate.

with C spinatus were characterized as havitig relatively tall herbs in the sumtner but low herbs in winter. Whereas some seasonal shifts might reflect changing distribution of environmental features (e,g,, grasses or herbaceous eover, shrub height), the distribution of logs did not chatige between these seasons, suggesting that this was a real shift in mierohabitat use by C. formosus.

Kangaroo rats Differences between these species are expected and largely reflect their general association with Lower and ECOGRAPHY 22:6 |I999|

Upper Sonoran (D. merriami) and Transition (D. agilis) life zones. Nonetheless these species exhibited many more similarities than differences in terms of the habitat tnetrics recorded. Of the 10 and 11 metrics on which both taxa deviated significantly in summer and winter, respectively, they deviated in the same direction on 9 and 7 of these (ca 15% similarity). These species differ in terms of canopy cover and litter depth in both seasons (not significant in winter for D. merriami), and distance to shrubs in the winter (all favored by D. agilis. disfavored by D. merriami). Additionally, both speeies were found at sites with signifieantly less ground eover by roeks, greater distances to oeotillo. 669

low cover by herbs and shrubs, low slope, and loose soils. Both were found at sites rnoderately (but signifieantly) close to shrubs, with less herbaceous ground cover than expected in summer but more in winter, greater log eover in summer but less in winter, and with fewer shrub species than expected in winter.

Sigmodontine rodents Only three sigmodontine rodents (P. crinitus, P. eremicus, and Neotoma) were captured in both seasons, and the remaining two species {P. boylii, P. truei) were captured at very low numbers. Many more similarities than differences were apparent in these species' use of habitat. In both seasons all three speeies favored sites with low canopy cover, close to oeotilio, low litter cover, and relatively hard soils. Additionally, all three of these species seleeted sites with low grass cover, close proximity to shrubs, and moderate but significant distanee to trees in the summer. All used sites low in herbaceous cover, except for Neotoma in summer. High slopes were used by these species, exeept for P. eremieus in summer, and they all selected sites with low ground cover by logs, except for P. crinitus in the summer, which favored sites with high log cover. In winter both P. crinitus and Neotoma were found relatively close to shrubs and with short herbaeeous growth, whereas P. eremicus favored sites with tall herbs located away from shrubs: interestingly, however, the former species also seleeted sites with relatively low cover by live shrubs, and the latter were found at sites with greater shrub cover, suggesting that the distribution of shrub sizes may be inversely related to shrub abundanee along this transect. It is worth noting that some of the significant results obtained in this analysis are trivial consequences of different macrohabitat use. For example, C. spinatus occurs on rocky slopes where logs are not likely to oecur, and D. merriami and all three Chaetodipns typically occupy sites completely or nearly laeking in trees. Nonetheless these results quantify important features of the ecological gradient that are important to the mammal species foutid there.

Discussion In spite of many years of study substantial gaps remain in our knowledge of the ecology of small mammals. The ecological distribution of many species, and the parameters that influence their distributions, have generally been studied only superficially, and much of our understanding is based on observations recorded by field workers early in this century. In this study I have documented distribution patterns of small terrestrial mammals from two faunal districts across a major

670

elevational gradient, in terms of species richness, population abundance, and habitat associations. Habitat associations repotted here supplement state-; ments found in the literature, and provide insights into the tneans by which some species may be segregating resources along this gradient. For example, although the literature reports that the three Cluietodipus species reported here generally have distinct macrohabitat preferences (C. spinatus, rocky habitats; C. formosus, hardpacked gravel, stable alluvium; C. falki.x, often near rocks but also into silt and fine sand habitats; Miller and Stebbins 1964, Ryan 1968, Hoffmeister 1986) they all may be captured sympatrically and even syntopically, albeit not commonly. It is not clear, however, whether this syrnpatry is a consequence of seasonal dispersal (and therefore eonstitutes a fleeting snapshot that would re-sort within a period of days), is an epiphenomenon of resource abundance (lasting until populations approach carrying capacity and again competitively sort among habitats), or if some alternate explanation exists. These speeies differ slightly in body size, but if this were sufficient to allow sympatry then we should see co-occurrence more frequently, and this is not the case. Abramsky et al, (1990) demonstrated that species with similar habitat preferences may be peteeived as seleeting different habitats if one of the speeies is a superior competitor, such that the second speeies is rnore abundant in less preferred habitat. Data presented here suggests that syntopy among these pocket mice rnay occur only ephemerally, such as during dispersal following spring recruitment. The role of interspecific competition vs true habitat preferenee, however, is not clear. Habitat assoeiations in principal components space also reflected the general associations that are observed between speeies, supporting the relevance of PCA to provide insights into patterns of habitat segregation. For example, Peromvsetis boylii, P. truei, Perognathus longimembris, and Dipodomys agilis (San Diegan taxa typical of shrubby habitats) anchored one end of PC axis I, whereas Peromvseus erinitus and Chaetodipus spinatus (desert taxa generally found in exposed rocky habitat) clustered at the opposite pole of this axis, Cluietodipus Jalla.x and Peromyseus eremicus are relatively generalist in their habitat requirements, and they lie at the middle of this axis, while the modei ate habitat specialists Neotoma, C. formosus, and D. merriami fall between the polar habitat specialists and the centrally located generalists. The distribution of richness and abundance was somewhat different from that reported by Ryan (1968) for rodents in the same habitats at Deep Canyon, Speeies riehness was lower at every point in the transition than reported by Ryan, and it peaked at a lower section of the gradient. Differences in richness likely reflect the scale of observation; Ryan listed taxa found in broad habitat types, whereas I have reported animals ECOGRAPHY 22:6 (1999)

captured at specific trapping grids. Although Ryan included diurnal rodents {Ammospermophilus, Spermophilus) in his report, inclusion of these ta.xa in the present report would not bring the observed richness to the level reported by Ryan. Differences in richness may also reflect the use of different trapping methodologies. To estimate abundances Ryan used a single trapping "quadrat" in each of the macrohabitats he studied, but these quadrats varied in size and in the distribution of traps according to the density of vegetation and of rodent burrows. Thus, quadrats ranged in size from 0.52 acres (pifion/juniper) to 1.08 acres (cholla/palo verde). and although traps were placed in rectangular arrays the inter-trap distance was largely set by "placing traps at approximately one per rodent burrow, or clump of plants containing a burrow, so that the area was saturated by the 100 traps used" (Ryan 1968, p. 126). While this is a perfectly justifiable protocol it also results in different sized sample units and difficult interpretation. I have chosen to use sample units of constant area (ca 1.1 ha) to avoid this potential confusion. Total abundances were distributed somewhat differently than reported by Ryan (1968), with a moderate peak at a higher elevation than in Ryan's study. This could also be a consequence of different trapping methods, but it might reflect real changes in the distribution of abundances since Ryan's surveys. The substantial variation in total population abundances and, to a lesser degree, in species richness, between seasons, underscores the temporal variability of such metrics. It seems reasonable to conclude, therefore, that patterns of distribution across this gradient may be dynatnic in time and that local effects (e.g., variations in shrub cover, forage availability, moisture, etc.) may substantially influence these patterns. For example, the El Nilio southern oscillation event of 1998 (a year after this study was conducted) resulted in greatly elevated precipitation throughout southern California, and small mammal numbers rose concomitantly. Habitats that are of low value to some species in years of normal precipitation may become suitable after events such as El Nifio, and emigration may substantially alter local composition and ecological dynamics. Elevated species richness or abundances at intermediate points along ecological gradients have been noted by many authors. Such bulges range from familiar edge effects at small spatial scales (e.g., Harris 1988, Yahner 1988, see also Heske 1995) to patterns over larger ecological transitions (Shmida and Wilson 1985, Patterson et al. 1989, Kelt 1994, Shepherd and Kelt in press) or even biogeographic scales (e.g., Owen 1988, Rosenzweig 1992). Factors resulting in such bulges generally are assumed either to reflect increased resource availability at intermediate regions, or overlap of faunas from either end of a transition. Under the former scenario one would predict a gradual addition of species towards intermediate points along the gradient. ECOGRAPHY 22:6 (1999)

and that titness would be greatest at intermediate sites along these gradients. In the latter scenario intermediate locations may be only marginally suited to the species present such that fitness is predicted to be lower for most species, or populations may only survive in intermediate regions by immigration (e.g., Shmida and Wilson's 1985 mass effects model; see Kunin 1998). It is unlikely that mass effects (sensu stricto) are responsible tor the patterns observed at Deep Canyon, as a number of animals within the regions of highest species richness were observed to have descended testes or to be pregnant, lactating, or otherwise reproductively active. Additionally, the distribution of abundances (especially in summer) supports an argument for declining resource availability at elevations above the agave/ocotillo zone, but only minimal evidence for similar declines at lower elevations. Finally, the small mammal fauna at the richest sites is almost entirely comprised of desert taxa; San Diegan faunal elements only become apparent in the lower and upper piiion-juniper habitats (Ryan reported D. agilis and P. loiii;iiiwmhri.s in agave/ocotillo habitat, but this was not confirmed in this study). Thus, although further research on this would be useful, it seems most likely that reduced resource availability is responsible for diminishing rodent abundances above the agave/ocotillo habitat. If abundance is positively related to species richness (e.g., Preston 1962) then this might also explain the distribution of species richness across this transition. This is a different interpretation, however, than offered by Shepherd and Kelt (in press) for the entire mammal fauna across the same elevational transect. Basing their arguments on the data reported by Ryan (1968) they noted that intermediate sites with the greatest number of species also contained a number of sites typical of both Sonoran and San Diegan faunas. This suggests that faunal overlap may be responsible for elevated richness of larger mammals, but does not preclude the possibility that this in turn is a consequence of a more hospitable habitat. The habitat metrics recorded along this transition suggest that the distributional patterns observed are not simple responses to changes in elevation. Selectivities by many species reflect differential use of habitat characteristics, many of which change markedly across the gradient, providing a broad gradient in habitat types and vegetative structure. Many variables (e.g., number of herbaceous species, agaves, trees, and both small and large logs; shrub height and area; distance to shrubs and ocotillo; litter depth; canopy cover; ground cover by bare ground, logs, grass, herbaceous plants, and by both live and dead shrubs) were significantly positively associated with elevation, whereas other variables (e.g., number of cacti and of holes; distance to agave, cholla, and trees; soil hardness; ground cover by rocks) were significantly negatively associated with elevation (all 671

p < 0,0001 after Bonferoni adjustment). However, most of these regressions explained very little of the variance in these relationships (r- > 0,50 for distance to tree and for ground cover by rock; r- > 0,20 for distance to agave, soil hardness, and ground cover by grass; r~ < 0,20 for all other analyses). Thus, significant associations with the habitat characteristics reported here do not imply gradients with elevation. Rather, these likely reflect selection for habitat types that partially transcend elevations, and it is clear that many species are more common in some habitats than others. For example, species that require loose soils and open habitats, such as D. merriami, are unlikely to fare well in the rocky soils characteristic of the rocky slopes and the agave/ocotillo habitats, but two individuals were captured at higher elevations in the lower pifion-juniper habitat, Chaetodipus spinatus and P. crinitus are relatively saxicolous and therefore unlikely to occur in either upper or lower pifion-juniper or valley floor habitats, although Ryan (1968) reported the latter species from flood plain habitats. It is worth reiterating that selectivities recorded here do not necessarily imply true preferences. Competitive exclusion by the larger C. formosus could be excluding C. spinatus from preferred habitat on the valley floor, resulting in apparent preference for sites with rocky substrate, similar to the situation documented for two gerbil species by Abramsky et al, (1990), Whether the three Chaetodipus that occur in the valley of Deep Canyon occupy different macrohabitats by preference or subsistence is not certain, therefore, and would require experimental manipulations which have not been implemented. Because of access limitations and concerns about fire hazard, this study was not able to sample the entire Deep Canyon Transect, At the lower end of the transect, sand dunes in the Coachella Valley have been largely altered by development or agriculture. At the other end of the transect, limited access and fire hazard prevented us from sampling chaparral and coniferous forest. Data presented by Ryan (1968) suggest that inclusion of these habitats would accentuate the hump in species richness by adding depauperate sites at either end of the gradient (Fig, 3), although this is not a certain outcome. The observed seasonal changes in community composition suggest that species at Deep Canyon may be focusing their activities in those seasons in which they are particularly efficient at foraging. Seasonal rotation in foraging efficiency has been documented among rodents in the Sonoran Desert (Brown 1989, see also Kotler and Brown 1988), but not in the Negev Desert (Brown et al, 1994), Further work on seasonal changes in small mammal communities would be a fruitful area of work. Across a non-elevational transition from forest to steppe in southern South America, Kelt (1996) noted 672

the general uniformity of population and community characteristics of small mammals, leading him to suggest that local processes (e,g,, resource availability) may compensate for geographical dynamics (e,g,, faunal overlap) to yield surprisingly uniform patterns of species richness, community evenness, and biomass across the transition. Such constancy does not characterize elevational gradients in southern Chile (Patterson et al, 1989, Keh et al, 1999) or southeast Asia (Philippines, Rickart et al, 1991; Malaysia, Langham 1983) and does not appear to characterize the elevaticMial gradient studied here. It may be that patterns and perhaps even processes operating across elevational transects are fundamentally different than those operating across "horizontal" ecological gradients involving minimal change in elevation. Since temperature and precipitation generally covary with elevation, and in turn influence ecological processes from rates of soil mineralization and decomposition to potential evapotranspiration, it may not be surprising that gradients lacking such pervasive influences would be structured very differently from gradients exhibiting such processes. The dynamics of ecological transitions continue to provide insightful natural experiments into the factors that structure ecological communities. Further research should be directed towards quantifying the differences between elevational and non-elevational gradients, and how these differ among different groups (e,g,, mammals, birds, insects, plants, etc), Aci<mm!edgemcnts This study was paitially funded by Faculty Research Grants from the Univ, of California. Al Muth and Mark Fisher provided critical logistical support at Deep Canyon, and Patrick Aldrich, Victoria Dye, Chantal Green, and Rodolfo Santos all assisted with collection of data. Burt Kotler reviewer provided useful critiques of this paper.

References Abramsky, Z. et al. 1990. Habitat selection: an experimental field test with two gerbil species. - Ecology 71: 2358-2369. Brown, J. S. 1989. Coexistence on a seasonal resource. - Am. Nat. 133: 168-182. Brown, J. S., Kotler, B. P. and Mitchell, W. A. 1994. Foraging theory, patch use, and the structure of a Negev Desert granivore community. - Ecology 75: 2286-2300. Burk, J. H. 1988. Sonoran Desert vegetation. - In: Barbour, M. G. and Major, J. (eds). Terrestrial vegetation of California, expanded ed, California Native Plant Soc. Spec, Publ. 9: 869-889. Harris, L. D. 1988. Edge effects and conservation of biotic diversity. - Conserv. Biol. 2: 330 332. Heske, E. J. 1995. Mammalian abundances on forest-farm edges versus forest interiors in southern Illinois: is there an edge effect? J. Mammal. 76: 562-568. Hoffmeister, D. F. 1986. Mammals of Arizona. - Univ. of Arizona Press. Karieva, P. 1994, Space: the final frontier for ecological theory. - Ecology 75: 1. Kelt, D. A. 1996. Ecology of small mammals across a strong environmental gradient in southern South America. - J. Mammal. 77: 205 219. ECOGRAPHY 22:6 (1999)

Kelt, D, A,, Meserve, P, L, and Lang, B. K. 1994, Quantitative habitat associations of small mammals in a temperate rainforest in southern Chile: empirical patterns and the importance of ecological scale, - J, Mammal, 75: 890-894. Kelt, D, A, et al, 1999, Scale dependence and scale independence in habitat associations of small mammals in southern temperate rainforest, - Oikos 85: 320-334, Kikkawa, J, and Williams, E, E, 1971. Altitude distribution of land birds in New Guinea, - Search 2: 64-65, Kotler, B. P. and Brown, J. S. 1988. Environmental heterogeneity and the coexistence of desert rodents. - Annu. Rev. Ecol. Syst. 19: 281 307. Kunin, W. E. 1998. Biodiversity at the edge: a test of the importance of spatial "mass effects" in the Rothamsted Park grass experiments. - Proc. Nat, Acad. Sci. U.S.A. 95: 207-212. Langham, N. 1983. Distribution and ecology of small mammals in three rain forest localities on peninsular Malaysia with particular reference to Kedah Peak, - Biotropica 15: 199-206. Levin, S, A, 1992. The problem of pattern and scale in ecology. - Ecology 73: 1943-1967. Miller, A. H. and Stebbins, R. C, 1964. The lives of desert animals in Joshua Tree National Monument. - Univ. of California Press. Owen, J. G. 1988, On productivity as a predictor of rodent and carnivore diversity. - Ecology 69: 1161-1165, Patterson, B. D., Meserve, P. L. and Lang. B, K, 1989, Distribution and abundance of small mammals along an elevational transect in temperate ramforests of Chile. - J. Mammal. 70: 67-78. Patterson, B. D., Meserve, P. L. and Lang. B. K. 1990. Quantitative habitat associations of small mammals along an elevational transect in temperate rainforests of Chile. J. Mammal, 71: 620-633, Pearson, O. P, and Pearson, A. K. 1982. Ecology and biogeography of the sothern rainforests of Argentina. - In: Mares, M. A. and Genoways, H. H. (eds). Mammalian biology in South America. Special Publ. Ser., Pymatuning Lab of Ecology, Univ. of Pittsburgh, Vol. 6: 129-142

ECOGRAPHY 22:6 (1999)

Preston, F. W. 1962. The canonical distribution of commonness and rarity. - Ecology 43: 410-432. Rahbek, C. 1995. The elevational gradient of species richness: a uniform pattern? - Ecography 18: 200-205. Ralph, C. J. 1985, Habitat association patterns of forest and steppe birds of northern Patagonia, Argentina. - The Condor 87: 471-483. Rickart, E. A., Heaney, L. R. and Utzurrum, R. C. B. 1991. Distribution and ecology of small mammals along an elevational transect in southeastern Luzon, Philippines. J. Mammal. 72: 458-469. Rosenzweig, M. L. 1992. Species diversity gradients: we know more and less than we thought. - J. Mammal. 73: 715730. Ryan, R. M. 1968. Mammals of Deep Canyon, Colorado Desert, California. - The Desert Museum, Palm Desert, CA. Shepherd, U, L. and Kelt, D. A. 1999. Mammalian species richness and morphological complexity along an elevational gradient in the arid southwest. - J. Biogeogr. 26: 843-855. Shmida, A. and Wilson, M. V. 1985. Biological determinants of species diversity. - J. Biogeogr. 12: 1-20. Terborgh, J. 1977. Bird species diversity on an Andean elevational gradient. - Ecology 58: 1007-1019. Thorne, R. F. 1988. Montane and subalpine forests of the Transverse and Peninsular Ranges. - In: Barbour, M. G. and Major, J. (eds). Terrestrial vegetation of California, expanded ed. California Native Plant Soc. Spec. Publ. 9: 537-557. Wiens, J. A. 1989. Spatial sealing in ecology. - Funct. Ecol. 3: 385-397. Whitaker, R. H. and Neiring, W. A. 1965. Vegetation of the Santa Catalina Mountains, Arizona: a gradient analysis of the sotith slope. - Ecology 46: 429-452, Yahner, R. H. 1988. Changes in wildlife communities near edges. - Conserv. Biol. 2: 333-339. Zabriskie, J. G. 1979. Plants of Deep Canyon and the eentral Coachella Valley, California, - Univ. of California Press.

673

Ecography 34: 123 131, 2011 doi: 10.1111/j.1600-0587.2010.06371.x # 2011 The Authors. Journal compilation # 2011 Ecography Subject Editor: Nathan J. Sanders. Accepted 6 April 2010

The impact of sterile populations on the perception of elevational richness patterns in ferns Michael Kessler, Sandra Hofmann, Thorsten Kro¨mer, Daniele Cicuzza and Ju¨rgen Kluge M. Kessler (michael.kessler@systbot.uzh.ch), D. Cicuzza and J. Kluge, Systematic Botany, Univ. of Zurich, Zollikerstrasse 107, CH-8008 Zurich, Switzerland. S. Hofmann, Albrecht-von-Haller-Inst. of Plant Sciences, Univ. of Go¨ttingen, Untere Karspu¨le 2, DE 37073 Go¨ttingen, Germany. T. Kro¨mer, Centro de Investigaciones Tropicales, Univ. Veracruzana, Interior de la Exhacienda Lucas Martı´n, Privada de Araucarias s/n, Col. 21 de Marzo, C.P. 91019 Xalapa, Veracruz, Me´xico.

Dispersal may influence the spatial distribution of species richness through mass or source-sink effects, but the extent of sink populations at the community level remains largely unknown due to difficulties of identifying such populations. We compared the richness patterns of ferns in 333 plots along six tropical elevational gradients in America, the Mascarenes, and southeast Asia, using sterile populations as an indication of sink populations. First, we tested whether sterile fern records were more common towards the elevational range limits of the individual species, but found this pattern in only one out of ten cases. It is therefore uncertain if sterile records correspond to marginal sink populations. Second, we compared the elevational richness patterns of sterile and fertile species. In several cases, elevational trends for sterile and fertile records were quite similar, but in others they differed distinctly. The percentage of sterile records per plot decreased with elevation among epiphytic ferns along all six transects, whereas terrestrials showed mixed results (decrease, increase, and U-shaped patterns). The percentage of sterile species records per plot relative to the number of species per plot recovered four significant patterns among the twelve cases analysed: higher percentages at higher species numbers among terrestrial ferns on two transects and lower percentages among epiphytes on two others. Despite the problems with equating sterile records to sink populations, we thus found distinct elevational patterns of sterile records that clearly affected our perception of the overall richness patterns. Ignoring the impact of population dynamics on diversity patterns is thus liable to result in misinterpretations of the diversity patterns.

Geographical patterns of species richness are determined by a wide range of factors at different spatial and temporal scales (Mittelbach et al. 2007). These factors involve such disparate aspects as habitat area (Rosenzweig and Ziv 1999, Lomolino 2001), evolutionary and historical factors (Wiens and Donoghue 2004, Ricklefs 2005), and local ecological conditions (Hawkins et al. 2003, Currie et al. 2004, Evans et al. 2005). Many of these aspects have been extensively studied in the last decades. However, the number of species co-occurring at a site is also influenced by dispersal, an aspect that has received less attention because of the difficulties involved in documenting and quantifying dispersal (Myers and Harms 2009). At the local scale, dispersal can influence patterns of species richness either through dispersal limitation, where potentially suitable habitats are not occupied because propagules fail to reach these sites (Eriksson and Ehrlen 1992, Tilman 1997, Turnbull et al. 2000), or due to mass or source-sink effects (Shmida and Wilson 1985, Pulliam 1988, 2000, Dias 1996). The latter occurs when diaspores of a species are dispersed to suboptimal habitats where the species survive but where they are unable to produce enough offspring to maintain

self-sustaining populations. Although sink populations have been documented in numerous individual species (Gilpin and Hanski 1991, Wilson 1992, Leibold et al. 2004), the extent of such populations at the community level and therefore their influence on richness patterns remains largely unexplored (Myers and Harms 2009). One geographical system along which the influence of such source-sink dynamics may be studied are elevational gradients (Kessler 2009). Elevational patterns of species richness have received increased attention in the last two decades (Rahbek 1995, 2005, Kessler 2001a, Lomolino 2001, McCain 2009), partly because they represent unambiguous gradients that are replicated in different mountain systems. Dispersal has been invoked as a factor to explain elevational patterns of species richness via two different mechanisms. Stevens (1992) proposed that lower elevations have increased richness because species occurring under less extreme, tropical environmental conditions have narrower ecological ranges, resulting in narrower geographical ranges and hence a higher frequency of range margins. This in turn increases the potential for dispersal beyond the source areas. In contrast, maximum species 123

In the present study, we compared the richness patterns of ferns along six tropical elevational gradients in America, the Mascarenes, and southeast Asia. Our aim was to assess the potential impact of source-sink dynamics on elevational richness patterns of ferns, using sterile populations as an indication of sink populations. Ferns are particularly suitable for this kind of study because they are distributed worldwide with a large, but manageable number of species, thus allowing for replicated, quantitative studies (Barrington 1993). Furthermore, in humid tropical mountain forests with their limited climatic seasonality, the majority of fern species is fertile during all or most of the year, reducing the potential impact of the timing of the surveys on the recorded patterns (Sharpe and Mehltreter 2010).

richness at mid-elevations has repeatedly been hypothesized to be due to dispersal of species from both lower and higher elevations, resulting in highest overlap of sink populations at mid-elevations (Rahbek 1997, Kessler 2000b, Lomolino 2001, Grytnes and Vetaas 2002, Grytnes 2003a, b, Kattan and Franco 2004). Already in 1961, van Steenis observed that comparatively low mountains in Java lacked plants at higher elevations that were present at similar elevations on higher mountains. He suggested that on high mountains these species dispersed downwards from the highest peaks to lower elevations and were thus absent on mountains lacking such source habitats (van Steenis 1961). Beyond such descriptive accounts, however, quantitative studies are very rare. One problem is that conducting detailed population-level studies of numerous species at multiple elevations is prohibitively expensive and time consuming (Dias 1996, Diffendorfer 1998). Typically, source populations have higher reproductive rates than sink populations (Pulliam 1988, Robinson et al. 1995, Hansen and Rotella 2002). This implies that, unless vegetative reproduction is involved, populations consisting entirely of sterile individuals are more likely to represent sink populations than populations with fertile individuals. Grytnes et al. (2008) showed that along several Norwegian elevational gradients sterile populations of individual plant species were significantly more frequently found at the upper and lower ends of the elevational ranges of these species. Thus, the distinction of sterile and fertile populations represents a suitable first approach to understanding the potential impact of source-sink dynamics on elevational richness patterns. This approach has been employed in two recent studies. In Bolivia, sterile individuals of numerous palm species are found at higher elevations than their fertile conspecifics, probably representing sink populations located above the elevations where these species can reproduce (Kessler 2000b). In Norway, along several elevational gradients, sterile plant populations are clumped at mid-elevations, emphasizing the hump-shaped richness patterns (Grytnes et al. 2008). These two studies suggest that source-sink dynamics may indeed modify elevational richness patterns, but they also differ from each other in one important aspect. Whereas among the Bolivian palms sterile populations were only found towards the upper elevational limits of the species, in Norway sterile plant populations were documented at both the upper and lower limits of the individual species ranges. Thus, in Bolivia the sterile populations lead to upwards shift in the palm species richness pattern whereas in Norway sterile populations accumulated at the middle of the gradients.

Material and methods Study sites Ferns were studied along six elevational transects, two in Bolivia and one each in Costa Rica, Mexico, La Re´union, and Indonesia (Table 1). Site description are available in previous publications for transects in Bolivia (Kessler 2000a, 2001a), Costa Rica (Kluge and Kessler 2006, Kluge et al. 2006, 2008), and Mexico (Kro¨mer and Acebey 2007, Acebey and Kro¨mer 2008). The island of La Re´union (2512 km2) is a volcanic island about three million years old. Our study was conducted from 300 m elevation in the southeast corner of the island up to the upper limit of closed forest at 2100 m on Piton de Neiges (3071 m). In our study region, mean annual precipitation increases from ca 4000 mm in the lowlands to ca 6000 mm at 550 m and then decreases to ca 2500 mm near tree line. June to September are the driest months. Mean annual temperatures are ca 248C at sea level and 128C at 2000 m. On the island of Sulawesi, Indonesia, we worked in Lore Lindu National Park in the centre of the island. Our transect extended from 300 m in the extreme northwest of the park to 2350 m near the summit of Gunung Rorekatimbu (2450 m). The geological substrate of the area consists of tertiary acid intrusives (Whitten et al. 2002). Mean annual precipitation increases from about 2000 mm in the lowlands to 4000 mm on the higher mountain slopes. There is no well-defined dry season. Field sampling Along all transects we used the same sampling method, detailed in Kessler and Bach (1999), Kessler (2000a,

Table 1. Overview of the study sites and key characteristics of field data and number fern of species. Site

Elevational extent

No. plots

Bolivia, Carrasco NP Bolivia, Masicurı´ Costa Rica, Braulio Carrillo NP La Re´union Indonesia, Sulawesi, Lore Lindu NP Mexico, Los Tuxtlas

220 3950 500 2500 100 3400 100 2750 250 2450 140 1670

81 51 96 18 45 42

124

No. species Source (time of field work in italics) 369 146 426 96 284 91

Kessler 2001a (MK 6-10.1996) Kessler 2000a (MK 7.1995, 5-6.1996) Kluge and Kessler 2005, Kluge et al. 2006 (JK 8.2002-9.2003) MK 3-4.2008 MK, SH, DC, JK 6-8.2007 Kro¨mer and Acebey 2007, Acebey and Kro¨mer 2008 (TK 4-12.2005)

2001a, b), and Kluge et al. (2006), and summarized here. Sampling was conducted in plots of 20 20 m2 each or of plots of similar area but different shape if local conditions did not allow the establishment of square plots. This plot size is small enough to be ecologically homogeneous within while containing a representative sample of the local fern flora (Kessler and Bach 1999). Plots were located in natural forests on slopes, excluding secondary or disturbed habitats as well as ridge or valley-bottom habitats. All fern species present in each plot were recorded, separating epiphytic and terrestrial records, and annotating whether at least one individual of a species was in reproductive condition (fertile records) or whether all individuals were sterile (sterile records). Epiphytic species were sampled by climbing lower parts of trees and by searching for fallen-down branches as well as with the use of trimming poles. In addition, large species, where fertility can also be assessed at a distance, were checked through binoculars. All species present in a study area (but not all species from each plot) were collected at least in triplicate and deposited at the herbaria GOET, UC, and the respective national herbaria. Data analysis Following the approach of Grytnes et al. (2008), we conducted two successive sets of analyses, albeit with different statistics due to different data structures. In the first step of the analysis, we tested whether sterile fern records were more common towards the elevational range limits of the respective species, in order to assess if sterile records indeed reflect sink populations. This was not possible for the La Re´union transect as it included too few plots for a statistically meaningful analysis. We further only included species with at least nine records in a given transect. For each species, we divided its elevational amplitude into quartiles (steps of 25% of the total elevational amplitude of the species), and calculated the percentage of sterile records separately for each quartile. Because the overall percentage of sterile records differed greatly between species (from 10 to 90%), percentages of sterile records in each quartile were transformed as to sum up to 100% for each species. We then used one-way two-tailed KruskalWallis tests to assess whether the percentage of sterile species records in the upper- and lowermost quartiles differed from those in the two central quartiles. Because for many species the upper- or lowermost records corresponded to the extreme points of the transects and may therefore not reflect the potential distributional limits of the species, in a second run of the test we only included range limits that were at least 500 elevational meters from the boundaries of the transects. In this way, for some species we only included either the upper or lower limits, whereas others were fully excluded. In the second part of the analysis, we compared the elevational richness patterns of sterile and fertile species. Elevational patterns of species richness were plotted separately for all species, epiphytes, and terrestrials, as well as for sterile and fertile records for the six study transects, and trend lines fitted by 2nd-order polynomial regressions. Additionally, to reduce the influence of variations in overall species richness, we plotted the percentage of species with

sterile records in each plot against elevation. Finally, to assess whether the distribution of sterile records influences the observed richness patterns, we plotted the percentage of sterile species records per plot against species number per plot. All analyses were conducted with R (R Development Core Team 2008).

Results The first series of tests, assessing whether sterile records are more frequent at the elevational range margins of the species, all resulted in non-significant results (Table 2). When the analyses were redone including only species boundaries at least 500 m from the transect limits to exclude potential biases by considering range limits determined by the transect extent rather than the biology of the species, no significant results were obtained either (data not shown). Comparing the elevational richness patterns of the ferns, five transects showed hump-shaped patterns (Los Tuxtlas only weakly so, because the transect is cut off at 1700 m), whereas on Sulawesi fern richness increased continuously with elevation (Fig. 1). When the records were divided into epiphytic and terrestrial life forms as well as by sterile and fertile records, the patterns became more complex. Epiphytes showed more pronounced hump-shaped patterns than terrestrials, except again in Sulawesi. In several cases (Los Tuxtlas, La Re´union) trend lines for sterile and fertile records were visually quite similar, whereas in others (especially Carrasco) they had distinctly different peaks. Among terrestrials, the differences between sterile and fertile records was even more pronounced, especially in Costa Rica where sterile records showed hump-shaped patterns whereas fertile records increased continuously with elevation, while this pattern is opposite in Masicurı´. In Los Tuxtlas, fertile records even showed an inversed hump-pattern. Focussing on the percentage of sterile species records per plot, epiphytic ferns along all six transects showed decreasing percentages with elevation, in two cases combined with slight increases at highest elevations (Fig. 2). Three of these patterns were statistically significant (Table 3). Among the terrestrial species, the patterns were more diverse, significantly decreasing with elevation in two cases (and not significantly in another one), increasing in one case, and with U-shaped patterns in two others. Table 2. Results of the Kruskall-Wallis tests comparing the frequency of sterile species records in the upper- and lower-most quartiles of the species distributions with that in the two intermediate quartiles. Study area

K value

p value

Epiphytes

Bolivia, Carrasco Bolivia, Masicurı´ Costa Rica Mexico, Los Tuxtlas Sulawesi

2, 2, 2, 2, 2,

210 30 288 60 180

5.78 1.55 2.55 2.14 3.76

0.053 0.415 0.261 0.298 0.172

Terrestrials

Bolivia, Carrasco Bolivia, Masicurı´ Costa Rica Mexico, Los Tuxtlas Sulawesi

2, 2, 2, 2, 2,

132 48 156 24 102

1.24 4.88 2.78 2.34 1.77

0.450 0.083 0.193 0.276 0.376

125

All species

Terrestrials

Epiphytes 50

Bolivia – Carrasco

0 0

1000

2000

3000

4000

0 0

1000

2000

3000

4000

1000

2000

3000

4000

0 0

500

1000 1500 2000 2500 3000

0 0

500

1000 1500 2000 2500 3000

Bolivia − Masicurí 500

1000 1500 2000 2500 3000

Number of species

30 20 10

Costa Rica 0

0 0

500 1000 1500 2000 2500 3000 3500

0 0

500 1000 1500 2000 2500 3000 3500

25 15

10 5

Mexico – Los Tuxtlas

0 0

500

1000

1500

2000

0 0

500

1000

1500

2000

500

1000

1500

2000

30 25 20

15 10 5

Réunion 0

500

1000 1500 2000 2500 3000

0 0

500

1000 1500 2000 2500 3000

500

1000 1500 2000 2500 3000

500

1000 1500 2000 2500 3000

40 15

30 10

20 5

Sulawesi 0

0 0

500

1000 1500 2000 2500 3000

0 0

500

1000 1500 2000 2500 3000

Elevation (m)

Figure 1. Elevational patterns of fern species richness along the six elevational study transects for all species (left), terrestrials (centre), and epiphytes (right). Black dots and black lines mark total species richness, dark grey dots and wide dashed lines fertile species plot records, and light grey dots and dotted lines sterile species plot records. Lines are 2nd-order polynomial regression lines.

126

100

0 0

1000

2000

3000

4000

100

Bolivia − Masicurí

0 0

Percent of sterile species

Bolivia – Carrasco

500

1000 1500 2000 2500 3000

100

Costa Rica 0

0 0

500 1000 1500 2000 2500 3000 3500

100

Mexico − Los Tuxtlas

0 0

500

1000

1500

2000

100

Réunion 0

0 0

500

1000 1500 2000 2500 3000

100

Sulawesi 0

0 0

500

1000 1500 2000 2500 3000

Elevation (m)

Number of species

Figure 2. Percentage of sterile species records relative to elevation (left) and number of species of the respective life form (right) along the six elevational transects. Data and regression lines (2nd-order polynomial regression lines) are given for terrestrial species (black) and epiphytic species (grey). Continuous lines mark regressions significant at the pB0.05 level, non-significant lines are dashed.

127

Table 3. Determination coefficients and significancy levels of polynomic models describing the relationship of percentage of sterile species records to elevation along the six study transects. p-valuesB0.05 marked in bold. Study area

R2 value

p value

Epiphytes

Bolivia, Carrasco Bolivia, Masicurı´ Costa Rica Mexico, Los Tuxtlas La Re´union Sulawesi

0.42 0.05 0.40 0.43 0.16 0.13

B0.001 0.370 B0.001 B0.001 0.352 0.071

Terrestrials

Bolivia, Carrasco Bolivia, Masicurı´ Costa Rica Mexico, Los Tuxtlas La Re´union Sulawesi

0.10 0.26 0.42 0.28 0.07 0.06

0.018 0.001 B0.001 0.002 0.629 0.273

Finally, relating the percentage of sterile species records per plot to the number of species per plot recovered four significant patterns among the twelve cases analysed (Fig. 2, Table 4). The significant cases documented higher percentages of sterile records at higher species numbers per plot among terrestrial ferns at Masicurı´ and in Costa Rica, and higher percentages of sterile records at lower species numbers per plot among epiphytic ferns in Los Tuxtlas and Sulawesi. The non-significant patterns showed four decreases among the terrestrial species, and three increases as well as one decrease among the epiphytic species.

Discussion Our study of the elevational distribution of sterile ferns addressed two complementary questions. First, we asked whether sterile species records are more frequent along species range margins and can hence be considered to represent sink populations. Second, we asked whether sterile records show distinct elevational richness patterns that influence our perception of the overall species richness of ferns. The first of these questions has to be answered negatively, whereas for the second question the answer is complex, with differences between life forms and study sites. Table 4. Determination coefficients and significancy levels of polynomic models describing the relationship of percentage of sterile species records per plot to number of species per plot along the six study transects. p-valuesB0.05 marked in bold. Study area

R2 value

p value

Epiphytes

Bolivia, Carrasco Bolivia, Masicurı´ Costa Rica Mexico, Los Tuxtlas La Re´union Sulawesi

0.03 0.02 0.02 0.19 0.12 0.12

0.140 0.418 0.126 0.005 0.216 0.023

Terrestrials

Bolivia, Carrasco Bolivia, Masicurı´ Costa Rica Mexico, Los Tuxtlas La Re´union Sulawesi

0.01 0.12 0.05 0.05 0.00 0.02

0.480 0.013 0.034 0.164 0.982 0.320

128

We were unable to document that sterile records are clumped at the elevational range margins of the species. This contrasts with the results of Grytnes et al. (2008) who showed that for flowering plants along several elevational transects in Norway sterile records were indeed more common in the upper- and lowermost quartiles than expected by chance. However, Grytnes et al. (2008) also observed that the ability to detect this pattern depended on the spatial size of the plots, with larger plots revealing more diffuse patterns. They believed that this was because species records were classified only as fertile or sterile, rather than reflecting the actual number of sterile and fertile plants. Thus, a plot record was considered as ‘‘fertile’’ if at least one individual was found in fertile condition, regardless of whether hundreds of other plants were sterile. The probability of finding such scattered fertile individuals is higher in larger plots, reducing the ability of this approach to detect sink populations. In our study we used the same method as Grytnes et al. (2008), and it is conceivable that our plot size of 400 m2 is also too large for this approach. In both studies, the data were originally gathered for a different suite of analyses, with fertility status only included as an additional parameter. Clearly, more detailed data are needed, including precise counts of sterile and fertile specimens per plot. Our difficulty in detecting increases of sterile records at the elevational range margins of the species may also have been determined by the dispersal mechanisms of ferns. Contrary to seed plants, where dispersal mostly takes place by seeds of varying but usually substantial size and accordingly relatively restricted dispersal distances, in ferns dispersal is by dust-like spores that have highly efficient wind dispersal (Barrington 1993). There are no comparative studies of the diaspore rain of ferns and angiosperms along elevational gradients, but it is certainly plausible that the dispersal kernels of ferns and angiosperms are different, and accordingly that the spatial positioning of sink populations may be different. In particular, ferns have an independent gametophytic life stage which we did not consider at all due the difficulty of sampling and identifying fern gametophytes in tropical forests. It is conceivable that much of the ecological filtering in fern sink populations takes place in this gametophytic phase and may thus no longer be evident in the sporophytic stage studied by us. Thus, the detection of sterile populations may require different approaches for ferns and angiosperms. Furthermore, sterile populations are certainly not an unambiguous measure of sink populations. Especially among long-lived organisms under extreme environmental conditions successful reproduction may only take place at long, sometimes decadal intervals. Moreover, sterile populations can result from processes other than source-sink dynamics. For example, many ferns have vegetative reproduction and such species are not randomly distributed with elevation (Kluge and Kessler 2007). However, detailed studies of the population dynamics along elevational gradients can realistically only be conducted for selected species and will probably never be achieved for entire, diverse communities. Thus, while these problems have to be taken into account, at present the distinction of sterile and fertile populations represents a suitable first approach to

understanding the potential impact of source-sink dynamics on elevational richness patterns. Despite the problems with equating sterile records to sink populations, in our study we found distinct elevational patterns of sterile records that clearly affected our perception of the overall richness patterns. This was most conspicuous for the terrestrial ferns at Los Tuxtlas and Costa Rica, where the fertile records showed completely different elevational patterns compared to the hump-shaped distribution of sterile records (Fig. 1). In these cases, the overall hump-shaped richness pattern of terrestrial ferns was clearly caused by the sterile records. Accordingly, our previous interpretations of such hump-shaped patterns (Kessler 2000a, 2001a, Kluge et al. 2006) actually attempted to explain the distribution of sterile populations, while unknowingly ignoring the fact that the fertile populations showed distinct patterns that might be explained in different ways. In such a situation, the explanatory factors that we considered, including climatic conditions, topography, or the mid-domain effect, may all have missed the actual point. However, not all transects showed such clearcut patterns. Indeed, we found a high variability of patterns along the different transects and for the two main life forms. Among epiphytes, all six transects showed a tendency towards decreasing percentages of sterile species records with elevation, although this was only significant for three transects. On the other hand, there was no clear relationship of total species richness per plot to the percentage of sterile records, suggesting that our overall perception of the richness patterns of epiphytic ferns along elevational transects is not strongly influenced by the sterile records. Terrestrial ferns, in contrast, showed a much more variable distribution of sterile records, with increasing, decreasing, and U-shaped patterns. The latter, found along the transects in Mexico and on La Re´union, are particularly interesting in that they show that sterile populations of terrestrial ferns can be most common in species-poor communities under (for ferns) stressful environmental conditions. The low diversity of ferns in tropical lowlands and at high elevations has repeatedly been interpreted as reflecting extreme environmental conditions, namely limited water availability in the lowlands and low temperatures in the highlands (Kessler 2001b, Bhattarai et al. 2004, Kluge et al. 2006). Our study suggests that under these conditions sterile populations may be more common, either because they represent sink populations derived from source populations at mid-elevations, or because under extreme environmental conditions vegetative reproduction may be more common. On the other hand, along the Masicurı´ and Costa Rica transects sterile records of terrestrial ferns were most common in the most species rich plots, suggesting that here potential sink populations were accumulated at midelevations, inflating the hump-shaped richness patterns. This observation is accordance with the hypotheses of numerous researchers (Rahbek 1997, Kessler 2000b, Lomolino 2001, Grytnes and Vetaas 2002, Grytnes 2003a, b, Kattan and Franco 2004) as well as with the observations of Grytnes et al. (2008) in Norway.

How might these contradictory results of different transects studies be reconciled? One possibility is the influence of topography in the different study areas. Both the Los Tuxtlas and La Reúnion transects are located on isolated volcanic mountains without large mountain massifs in close proximity. Due to the conical shape of these mountains, their high-elevation habitats have a very limited spatial extent and accordingly species-poor floras. The young age of the mountains and their vegetation (last eruption at Volcań San Martıń Tuxtla was in 1793, Guevara et al. 2004) and their active volcanic history further limit their species pools. In such a situation, it is conceivable that most plant species found near the mountain tops are sink populations derived from sources lower down on the mountain, as indeed suggested by our data. In contrast, the transects in Bolivia and Costa Rica are located on older, more extensive mountain ranges with a well-developed high elevation flora. Other geographic, climatic, or historical factors might further influence the spatial distribution of source and sink plant populations on mountains. A comparative analysis of numerous mountains with contrasting conditions might bring us a long step forward in understanding the role of population dynamics in shaping elevational richness patterns. For now, we can conclude that sterile populations of especially terrestrial ferns show distinct elevational distribution patterns that influence our perception of elevational richness patterns among ferns. Future studies of such patterns should take into account the possibility that source-sink dynamics determine or at least modify the observed patterns, both for ferns and other organisms. This does not only apply to elevational gradients but also to other ecological or geographical gradients. It is likely that the impact of population dynamics will be more pronounced over distances within the scale of regular dispersal of the taxon under consideration and less so for extensive gradients exceeding the scale of usual dispersal such as the latitudinal gradient (Kessler 2009). In any case, ignoring the impact of population dynamics on diversity patterns is liable to result in misinterpretations of the diversity patterns, as shown in our study for the terrestrial ferns. Acknowledgements First of all, we wish to thank A. R. Smith for tireless plant identification and J.-A. Grytnes for valuable comments on the manuscript. In Bolivia, we thank the Herbario Nacional de Bolivia, A. Acebey, K. Bach, S. G. Beck, M. Cusicanqui, S. K. Herzog, J. Gonzales, I. Jimenez, A. de Lima, R. de Michel, M. Moraes, the Direccioń Nacional de Conservacioń de la Biodiversidad, the park guards of Carrasco National Park, the mayor of Villa Tunari, Hotel El Puente, and the Facultad de Agricultura, Univ. Mayor de San Simoń, Cochabamba. For the study in Costa Rica, we thank F. Corrales for invaluable help during the field work, the park rangers of the Sistema Nacional de Area de Conservaciones (SINAC) and the Area de Conservacioń Cordillera Volcańica Central (ACCVC) in Costa Rica, the staff of Biological Station La Selva and the Organisation for Tropical Studies OTS, and the National Herbarium in San Jose´ for logistical support. The study in Los Tuxtlas was supported by a postdoctoral grant to TK from the Univ. Nacional Autońoma de Me´xico, who also thanks A. Acebey for fieldwork assistance, and the directors of the Los Tuxtlas Biological Research Station, R. Coates and M. Ricker, for logistical support. Field work on La

129

Re´union was funded by am ERASMUS grant to MK and would have been impossible without the help of D. Strasberg and C. Ah Peng. The Sulawesi data were collected within the field work of the STORMA project, whose framework conditions were of invaluable importance. Field work was partly funded by the Deutsche Forschungsgemeinschaft DFG (grants to MK) and the Deutscher Akademischer Austauschdienst DAAD (to JK).

References Acebey, A. and Kro¨mer, T. 2008. Diversidad y distribucioń de las araćeas de la Reserva de la Biosfera Los Tuxtlas, Veracruz, Me´xico. Rev. Mex. Biodiv. 79: 465 471. Barrington, D. S. 1993. Ecological and historical factors in fern biogeography. J. Biogeogr. 20: 275 280. Bhattarai, K. R. et al. 2004. Fern species richness along a central Himalayan elevational gradient, Nepal. J. Biogeogr. 31: 389 400. Currie, D. J. et al. 2004. Predictions and tests of climate-based hypotheses of broad-scale variation in taxonomic richness. Ecol. Lett. 7: 1121 1134. Dias, P. C. 1996. Sources and sinks in population biology. Trends Ecol. Evol. 11: 326 330. Diffendorfer, J. E. 1998. Testing models of source-sink dynamics and balanced dispersal. Oikos 81: 417 433. Eriksson, O. and Ehrlen, J. 1992. Seed and microsite limitation of recruitment in plant populations. Oecologia 91: 360 364. Evans, K. L. et al. 2005. Species energy relationships at the macroecological scale: a review of the mechanisms. Biol. Rev. 80: 1 25. Gilpin, M. E. and Hanski, I. A. 1991. Metapopulation dynamics: empirical and theoretical investigations. Academic Press. Grytnes, J. A. 2003a. Ecological interpretations of the middomain effect. Ecol. Lett. 6: 883 888. Grytnes, J. A. 2003b. Species richness patterns of vascular plants along seven altitudinal transects in Norway. Ecography 26: 291 300. Grytnes, J. A. and Vetaas, O. R. 2002. Species richness and altitude: a comparison between null models and interpolated plant species richness along the Himalayan altitudinal gradient, Nepal. Am. Nat. 159: 294 304. Grytnes, J. A. et al. 2008. Can the mass effect explain the midaltitudinal peak in vascular species richness? Basic Appl. Ecol. 9: 373 382. Guevara, S. et al. 2004. Los Tuxtlas. El paisaje de la sierra. Inst. de Ecologıá, A.C., Xalapa. Hansen, A. J. and Rotella, J. J. 2002. Biophysical factors, land use, and species viability in and around nature reserves. Conserv. Biol. 16: 1112 1122. Hawkins, B. A. et al. 2003. Energy, water, and broad-scale geographic patterns of species richness. Ecology 84: 3105 3117. Kattan, G. H. and Franco, P. 2004. Bird diversity along elevational gradients in the Andes of Colombia: area and mass effects. Global Ecol. Biogeogr. 13: 451 458. Kessler, M. 2000a. Elevational gradients in species richness and endemism of selected plant groups in the central Bolivian Andes. Plant Ecol. 149: 181 193. Kessler, M. 2000b. Upslope-directed mass effect in palms along an Andean elevational gradient: a cause for high diversity at midelevations? Biotropica 32: 756 759. Kessler, M. 2001a. Patterns of diversity and range size of selected plant groups along an elevational transect in the Bolivian Andes. Biodivers. Conserv. 10: 1897 1920.

130

Kessler, M. 2001b. Pteridophyte species richness in Andean forests in Bolivia. Biodivers. Conserv. 10: 1473 1495. Kessler, M. 2009. The impact of population processes on patterns of species richness: lessons from elevational gradients. Basic Appl. Ecol. 10: 295 299. Kessler, M. and Bach, K. 1999. Using indicator groups for vegetation classification in species-rich Neotropical forests. Phytocoenologia 29: 485 502. Kluge, J. and Kessler, M. 2005. Inventory of pteridophytes along an elevational transect in Braulio Carrillo National Park, La Selva Biological Stationand Cerro de la Muerte, Costa Rica. Brenesia 63 64: 11 34. Kluge, J. and Kessler, M. 2006. Fern endemism and its correlates: constribution from an elevational transect in Costa Rica. Divers. Distrib. 12: 535 545. Kluge, J. and Kessler, M. 2007. Morphological characteristics of fern assemblages along an elevational gradient: patterns and causes. Ecotropica 13: 27 43. Kluge, J. et al. 2006. What drives elevational patterns of diversity? A test of geometric constraints, climate, and species pool effects for pteridophytes on an elevational gradient in Costa Rica. Global Ecol. Biogeogr. 15: 358 371. Kluge, J. et al. 2008. Elevational distribution and zonation of tropical pteridophyte assemblages in Costa Rica. Basic Appl. Ecol. 9: 35 43. Kro¨mer, T. and Acebey, A. 2007. The bromeliad flora of the San Martı´n Tuxtla volcano, Veracruz, Mexico. J. Bromeliad Soc. 57: 62 69. Leibold, M. A. et al. 2004. The metacommunity concept: a framework for multi-scale community ecology. Ecol. Lett. 7: 601 613. Lomolino, M. V. 2001. Elevational gradients of species-density: historical and prospective views. Global Ecol. Biogeogr. 10: 3 13. McCain, C. M. 2009. Global analysis of bird elevational diversity. Global Ecol. Biogeogr. 18: 346 360. Mittelbach, G. G. et al. 2007. Evolution and the latitudinal diversity gradient: speciation, extinction and biogeography. Ecol. Lett. 10: 315 331. Myers, J. A. and Harms, K. E. 2009. Seed arrival, ecological filters, and plant species richness: a meta-analysis. Ecol. Lett. 12: 1250 1260. Pulliam, H. R. 1988. Sources, sinks, and population regulation. Am. Nat. 132: 652 661. Pulliam, H. R. 2000. On the relationship between niche and distribution. Ecol. Lett. 3: 349 361. R Development Core Team 2008. R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna. Rahbek, C. 1995. The elevational gradient of species richness: a uniform pattern? Ecography 18: 200 205. Rahbek, C. 1997. The relationship among area, elevation, and regional species richness in neotropical birds. Am. Nat. 149: 875 902. Rahbek, C. 2005. The role of spatial scale and the perception of large-scale species-richness patterns. Ecol. Lett. 8: 224 239. Ricklefs, R. E. 2005. Phylogenetic perspectives on patterns of regional and local richness. In: Bermingham, E. et al. (eds), Tropical rainforest: past, present, and future. Univ. Chicago Press, pp. 16 40. Robinson, S. K. et al. 1995. Regional forest fragmentation and the nesting success of migratory birds. Science 267: 1987 1990. Rosenzweig, M. L. and Ziv, Y. 1999. The echo pattern of species diversity: pattern and process. Ecography 22: 614 628.

Sharpe, J. M. and Mehltreter, K. 2010. Ecological insights from fern population dynamics. In: Mehltreter, K. and Sharpe, J. M. (eds), Fern ecology. Cambridge Univ. Press, in press. Shmida, A. and Wilson, M. W. 1985. Biological determinants of species diversity. J. Biogeogr. 12: 1 20. Stevens, G. C. 1992. The elevational gradient in altitudinal range: an extension of Rapoportâ&#x20AC;&#x2122;s latitudinal rule to altitude. Am. Nat. 140: 893 911. Tilman, D. 1997. Community invasibility, recruitment limitation, and grassland biodiversity. Ecology 78: 81 92. Turnbull, L. A. et al. 2000. Are plant populations seed-limited? A review of seed sowing experiments. Oikos 88: 225 238.

van Steenis, C. G. G. J. 1961. An attempt towards an explanation of the effect of mountain mass elevation. Proc. Kon. Ned. Akad. Wetensch. C64: 435 442. Whitten, T. et al. 2002. Ecology of Sulawesi. Periplus Editions. Wiens, J. J. and Donoghue, M. J. 2004. Historical biogeography, ecology and species richness. Trends Ecol. Evol. 19: 639 644. Wilson, D. S. 1992. Complex interactions in metacommunities, with implications for biodiversity and higher levels of selection. Ecology 73: 1984 2000.

131

Ecography 34: 372 383, 2011 doi: 10.1111/j.1600-0587.2010.06460.x # 2011 The Authors. Ecography # 2011 Ecography Subject Editor: Jeremy T. Kerr. Accepted 5 July 2010 The first two authors (O. L. and N. L.) contributed to this paper equally and are listed alphabetically

Can we predict butterfly diversity along an elevation gradient from space? Oded Levanoni, Noam Levin, Guy Pe’er, Anne Turbe´ and Salit Kark O. Levanoni, G. Pe’er, A. Turbe´ and S. Kark (salit@hebrew.edu), The Biodiversity Research Group, Dept of Ecology, Evolution and Behavior, The Silberman Inst. of Life Science, The Hebrew Univ. of Jerusalem, IL-91904 Jerusalem, Israel. Present address of GP: Dept of Conservation Biology, UFZ Helmholtz Centre for Environmental Research, Permoserstr. 15, DE-04318 Leipzig, Germany. N. Levin, Dept of Geography, Faculty of Social Sciences, The Hebrew Univ. of Jerusalem, Mount Scopus, IL-91905 Jerusalem, Israel.

An important challenge in ecology is to predict patterns of biodiversity across eco-geographical gradients. This is particularly relevant in areas that are inaccessible, but are of high research and conservation value, such as mountains. Potentially, remotely-sensed vegetation indices derived from satellite images can help in predicting species diversity in vast and remote areas via their relationship with two of the major factors that are known to affect biodiversity: productivity and spatial heterogeneity in productivity. Here, we examined whether the Normalized Difference Vegetation Index (NDVI) can be used effectively to predict changes in butterfly richness, range size rarity and beta diversity along an elevation gradient. We examined the relationship between butterfly diversity and both the mean NDVI within elevation belts (a surrogate of productivity) and the variability in NDVI within and among elevation belts (surrogates for spatial heterogeneity in productivity). We calculated NDVI at three spatial extents, using a high spatial resolution QuickBird satellite image. We obtained data on butterfly richness, rarity and beta diversity by field sampling 100 m quadrats and transects between 500 and 2200 m in Mt Hermon, Israel. We found that the variability in NDVI, as measured both within and among adjacent elevation belts, was strongly and significantly correlated with butterfly richness. Butterfly range size rarity was strongly correlated with the mean and the standard deviation of NDVI within belts. In our system it appears that it is spatial heterogeneity in productivity rather than productivity per se that explained butterfly richness. These results suggest that remotely-sensed data can provide a useful tool for assessing spatial patterns of butterfly richness in inaccessible areas. The results further indicate the importance of considering spatial heterogeneity in productivity along elevation gradients, which has no lesser importance than productivity in shaping richness and rarity, especially at the local scale.

Elevation gradients and species diversity Ecologists have a long lasting interest in diversity patterns across spatial gradients (Rosenzweig 1995, Lomolino 2001). Many earlier studies have examined changes in biodiversity along elevation gradients, yet no single spatial pattern has been identified thus far (Shmida and Wilson 1985, Rahbek 1995, 2005, Lomolino 2001, Grytnes and McCain 2007, Nogue´s-Bravo et al. 2008). Therefore, despite the interest in predicting patterns along climatic gradients, such predictions remain challenging. This is especially true in remote areas that are difficult to access and to sample in the field, such as mountains (Levin et al. 2007). The current availability of satellite imagery at detailed spatial resolutions (Kark et al. 2008) has created an opportunity to study and gain information about remote areas (Levin et al. 2007). Several studies (Bawa et al. 2002, Oindo 2002, Kerr and Ostrovsky 2003, Foody and Cutler 2006, Levin et al. 2007, Gillespie et al. 2008) have suggested that plant species richness can be

372

effectively predicted using simple indices derived from remotely-sensed images, such as the Normalized Difference Vegetation Index (NDVI; Tucker 1979, Tucker and Sellers 1986). This predictive ability is likely related to the fact that both primary productivity and habitat heterogeneity, two of the major factors shaping biodiversity patterns, can be relatively easily estimated by calculating satellite-derived vegetation indices (Kerr and Ostrovsky 2003, Gillespie et al. 2008, Rocchini et al. 2010, for reviews on the use of satellite images in ecology and biodiversity research). While predicting animal diversity using vegetation indices is more challenging, because NDVI is based on vegetation-related variables, Kerr et al. (2001) have shown that remote sensing tools can accurately predict butterfly richness at a semi-continental scale using low spatial resolution satellite imagery (1 and 8 km) and biodiversity data at a coarse scale (2.58 grid cells). They found that satellite-derived heterogeneity measures of land cover were strongly correlated with butterfly richness when examined

across Canada. However, the potential of remotely-sensed estimates of spatial heterogeneity in productivity to predict animal diversity at smaller scales (covering a small area using high spatial resolution satellite imagery and high spatial resolution biodiversity data) is less well understood. Our study focuses on butterflies, which are often considered to be good surrogates of biodiversity, being tightly dependant on a range of plants. They are known to respond to various environmental factors, to vegetation changes (reviewed in Peâ&#x20AC;&#x2122;er and Settele 2008a) and to climate changes (Parmesan et al. 1999, Thomas et al. 2004). Butterflies are relatively easy to sample in the field (Nowicki et al. 2008, Peâ&#x20AC;&#x2122;er and Settele 2008b), and have been successfully used in studies of ecological gradients (Blair 1999, Fleishman et al. 2000) and in conservation and global change research (Samways 1989, Kremen 1992, Kim 1993, Parmesan et al. 1999, Thomas et al. 2004, Pin Koh and Sodhi 2005, Thomas 2005, Parmesan 2006). However, fewer studies have tested the productivity richness relationship in butterflies using remotely sensed indices of vegetation (but see Kerr et al. 2001, Bailey et al. 2004, Seto et al. 2004). The relationship between spatial heterogeneity in productivity (as estimated using remotely sensed vegetation indices) and species richness along local gradients has especially remained under-explored (Bailey et al. 2004). This is surprising, since spatial environmental heterogeneity is hypothesized to be an important factor shaping ecological communities and is related with species richness (Rosenzweig 1995, Atauri and de Lucio 2001, Rocchini et al. 2010). Furthermore, spatial heterogeneity in productivity has received much attention in studies at large scales (Kerr and Packer 1997, Jetz and Rahbek 2002). Here, we aim to examine whether butterfly richness, rarity, and beta diversity along an elevation gradient (hereby termed diversity estimates, see Methods) can be accurately predicted using satellite-derived vegetation indices. We asked whether the mean NDVI and estimates of spatial variation in NDVI can successfully predict: 1) butterfly richness within elevation belts along the elevation gradient, 2) changes in species composition among elevation belts, and 3) changes in range size rarity along the elevation gradient. We predicted that the mean values of NDVI and the spatial heterogeneity in NDVI, both within and among 100 m elevation belts, will be useful predictors for butterfly richness, beta diversity and rarity along the elevation gradient. While we did not aim to examine the effect of all potential environmental variables affecting species diversity along the altitudinal gradient, we did examine the effect of two major factors that can potentially confound the NDVIrichness relationship along the elevation gradient, namely area and the mid-domain effect (Grytnes and McCain 2007). A mid-domain effect (a peak in richness at midelevations, or mid landmass, due to spatial geometric constraints), is predicted where landmass boundaries, such as mountain tops, restrict species ranges and the overlap of variously sized ranges creates a peak in species richness at mid-elevations (Colwell and Lees 2000, Colwell et al. 2004, Grytnes and McCain 2007).

Methods Study area Located in north-eastern Israel along the border with Lebanon and Syria, Mt Hermon (33.25?N, 35.48?E), is part of the Anti-Lebanon Mountains, which are isolated from the main mountain ranges of the Middle East, Asia and Europe (Shmida 1977). The parent material is homogeneous, hard Jurassic limestone and dolomite, forming Terra Rossa soils (Shmida 1977). The terrain is characterized by steep rocky limestone Karst slopes (Auerbach and Shmida 1993). Mt Hermon is an elongated anticline that extends NE-SW over 35 km and rises from 300 to 2814 m over a 13-km distance on its SW side in Israel, where our study area is located (the highest point is 2224 m). The climate is Mediterranean, with rainy or snowy winters, and hot dry summers. Precipitation ranges from 600 to 1500 mm yr 1 along the mountain, and above an elevation of 1500 m consists mostly of snow. As in other mountains, temperature decreases, while solar radiation and precipitation increase with elevation. Snow usually begins to cover the slopes of Mt Hermon in the first half of January and lasts until April. Snow patches usually remain until June above 1900 m, mainly on SE facing slopes and in the valleys (Shmida 1980). Three main vegetation belts have been defined in earlier studies of Mt Hermon (Shmida 1977, 1980). These include: 1) evergreen Mediterranean maquis (300 1200 m); 2) the xero-montane open forest (1200 1900 m) and 3) the subalpine Tragacanthic belt (1900 2814 m) (Fig. 1). The part of the Hermon in Israel ranges approximately 7300 ha, most of which is a nature reserve since 1972 (Levin et al. 2007). Selection of study sites and sampling design Most studies of butterfly richness, as well as systematic monitoring schemes, rely on line transect sampling (van Swaay et al. 1997, KuÂ¨ hn et al. 2008, Nowicki et al. 2008), as described by Pollard (1977) and standardized by Pollard and Yates (1993). Quadrat sampling is less often used for butterfly sampling (but see Su et al. 2004, Grill et al. 2005). However, quadrat sampling enables one to concentrate higher sampling effort in given locations and provides comparative ability with sampling methodologies used for other taxa, such as quadrat sampling for plants or point-counts for birds. To examine changes in diversity both among and within elevation belts, we conducted both line transect and quadrat sampling. We divided the elevation range of our study area (500 2200 m) into elevation classes 100 m high, thus obtaining 17 elevation belts. The area of the elevation belts ranged between a minimum of 5.8 ha (2100 2200 m) to 117 ha (1300 1400 m). To reduce variability resulting from different surface aspects, all quadrats and transects were located on SW facing slopes (following Levin et al. 2007), which are the most common slopes in the study area, corresponding with the shape of the Mt Hermon anticline that extends from SW to NE (see Levin et al. 2007 for

373

Figure 1. Map showing NDVI values in the study area on Mt Hermon. The photos on the right show the three major vegetation belts, from top to bottom: subalpine Tragacanthic belt (above 1900), xero-montane open forest (1200 1900 m) and evergreen Mediterranean maquis (300 1200 m). Photos by S.K.

further details). We marked two grid squares of 50 20 m (1000 m2) within each 100 m elevation belt, totalling 34 quadrats. This enabled us to examine changes along the elevation gradient and to obtain repetition within elevation belts. The grid squares were pre-selected using remote sensing tools, being those with NDVI values close to the median NDVI of the whole elevation belt in the SW slope of the study area, so as to assure that they indeed represent vegetation of their respective elevation belt. The same quadrats were used for plant sampling in an earlier study (Levin et al. 2007). Butterfly sampling We sampled the butterflies along the elevation gradient between 500 and 2200 m. Butterfly sampling was conducted during two years in the peak activity season of most Hermon butterflies (February September 2005, and March August 2006; sensu Benyamini 2002), with each of the vegetation belts visited on average over 15 separate dates. Butterfly sampling was conducted by two Lepidoptera experts capable of recognizing all species in the field (O.L. and G.P.). A butterfly expert and a note-taker walked along parallel lines inside each sampling quadrat for 20 min. The sampling duration of 20 min in the quadrats was determined based on species-accumulation curves generated in preliminary work, using EstimateS ver. 8.0 (Colwell 2006). These indicated that accumulation is reached, on average, after 1293 min of sampling (sampling ]95% of the species). The transect lines began from the corner of each quadrat and led to the next quadrat (i.e. they were located outside the quadrats). Each line transect was 300 m long (the length was determined based again on species-accumulation curves (Colwell 2006), indicating the accumulation of at least 95% of the species within 250950 m of sampling). The transect line was divided into sections 50 m long (Pollard and Yates 1993) in order to allow the calculation of species-accumulation curves with distance. The line transects were marked by metal rods 374

(0.5 m high) in order to ensure accurate repetition of transects during all sampling visits. Butterfly species recognition in both quadrats and transect lines was performed visually. All individuals first seen within a distance of 5 m from the observers were recorded. When needed, in order to verify identification, we captured the butterfly with a sweeping net and then immediately released it at same location. Line transect sampling and quadrat sampling were performed sequentially. All sampling was performed between 9:00 am (10:00 am in elevations 1500 m) and 15:00, when ambient temperatures were 208C, cloud cover was B50%, and wind speed was B4 km h 1. Weather conditions were recorded using a hand-held Kestrel 4000 weather station. Repeated sampling visits to each 100 m elevation belt were organized so that there were different starting times for each belt in order to reduce potential biases that are related to the timing of sampling. Diversity estimates In order to increase statistical power, we pooled the data from the two quadrats and transects within each 100 m elevation belt. This was done after a preliminary analysis, which showed that the butterfly diversity estimates were rather similar for quadrats and for transects when analysed separately and since our goal in this paper was not to compare the different methodologies. We calculated values of richness, beta diversity and range size rarity as sampled in each 100 m elevation belt. Species richness (alpha diversity) was calculated by summing up all the species that were observed in the two quadrats and transects within each 100 m elevation belt. Following McCain (2004), species were assumed to be present at an elevation if they were detected at both higher and lower elevations adjacent to a given belt. In cases where larger elevation-belt gaps in appearance were found (over 100 m), we only ‘‘filled in’’ the occurrence of a species in a given elevation belt if the data was consistent with the known range of distribution of

the species in Mt Hermon based on the literature (Benyamini 1993, 2002). Filling in was done for 25 of the 83 butterfly species sampled in this study. For 15 of these 25 species only a single elevation belt was filled-in. Results were very similar when analysis was repeated without filling in. Various estimators for beta diversity have been suggested in the literature, as reviewed by Koleff et al. (2003). We adopted the bsim (beta sim) estimator, which was considered by Lennon et al. (2001) and Koleff et al. (2003) as a reliable estimate. Preliminary results indicated that it produced very similar results to the estimator bt used by Wilson and Shmida (1984) in their earlier study of plants in Mt Hermon. Beta sim (bsim) was calculated as follows: bsim

min(b; c) min(b; c) a

(1)

where for each two neighbouring 100 m elevation belts (X and Y): a the number of species observed in both X and Y, b the number of species in Y that are not observed in X, c the number of species in X that are not observed in Y. High values of bsim indicate that there were few species shared between two adjacent elevation belts (i.e. a high turnover rate). Rarity has been defined and estimated in the literature using many different approaches (Gaston 1994, Izco 1998). Because we were interested in the relative range size rarity within the mountain area, rather than in rarity over the whole distribution range, we used an estimate that quantifies the confinement of species to a small number of elevation belts in the mountain range. This approach has been used in many recent spatial ecology and large-scale conservation studies (Myers et al. 2000), and in Mt Hermon in a study by Levin et al. (2007). Range size rarity (RSR) was calculated for each elevation belt as the sum of the inverse of the range sizes of all the species occurring in it (Williams et al. 1996, Williams 2000): X RSR (1=Ci) (2) where Ci is the number of elevation belts occupies by species i. We estimated range size as the number of elevation belts in which the species occurred (of the seventeen 100 m altitude belts sampled). Remote sensing analyses We used a high spatial resolution QuickBird satellite image of the study area that was acquired during mid-spring (26 May 2004), when vegetation flowering is at its peak (Shmida 1977, 1980, Levin et al. 2007). The image has a spatial resolution of 2.4 m in its four spectral bands that cover the visible and near infrared spectral regions. We corrected the satellite image for atmospheric scattering and absorption and for topographic effects of shading using the atmospheric/topographic correction of multispectral sensors for rugged terrain as applied in ATCOR 3 ver. 7.1 (Richter 1998), which is considered a reliable model for atmospheric corrections (Ben-Dor et al. 2005). We used a Digital Elevation Model (DEM) obtained from the Survey of Israel at a spatial resolution of 25 m (Hall et al. 1999) for calculating the slope, aspect and the sky view factor (i.e. the

visible area of the sky as dependent upon the surrounding topography). We then calculated normalized difference vegetation index (NDVI), one of the earliest remotely sensed vegetation indices applied in the literature (Rouse et al. 1973, Tucker 1979). Its relationship with vegetation productivity is well established, and it is one of the most commonly used vegetation indices (Kerr and Ostrovsky 2003, Pettorelli et al. 2005, Levin et al. 2007), especially in biodiversity studies (Gillespie et al. 2008). NDVI was calculated as follows: NDVI (NIR R)=NIR R (3) where NIR reflectance in the near infrared band of an image pixel, R reflectance in the red band of an image pixel. Because NDVI is a ratio index shading effects have only a minor effect on it (Lillesand and Kiefer 1994). We also compared the results with three other remotely sensed vegetation indices designed for overcoming issues of variability in the soil background, atmospheric haze, and saturation of the NDVI in cases of dense vegetation, including the Soil Adjusted Vegetation Index (Huete 1988), the Enhanced Vegetation Index (Huete et al. 2002) and the percentage of tree cover (as in Levin et al. 2007). Because results were generally similar for the four satellite-derived vegetation indices and because correlations with diversity estimates were strongest for NDVI, we report here only the results for NDVI (detailed results for the three other indices are available from the authors upon request). In addition to calculating the mean and standard deviation (SD) of NDVI within the elevation belts, we quantified the change in NDVI among elevation belts along the elevation gradient. When examining changes in the values of NDVI along a gradient, multiple statistics can be used, such as the difference in NDVI among elevation belts and the ratio between adjacent belts (compare with Walker et al. 2003). Here, we initially calculated four different estimates for the change in NDVI along the elevation gradient. These include: rate of change in NDVI between each two neighbouring elevation belts (RC1): RC1X NDVIX =NDVIX 1

(4)

degree of change in NDVI between each two neighbouring elevation belts (DC1): DC1X NDVIX NDVIX 1

(5)

rate of change in NDVI between the elevation belt above and below the belt in focus (RC2): RC2X NDVIX 1 =NDVIX 1

(6)

degree of change in NDVI between the elevation belt above and below the belt in focus (DC2): DC2X NDVIX 1 NDVIX 1

(7)

where subscript x represents elevation belt x, subscript x 1 stands for the elevation belt adjacent to and above elevation belt x, and subscript x 1 stands for the elevation belt adjacent to and below elevation belt x. RC2 and DC2 were used to examine elevation gradients at a somewhat larger vertical distance, one that is still relevant for butterflies (200 vs 100 m). Negative values of 375

(mean and SD of NDVI per belt, DC1, DC2, RC1, RC2) and the residuals from the relationship between area and butterfly diversity, which was used as the dependent variable. This was calculated for the total area (log transformed) at each of the three spatial extents considered in this study.

Results Butterfly diversity along the elevation gradient Overall, in a total of 120 km of line transect sampling and 116 h of quadrat sampling, we recorded 10 513 individual butterflies belonging to 83 species and six butterfly families. Butterfly species richness showed a bimodal pattern with elevation, peaking between 1300 and 1500 m (48 species within each of the two 100 m belts) and between 1800 and 1900 m (46 species; Fig. 2a). Range size rarity showed local maxima at two intermediate elevation ranges (900 1000 and 1300 1400 m) and then increased sharply towards the highest elevation belts (Fig. 2b). Beta sim (bsim) diversity showed multiple peaks along the gradient, the largest of which was at the elevation range between 1900 and 50

(a)

376

40 35 30 25 20 9

(b)

8 7 RSR

6 5 4 3 2 0.14

(c)

0.12 0.10 0.08 0.06 0.04 0.02

2200

2100

2000

1900

1800

1700

1600

1500

1400

1300

1200

1100

900

1000

800

700

0.00 500

To examine the mid-domain effect, we ran 1000 Monte Carlo simulations without replacement. This was done in order to compare the observed species richness with the predictions of a null model simulating species richness based on empirical range sizes for each belt between 500 and 2200 m. The empirical range sizes were derived from the field data. We compared our results with 95% confidence intervals generated by the simulations. This was done based on analytical stochastic models (Colwell and Hurtt 1994) using the Mid-Domain-Null program, which takes into account the lower and upper elevations of the species ranges (McCain 2004). We calculated the linear regressions with the mean and SD of NDVI (within vegetation belts) or between adjacent belts (DC1, DC2, RC1 and RC2 among vegetation belts) as the independent variable (at the three spatial scales of our analysis) and each of the butterfly diversity estimates, including richness, bsim and range size rarity as the dependent variables. Variables were log-transformed when needed. To account for autocorrelation along the elevation gradient, we used the method developed by Dutilleul (1993), as applied in PASSAGE 1.1 (<www.passagesoft ware.net/>). To account for the potential effect of area of the different vegetation belts along the elevation gradient on the relationships and to examine whether the area of the elevation belts had a confounding effect on the observed relationships between NDVI and diversity estimates, we performed a partial correlation analysis, as well as a multiple regression analysis (using JMP 7.0 SAS Inst.). We examined whether statistical significance is maintained after calculating the relationship between each of the NDVI estimates

Beta SIM

Data analysis

Richness

600

DC indicate that NDVI values increase with elevation. We deliberately avoided using absolute values of DC, as the sign provided also an index for the directionality of changes (i.e. upwards or downwards) along the elevation gradient. This is especially important for butterflies, since some species perform directional hilltopping behaviour in which they ascend to mountain summits for the purpose of mating (Shields 1967, Alcock 1987, Ehrlich and Wheye 1988, Peâ&#x20AC;&#x2122;er et al. 2004). Because butterflies are usually not limited in their activity to a single quadrat, in order to calculate NDVI statistics comparing the different 100 m belts, we used three spatial extents (coverage areas) in the mountain. These included: 1) the total area within each of the 0.1 ha quadrats and a buffer zone of 5 m on both sides of the transects, within each of the 100 m elevation belts; 2) the total area of the SW facing aspect of each 100 m elevation belt in our study region. This was the aspect in which our sampling quadrats and transects were located; 3) the total area of the 100 m elevation belt in our study region (all aspects). At each of these spatial extents, we examined the relationship between the different butterfly diversity estimates and the mean NDVI, SD, RC1, RC2, DC1 and DC2. This enabled us to examine which spatial scale of NDVI best predicts changes in local butterfly diversity within and among elevation belts.

Elevation (m)

Figure 2. Changes in butterfly diversity along the elevation gradient in Mt Hermon showing: (a) species richness, (b) range size rarity (RSR) and (c) beta diversity (bsim).

2000 m (Fig. 2c). Range size rarity showed a significant positive relationship with bsim (R2 0.47, p B0.01, n 16), and a significant (weaker) positive relationship with species richness (R2 0.36, pB0.02, n 17). Changes in vegetation indices along the elevation gradient Mean NDVI showed a hump-shaped pattern along the elevation gradient, peaking between 900 and 1200 m and gradually declined to its minimum value at the highest elevation belts (Fig. 3). The SD of NDVI remained relatively constant up to 1200 m, above which it gradually declined with increasing elevation (Fig. 3). The values of DC1, DC2, RC1 and RC2, however, increased with elevation up to 1200 m and 1300 m, respectively, after which they declined (with a minor peak around 1850 m; Fig. 3; the changes in RC1 and RC2 with elevation are similar to those of DC1 and DC2, and are therefore not shown). Relationship between butterfly diversity and NDVI

0.7

Average DC1 DC2 Standard deviation

0.6 0.5 0.4 0.3 0.2 0.1 0 â&#x20AC;&#x201C;0.1

2200

2100

2000

1900

1800

1700

1600

1500

1400

1300

1200

1100

900

1000

800

700

600

â&#x20AC;&#x201C;0.2 500

NDVI values for 100 m elevation belts

Richness was strongly correlated with all variables estimating spatial heterogeneity in productivity, including DC1, DC2, RC1 and RC2 (Table 1, Fig. 4). However, the relationship between butterfly richness and mean and standard deviation (SD) of NDVI was weak and was not significant (Table 1). The relationship between richness and the rates of change in NDVI between elevation belts was strong, and remained significant after correcting for autocorrelation effects (Table 1). Unlike richness, range size rarity was significantly correlated with mean NDVI as well as with all the variables expressing the spatial heterogeneity in NDVI: SD, DC1, DC2, RC1, RC2 (Table 1, Fig. 5). Of the diversity estimates tested (richness, bsim and range size rarity), beta diversity (bsim) was in most cases the least strongly correlated with NDVI variables.

Elevation (m)

Figure 3. NDVI values (mean, standard deviation, DC1 and DC2) per 100 m elevation belt. DC1 stands for the difference between the mean of a given elevation belt and the elevation belt above it, while DC2 stands for the difference between the mean values of the elevation belts below and above each elevation belt in Mt Hermon.

When comparing NDVI estimates deriving from the three spatial extents examined (quadrats and transects, SW slope, and the area of the whole elevation belt within the study area), NDVI estimates deriving from the largest spatial extent (the entire elevation belt), were in most cases more strongly correlated with all butterfly diversity measures than those deriving from the quadrats and transects alone (Table 1). The area of the vegetation belts was not significantly correlated with butterfly richness at any spatial extent (Table 2). Butterfly rarity (RSR) was, however, significantly correlated with area (after log transformation) at the spatial scale of the SW aspect of the elevation belt (Table 2). However, in most cases, the correlations between NDVI and both richness and range size rarity remained statistically significant after taking into account the effect of the (log transformed) area of the elevation belts using a residual analysis (Table 3), as well as when performing a multiple regression analysis (Table 4). In most cases area was nonsignificant in the multiple regression analysis (Table 4). No mid-domain effect was detected. Species richness did not fall within the 95% prediction curves of the model based on the 1000 simulations of the Mid-Domain Null model (Fig. 6). When we included the effect of autocorrelation on the significance of the correlations, the significance of the regression model declined, as expected (Table 1). However, in some of the cases, the correlation remained significant between richness and spatial heterogeneity between the elevation belts (DC1, DC2, and RC1), also when spatial autocorrelation was taken into account.

Discussion We found that the NDVI was a strong predictor of butterfly richness along the elevation gradient in Mt Hermon, explaining up to 80% of the total variation in butterfly richness (Fig. 4). However, it was not the mean NDVI, but rather its variability among elevation belts, that best predicted butterfly richness within the elevation belts. Mean NDVI is considered a good surrogate for net primary productivity (Gillespie et al. 2008). Butterfly richness along the elevation gradient appears to be more strongly shaped by spatial heterogeneity in productivity than by productivity per se at the local spatial scale examined here. This suggests that the most commonly used remotely-sensed vegetation statistic, namely the mean NDVI is in some cases not the most efficient estimate if one aims to predict richness and diversity patterns along spatial gradients (e.g. elevation). In such cases, it may be more useful to study the spatial heterogeneity in NDVI. Here we show the importance of spatial heterogeneity in productivity at the small regional scale. The importance of heterogeneity in studies using remote sensing indices has been shown at much larger (e.g. continental) scales. For example, in their work on mammals, Kerr and Packer (1997) found that in the higher energy regions of North America, the best predictor of mammal richness was topographic heterogeneity and local variation in energy availability. At a regional scale, Atauri and de Lucio (2001) examined the relationships between landscape structure, land use and richness of birds, amphibians, reptiles and 377

Table 1. Pearson correlation coefficients between NDVI statistics calculated at three spatial extents and butterfly diversity estimates, including richness, range size rarity (RSR) and beta sim (bsim). The number sign (#) marks significance at the 0.05 level when taking into account the effect of autocorrelation. Correlation coefficients between each NDVI statistic and butterfly diversity estimates NDVI statistic

Spatial extent

Butterfly diversity

Entire 100 m elevation belt

SW aspect of the 100 m elevation belt

Quadrats transects within the 100 m elevation belt

0.07 0.68** 0.55* 0.17 0.76*** 0.64**

0.08 0.66** 0.41 0.25 0.63** 0.16

Average Log (average) Average Standard deviation Log (standard deviation) Standard deviation

Richness RSR Beta sim Richness RSR Beta sim

0.33 0.79*** 0.55* 0.46 0.86*** 0.67**

DC1

Richness Log RSR Beta sim

0.85*** 0.55* 0.04

0.78*** 0.56* 0.01

0.33 0.43 0.25

DC2

Richness Log RSR Beta sim

0.90*** 0.68** 0.03

0.81*** 0.70** 0.14

0.44 0.47 0.15

RC1

Richness Log RSR Beta sim

0.88*** 0.68** 0.07

0.81*** 0.68** 0.11

0.51* 0.28 0.05

RC2

Richness Log RSR Beta sim

0.90*** 0.81*** 0.18

0.79*** 0.82*** 0.27

0.59* 0.37 0.00

*pB0.05, **pB0.01, ***pB0.001.

(a)

9 8

Butterfly RSR

butterflies in a Mediterranean landscape (Madrid, Spain). They found that the response of species richness to land use heterogeneity varied depending on the group of species considered. The most important factor affecting bird and butterfly richness in their study was landscape heterogeneity, while other factors, such as the specific composition of land use, played a secondary role (Atauri and de Lucio 2001). In the montane ecosystem examined here, spatial heterogeneity in productivity between elevation belts explains butterfly richness better than mean productivity. What biological factors may lead to these results? One possibility is that the spatial heterogeneity in productivity estimated here represents the variety of habitat types available to the butterflies at local spatial scales and within relatively short distances (dozens of meters to kilometres). Such heterogeneity is particularly beneficial for adult

y = –4.866ln(x) + 1.1911

R2 = 0.6209

6 5 4 3 2 0.2

0.3

0.4

0.5

0.7

(b) 10 y = –5.18ln(x) –4.7109

R2 = 0.7441

40 y = 81.054x + 37.679

R2 = 0.8015

Butterfly RSR

Butterfly richness

0.6

Average NDVI

7 6 5 4 3

2 0.05

0.07

0.09

0.11

0.13

0.15

0.17

0.19

0.21

0.23

Standard deviation of NDVI 20 –0.2

–0.15

–0.1

–0.05

0.05

0.1

0.15

0.2

DC2

Figure 4. The relationship between DC2 (defined in eq. 7) and butterfly species richness in Mt Hermon.

378

Figure 5. (a) The relationship between average NDVI values within each 100 m elevation belt and butterfly range size rarity (RSR) in Mt Hermon; (b) the relationship between the standard deviation of NDVI within each 100 m elevation belt and butterfly range size rarity (RSR).

Table 2. Pearson correlation coefficients between the area of the three spatial extents examined in the study and butterfly diversity estimates, including richness, range size rarity (RSR) and beta sim (bsim). Correlation coefficients between area and butterfly diversity estimates

Spatial extent

Area

Butterfly diversity

Entire 100 m elevation belt

SW aspect of the 100 m elevation belt

Quadrats transects within the 100 m elevation belt

Area Log (area) Area Log (area) Area Log (area)

Richness Richness RSR RSR Beta sim Beta sim

0.39 0.23 0.32 0.55* 0.30 0.40

0.01 0.01 0.68** 0.74*** 0.52 0.55*

0.16 0.10 0.15 0.07 0.39 0.34

*pB0.05, **pB0.01, ***pB0.001.

butterflies, allowing them to utilize a variety of resources. While being highly dependent on particular host plants during larval development, adult butterflies often have different habitat requirements than those of the larvae (Benyamini 2002, Settele et al. 2009). Our results may also be partly shaped by the specific habitat and host plants of the butterfliesâ&#x20AC;&#x2122; larvae, but information about larval spatial distribution along the elevation gradient is not sufficient for analyzing such potential effect. Another explanation may be that the high spatial heterogeneity in productivity represents high turnover of habitats and changes in conditions, which allow more species to co-occur in transitional areas (Shmida and Wilson 1985). This supports findings from earlier studies, which suggest that areas of sharp environmental transition (ecotones) are especially rich both in species richness and Table 3. Partial correlations: Pearson correlation coefficients between NDVI statistics calculated at the spatial extent of the entire elevation belt and residuals of the butterfly diversity estimates (after predicting their values with area as the independent variable), including richness, range size rarity (RSR) and beta sim (bsim). Correlation coefficients between each NDVI statistic and residuals of butterfly diversity estimates

Spatial extent

NDVI statistic

Butterfly diversity estimate

Entire 100 m elevation belt

Average

Richness RSR Beta sim

0.60* 0.30 0.59*

Standard deviation

Richness RSR Beta sim

0.75*** 0.41 0.76***

DC1

Richness RSR Beta sim

0.75*** 0.71** 0.14

DC2

Richness RSR Beta sim

0.81*** 0.79*** 0.16

RC1

Richness RSR Beta sim

0.84*** 0.78*** 0.25

RC2

Richness RSR Beta sim

0.88*** 0.84*** 0.31

*pB0.05, **pB0.01, ***pB0.001.

in rare species because they serve as meeting areas between different communities and/or due to the unique environmental conditions found in ecotonal environments (reviewed in Kark and van Rensburg 2006). We found two peaks in beta diversity in the transition areas between Mt Hermonâ&#x20AC;&#x2122;s three vegetation belts. A peak in the betadiversity of plants was also found between 1200 and 1300 m on Mt Hermon, corresponding to the transition between a maquis and montane flora (Wilson and Shmida 1984). This supports the hypothesis that transition areas are zones of high turnover, where spatial heterogeneity is high (Shmida and Wilson 1985, Kark and Van Rensburg 2006). Our findings here also support the prediction and recent findings at continental and regional scales that areas with high turnover tend to show higher levels of rarity and local endemism (Kark et al. 2007, van Rensburg et al. 2009). Thus far, relatively few studies have examined the relationship between butterfly richness and satellite-derived vegetation indices that estimate productivity (but see Kerr et al. 2001). Some studies found relatively weak positive relationships, while others showed none (see Bailey et al. 2004, Seto et al. 2004 and references therein). Few studies have examined the relationship between butterfly richness, rarity and NDVI heterogeneity in space at local scales. Bailey et al. (2004) studied butterfly and bird richness and its correlation with NDVI heterogeneity using Simpsonâ&#x20AC;&#x2122;s diversity index. While heterogeneity in NDVI predicted the total species richness of birds (R2 0.75), no association occurred between NDVI heterogeneity and species richness of butterflies in any of the vagility classes tested in their work (Bailey et al. 2004). These included low vagility (an individual is likely to move on the order of dozens of meters in its lifetime); intermediate (an individual may move hundreds of meters); and high (an individual may move more than a kilometer) (Bailey et al. 2004). The authors suggested that for butterflies, NDVI may not be the best measure of environmental heterogeneity and that other measures (e.g. elevation) may be more appropriate (Bailey et al. 2004). However, the lack of relationship between spatial heterogeneity in productivity and butterfly richness may partly result from the estimates used to measure heterogeneity in NDVI, rather than from the lack of suitability of NDVI in predicting environmental heterogeneity. We propose that future studies should calculate spatial heterogeneity in productivity along the elevation gradient, quantifying changes in productivity between altitudinal gradients. This approach is more equivalent 379

Table 4. Multiple regression coefficients and the adjusted R2 between the area and NDVI variables as the independent variables and butterfly diversity (richness, RSR or bsim) at the spatial extent of the entire elevation belt. NDVI variable 2

R Coefficient of area NDVI variable R2 Coefficient of area

RC1

Standard deviation of NDVI within an elevation belt

Coefficient of NDVI

Richness

RSR

Beta-sim

0.77 *** 1.355 69.28 ***

0.71 *** 1.54 *** 13.32 ***

0.85 *** 0.24 ***

0.71 *** 0.01

0.43 ** 0.0002

42.14 ***

0.67 **

177.8 ***

0.15 0.02 0.12

*pB0.05, **pB0.01, ***pB0.001.

to beta diversity estimates used for estimating turnover of species in space, focusing on changes between neighbouring cells. The relatively weak correlations found here between productivity (mean NDVI) and butterfly richness may result from the fact that in some cases butterfly species are constrained by the identity of the plant species available, and particularly by the presence of specific host plants, rather than by total plant richness or vegetation cover (Kelly and Debinski 1998, Peâ&#x20AC;&#x2122;er and Settele 2008b). In such cases, productivity would not be a good predictor of butterfly richness, compared with, for instance, larger or more generalistic taxa that depend more directly on productivity (see Shochat 1999 for birds). Whereas plant richness above 2000 m amounted to B40% of the peak plant richness (found at about 1000 m; Levin et al. 2007), butterfly richness above 2000 m reached almost 80% of the peak butterfly richness (found at about 1400 m; Fig. 2a). We found higher levels of butterfly species richness at the high altitudes (above 2000 m), which are characterised by lower productivity, relatively low plant richness and harsher weather conditions (e.g. winds; Shmida 1977) compared with lower elevations (B900 m; Fig. 2a). This supports our knowledge that butterfly ranges are often limited by factors other than the diversity and distribution of plants (Dennis and Shreeve 1991, Dennis et al. 1991, Quinn et al. 1997, Hawkins and Porter 2003) or by temperature and rainfall (Pollard 1988). Observed species richness Lower 95%

Upper 95%

Species richness

60 50 40 30 20 10 0 400

600

800

1000 1200 1400 1600 1800 2000 2200

Elevation (m)

Figure 6. Mid domain analysis of the butterfly species richness. The upper and lower 95% confidence intervals were generated from 1000 Monte Carlo simulations.

380

Here, we examined the effect of the spatial extent at which the variability in NDVI was calculated. Interestingly, we found that the NDVI estimates that were based on the whole area of the elevation belt often provided better predictors of butterfly richness within each of the quadrats than the data from the quadrats themselves. It is not easy to conclude why this was found, but we can hypothesize that it results from the fact that scaling up when calculating NDVI better represented the relevant habitat as perceived by the butterflies (compare with Kumar et al. 2009). Because adult butterflies are mobile and move among host plants, which are not distributed in the area uniformly, this may better capture their preferences than local sampling of NDVI. Rowe and Lidgard (2009), following a detailed analysis of the effect of sampling methodology on patterns of elevation diversity, suggest that it may be advantageous to adopt more than a single spatial sampling method as empirical evidence because organisms relate to factors at a variety of spatial scales. This is especially true in the case of insects such as butterflies, in which the response of the different life stages may be quite different, with the adults being more mobile than the larval stages. Multiple factors affect species diversity in mountains and its spatial variation along the elevation gradient, such as climate, soil type, water availability, snow cover and topographic heterogeneity (reviewed in Grytnes and McCain 2007). Additional factors that have been studied are area and the mid-domain effect (Grytnes and McCain 2007). Given that many factors can affect changes in diversity along elevation gradient, we find it interesting that such a large portion of the variation in butterfly diversity, and especially in richness and rarity, was explained by satellite-derived vegetation indices. Mid-domain was not an important factor in this system. After removing the effect of area on the relationship between NDVI and the diversity indices examined, NDVI remained a strong predictor of butterfly diversity when using the degree of change among adjacent belts. This suggests that the variation among belts in their productivity is a good indicator (or even surrogate) of spatial heterogeneity or other processes that shape diversity patterns across the mountain. Range size rarity showed a strong negative correlation with both mean and SD of NDVI, whereas its relationship with the heterogeneity between elevation belts was somewhat weaker (Table 1). As we ascend the mountain into areas with lower vegetation cover, few or no trees, and with lower productivity, the proportion of rare species (those found in few elevation belts) increases. This is in

accordance with the prediction that the highest elevations, being more limited in area, more isolated from other areas and having conditions that require high levels of specialization (e.g. strong winds), will tend to show higher levels of rarity and endemism (this was also found for the range size rarity of plants on Mt Hermon; Levin et al. 2007). For locally rare species with smaller ranges in the mountain, both productivity and small-scale spatial heterogeneity in productivity are important. In our case, this may result from the fact that most of the locally-rare species are highaltitude butterflies that are not found elsewhere in Mt Hermon (Benyamini 2002, Pe’er and Benyamini 2008). Many butterfly species of the higher elevation belts reach the southern edge of their global distribution range on Mt Hermon, and comprise of peripheral populations (Benyamini 2002, Pe’er and Benyamini 2008). Lower NDVI values, indicating the lower productivity of highaltitude habitats and high local heterogeneity, are strongly correlated with the occurrence of unique species and high rarity. Species occurring in higher elevations are at higher risk in the face of climatic changes (Parmesan 2006 and references therein), as they occupy particularly small areas along mountain ranges. We should note, however, that our research was constrained to an elevation-range between 500 and 2200 m, and did not reach the summit of Mt Hermon, located in Syria at 2814 m. Thus, rarity patterns may be underestimated, as they are partly affected by the low number of elevation belts that were sampled above the tree-line ( 1850 1900 m). Indeed, Nogue´s-Bravo et al. (2008) have shown that the different sampling extents along the elevation gradient can affect the relationships found between richness and productivity, which may partly result from the effect of under-sampled rare species. Reports from the uppermost sections of the mountain indicate that the area near the peak of Mt Hermon actually harbours several additional rare butterfly species (Benyamini 1993, 2002). However, different land-use practices (e.g. cutting of trees and overgrazing) in parts of Mt Hermon located in Syria and in Lebanon, beyond the study area, likely lead to a reduction in butterfly diversity and rarity there. It would be interesting to collaborate across the political borders and sample the upper areas of Mt Hermon. In summary, we propose that estimates of local spatial heterogeneity in productivity based on remotely sensed vegetation indices may be useful in predicting butterfly richness along elevation gradients and should be examined in future studies. Such tools may be very useful in predicting and monitoring both plant richness (Levin et al. 2007) and animal richness in remote and inaccessible regions of high conservation importance. This is especially relevant in the face of the rapid climate changes and other environmental changes.

Acknowledgements We would like to thank Ben Inbar, Hava Goldstein and Didi Kaplan from the Israel Nature and Park Authority for their cooperation, Dubi Benyamini for advice and Catharine van Maanen, Rivka Peretz and Michal Hen-Gal for their help in conducting the field work. Christy McCain provided advice on the mid-domain simulations. This research was supported by the Israel Science Foundation (grant no. 740/04 to

SK). GP was partly supported by a Raymond and Jenine Bollag post-doctoral fellowship at the Hebrew Univ. of Jerusalem.

References Alcock, J. 1987. Leks and hilltopping in insects. J. Nat. Hist. 21: 319 328. Atauri, J. A. and de Lucio, J. V. 2001. The role of landscape structure in species richness distribution of birds, amphibians, reptiles and lepidopterans in Mediterranean landscapes. Landscape Ecol. 16: 147 159. Auerbach, M. and Shmida, A. 1993. Vegetation change along an altitudinal gradient on Mt-Hermon, Israel no evidence for discrete communities. J. Ecol. 81: 25 33. Bailey, S. A. et al. 2004. Primary productivity and species richness: relationships among functional guilds, residency groups and vagility classes at multiple spatial scales. Ecography 27: 207 217. Bawa, K. et al. 2002. Assessing biodiversity from space: an example from the Western Ghats, India. Conserv. Ecol. 6: 7, <www.consecol.org/vol6/iss2/art7>. Ben-Dor, E. et al. 2005. Comparison between six model-based methods to retrieve surface reflectance and water vapor content from hyperspectral data: a case study using synthetic AVIRIS data. Presented at International Conference on Optics and Optoelectronics ICOL 2005, Dehradun, India. Benyamini, D. 1993. The butterflies of Mt Hermon (Lepidoptera, Rhopalocera and Hesperiidae). Linn. Belg. 14: 167 203. Benyamini, D. 2002. A field guide to the butterflies of Israel; including butterflies of Mt Hermon, Sinai and Jordan. Keter Publ. House, Jerusalem. Blair, R. B. 1999. Birds and butterflies along an urban gradient: surrogate taxa for assessing biodiversity? Ecol. Appl. 9: 164 170. Colwell, R. K. 2006. EstimateS: statistical estimation of species richness and shared species from samples. Version 8.0, user’s guide and application. <http://purl.oclc.org/estimates>. Colwell, R. K. and Hurtt, G. C. 1994. Nonbiological gradients in species richness and a spurious Rapoport effect. Am. Nat. 144: 570 595. Colwell, R. K. and Lees, D. C. 2000. The mid-domain effect: geometric constraints on the geography of species richness. Trends Ecol. Evol. 15: 70 76. Colwell, R. K. et al. 2004. The mid-domain effect and species richness patterns: what have we learned so far? Am. Nat. 163: E1 E23. Dennis, R. L. H. and Shreeve, T. G. 1991. Climatic change and the British butterfly fauna: opportunities and constraints. Biol. Conserv. 55: 1 16. Dennis, R. L. H. et al. 1991. A multivariate approach to the determination of faunal structures among European butterfly species (Lepidoptera, Rhopalocera). Zool. J. Linn. Soc. 101: 1 49. Dutilleul, P. 1993. Modifying the t test for assessing the correlation between two spatial processes. Biometrics 49: 305 314. Ehrlich, P. R. and Wheye, A. 1988. Hilltopping butterflies revisited. Am. Nat. 132: 460 461. Fleishman, E. et al. 2000. Upsides and downsides: contrasting topographic gradients in species richness and associated scenarios for climate change. J. Biogeogr. 27: 1209 1219. Foody, G. M. and Cutler, M. E. J. 2006. Mapping the species richness and composition of tropical forests from remotely sensed data with neural networks. Ecol. Model. 195: 37 42. Gaston, K. J. 1994. Rarity. Chapman and Hall. Gillespie, T. W. et al. 2008. Measuring and modelling biodiversity from space. Prog. Phys. Geogr. 32: 203 221.

381

Grill, A. et al. 2005. Butterfly, spider, and plant communities in different land-use types in Sardinia, Italy. Biodivers. Conserv. 14: 1281 1300. Grytnes, J. A. and McCain, C. M. 2007. Elevation patterns in species richness. In: Levin, S. (ed.), Encyclopedia of biodiversity. Elsevier, pp. 1 8. Hall, J. K. et al. 1999. Test of the accuracy of the DEM of Israel Geological Survey of Israel, Jerusalem. Hawkins, B. A. and Porter, E. E. 2003. Does herbivore diversity depend on plant diversity? The case of California butterflies. Am. Nat. 161: 40 49. Huete, A. R. 1988. A soil-adjusted vegetation index (SAVI). Remote Sens. Environ. 25: 295 309. Huete, A. et al. 2002. Overview of the radiometric and biophysical performance of the MODIS vegetation indices. Remote Sens. Environ. 83: 195 213. Izco, J. 1998. Types of rarity of plant communities. J. Veg. Sci. 9: 641 646. Jetz, W. and Rahbek, C. 2002. Geographic range size and determinants of avian species richness. Science 297: 1548 1551. Kark, S. and van Rensburg, B. J. 2006. Ecotones: marginal or central areas of transition? Isr. J. Ecol. Evol. 52: 29 53. Kark, S. et al. 2007. The role of transitional areas as avian biodiversity centers. Global Ecol. Biogeogr. 16: 187 196. Kark, S. et al. 2008. Global environmental priorities: making sense of remote sensing. Trends Ecol. Evol. 23: 181 182. Kelly, L. and Debinski, D. M. 1998. Relationship of host plant density to size and abundance of the regal fritillary Speyeria idalia Drury (Nymphalidae). J. Lepidopterists’ Soc. 52: 262 276. Kerr, J. T. and Packer, L. 1997. Habitat heterogeneity as a determinant of mammal species richness in high-energy regions. Nature 385: 252 254. Kerr, J. T. and Ostrovsky, M. 2003. From space to species: ecological applications for remote sensing. Trends Ecol. Evol. 18: 205 299. Kerr, J. T. et al. 2001. Remotely sensed habitat diversity predicts butterfly species richness and community similarity in Canada. Proc. Natl Acad. Sci. USA 98: 11365 11370. Kim, K. C. 1993. Biodiversity, conservation and inventory why insects matter. Biodivers. Conserv. 2: 191 214. Koleff, P. et al. 2003. Measuring beta diversity for presenceabsence data. J. Anim. Ecol. 72: 367 382. Kremen, C. 1992. Assessing the indicator properties of species assemblages for natural areas monitoring. Ecol. Appl. 2: 203 217. Ku¨ hn, E. et al. 2008. Getting the public involved in butterfly conservation: lessons learned from a new monitoring scheme in Germany. Isr. J. Ecol. Evol. 54: 89 103. Kumar, S. et al. 2009. Effects of spatial heterogeneity on butterfly species richness in Rocky Mountain National Park, CO, USA. Biodivers. Conserv. 18: 739 763. Lennon, J. J. et al. 2001. The geographical structure of British bird distributions: diversity, spatial turnover and scale. J. Anim. Ecol. 70: 966 979. Levin, N. et al. 2007. Predicting mountain plant richness and rarity from space using satellite-derived vegetation indices. Divers. Distrib. 13: 692 703. Lillesand, T. M. and Kiefer, R. W. 1994. Remote sensing and image interpretation. Wiley. Lomolino, M. V. 2001. Elevation gradients of species-density: historical and prospective views. Global Ecol. Biogeogr. 10: 3 13. McCain, C. M. 2004. The mid-domain effect applied to elevation gradients: species richness of small mammals in Costa Rica. J. Biogeogr. 31: 19 31.

382

Myers, N. et al. 2000. Biodiversity hotspots for conservation priorities. Nature 403: 853 858. Nogue´s-Bravo, D. et al. 2008. Scale effects and human impact on the elevation species richness gradients. Nature 453: 216 219. Nowicki, P. et al. 2008. Butterfly monitoring methods: the ideal and the real world. Isr. J. Ecol. Evol. 54: 69 88. Oindo, B. O. 2002. Predicting mammal species richness and abundance using multi-temporal NDVI. Photogramm. Eng. Remote Sens. 68: 623 629. Parmesan, C. 2006. Ecological and evolutionary responses to recent climate change. Annu. Rev. Ecol. Evol. Syst. 37: 637 669. Parmesan, C. et al. 1999. Poleward shifts in geographical ranges of butterfly species associated with regional warming. Nature 399: 579 583. Pe’er, G. and Benyamini, D. 2008. A template for publishing the ‘‘conservation chain’’ from problem identification to practical action, exemplified through the campaign for butterfly protection in Israel. Isr. J. Ecol. Evol. 54: 19 39. Pe’er, G. and Settele, J. 2008a. The rare butterfly Tomares nesimachus (Lycaenidae) as a bioinicator for pollination services and ecosystem functioning in the north of Israel. Isr. J. Ecol. Evol. 54: 111 136. Pe’er, G. and Settele, J. 2008b. Butterflies in and for conservation: trends and prospects. Isr. J. Ecol. Evol. 54: 7 17. Pe’er, G. et al. 2004. Response to topography in a hilltopping butterfly and implications for modeling nonrandom dispersal. Anim. Behav. 68: 825 839. Pettorelli, N. et al. 2005. Using the satellite-derived NDVI to assess ecological responses to environmental change. Trends Ecol. Evol. 20: 503 510. Pin Koh, L. and Sodhi, N. S. 2005. Importance of reserves, fragments, and parks for butterfly conservation in a tropical urban landscape. Ecol. Appl. 14: 1695 1708. Pollard, E. 1977. A method for assessing changes in the abundance of butterflies. Biol. Conserv. 12: 115 134. Pollard, E. 1988. Temperature, rainfall and butterfly numbers. J. Appl. Ecol. 25: 819 828. Pollard, E. and Yates, T. 1993. Monitoring butterflies for ecology and conservation. Chapman and Hall. Quinn, R. M. et al. 1997. Abundance range size relationships of macrolepidoptera in Britain: the effects of taxonomy and life history variables. Ecol. Entomol. 22: 453 461. Rahbek, C. 1995. The elevation gradient of species richness: a uniform pattern? Ecography 18: 200 205. Rahbek, C. 2005. The role of spatial scale and the perception of large-scale species-richness patterns. Ecol. Lett. 8: 224 239. Richter, R. 1998. Correction of satellite imagery over mountainous terrain. Appl. Optics 37: 4004 4015. Rocchini, D. et al. 2010. Remotely sensed spectral heterogeneity as a proxy of species diversity: recent advances and open challenges. Ecol. Inform. 5: 318 329. Rosenzweig, M. 1995. Species diversity in space and time. Cambridge Univ. Press. Rouse, J. W. et al. 1973. Monitoring vegetation systems in the Great Plains with ERTS. In: Freden, S. C. et al. (eds), Third Earth Resources Technology Satellite-1 Symposium Technical presentations, section A, Vol. I. National Aeronautics and Space Administration, Washington, DC, pp. 309 317. Rowe, R. J. and Lidgard, S. 2009. Elevation gradients and species richness: do methods change pattern perception? Global Ecol. Biogeogr. 18: 163 177. Samways, M. J. 1989. Insect conservation and the disturbance landscape. Agric. Ecosyst. Environ. 27: 183 194. Seto, K. C. et al. 2004. Linking spatial patterns of bird and butterfly species richness with Landsat TM derived NDVI. Int. J. Remote Sens. 25: 4309 4324.

Settele, J. et al. 2009. Ecology of butterflies in Europe. Cambridge Univ. Press. Shields, O. 1967. Hilltopping: an ecological study of summit congregation behavior of butterflies on a southern California hill. J. Res. Lepidoptera 6: 69 178. Shmida, A. 1977. A quantitative analysis of the Tragacanthic vegetation of Mt Hermon and its relation to environmental factors. PhD thesis, The Hebrew Univ. of Jerusalem. Shmida, A. 1980. Vegetation and flora of Mt Hermon. In: Shmida, A. and Livne, M. (eds), Mt Hermon nature and landscape. Kibbutz Hameouchad, Tel Aviv, pp. 97 158. Shmida, A. and Wilson, M. V. 1985. Biological determinants of species diversity. J. Biogeogr. 12: 1 20. Shochat, E. 1999. The effect of scrub fragmentation by planted woods on bird communities in the northern Negev. PhD thesis, Ben-Gurion Univ. of the Negev, Beer-Sheva. Su, J. C. et al. 2004. Beyond species richness: community similarity as a measure of cross-taxon congruence for coarsefilter conservation. Conserv. Biol. 18: 167 173. Thomas, J. A. 2005. Monitoring change in the abundance and distribution of insects using butterflies and other indicator groups. Phil. Trans. R. Soc. B 360: 339 357. Thomas, J. A. et al. 2004. Comparative losses of British butterflies, birds and plants and the global extinction crisis. Science 303: 1879 1881.

Tucker, C. J. 1979. Red and photographic infrared linear combinations for monitoring vegetation. Remote Sens. Environ. 8: 127 150. Tucker, C. J. and Sellers, P. J. 1986. Satellite remote-sensing of primary production. Int. J. Remote Sens. 7: 1395 1416. van Rensburg, B. J. et al. 2009. Spatial congruence between ecotones and range restricted species: implications for conservation biogeography at the national scale. Divers. Distrib. 15: 379 389. van Swaay, C. A. M. et al. 1997. Monitoring butterflies in the Netherlands and Flanders: the first results. J. Insect Conserv. 1: 81 87. Walker, S. et al. 2003. Properties of ecotones: evidence from five ecotones objectively determined from a coastal vegetation gradient. J. Veg. Sci. 14: 579 590. Williams, P. H. 2000. Some properties of rarity scores used in site quality assessment. Br. J. Entomol. Nat. Hist. 13: 73 86. Williams, P. H. et al. 1996. A comparison of richness hotspots, rarity hotspots, and complementary areas for conserving diversity of British birds. Conserv. Biol. 10: 144 174. Wilson, M. V. and Shmida, A. 1984. Measuring beta diversity with presence absence data. J. Ecol. 72: 1055 1064.

383

Ecography 34: 364 371, 2011 doi: 10.1111/j.1600-0587.2010.06629.x # 2011 The Authors. Ecography # 2011 Ecography Subject Editor: Ken Kozak. Accepted 16 August 2010

Elevational gradients in phylogenetic structure of ant communities reveal the interplay of biotic and abiotic constraints on diversity Antonin Machac, Milan Janda, Robert R. Dunn and Nathan J. Sanders A. Machac (A.Machac@email.cz), Dept of Zoology, Faculty of Science, Univ. of South Bohemia, Branisovska 31, CZ-37005 Ceske Budejovice, Czech Republic. M. Janda, Biology Center, Czech Academy of Sciences and Faculty of Sciences Branisovska 31, CZ-37005 Ceske Budejovice, Czech Republic. (Present address of MJ: Museum of Comparative Zoology, Harvard Univ., Cambridge, MA 02138, USA.) R. R. Dunn, Dept of Biology and Keck Center for Behavioral Biology, North Carolina State Univ., Raleigh, NC 27695, USA. N. J. Sanders, Dept of Ecology and Evolutionary Biology, Univ. of Tennessee 28, Knoxville, TN 37996, USA, and Center for Macroecology, Evolution and Climate, Dept of Biology, Univ. of Copenhagen, DK-2100 Copenhagen, Denmark.

A central focus of ecology and biogeography is to determine the factors that govern spatial variation in biodiversity. Here, we examined patterns of ant diversity along climatic gradients in three temperate montane systems: Great Smoky Mountains National Park (USA), Chiricahua Mountains (USA), and Vorarlberg (Austria). To identify the factors which potentially shape these elevational diversity gradients, we analyzed patterns of community phylogenetic structure (i.e. the evolutionary relationships among species coexisting in local communities). We found that species at low-elevation sites tended to be evenly dispersed across phylogeny, suggesting that these communities are structured by interspecific competition. In contrast, species occurring at high-elevation sites tended to be more closely related than expected by chance, implying that these communities are structured primarily by environmental filtering caused by low temperatures. Taken together, the results of our study highlight the potential role of niche constraints, environmental temperature, and competition in shaping broad-scale diversity gradients. We conclude that phylogenetic structure indeed accounts for some variation in species density, yet it does not entirely explain why temperature and species density are correlated.

A fundamental pattern in biogeography is that both the number of species in a local community (i.e. species density; Gotelli and Colwell 2001) and the composition of communities vary, often systematically, along elevational gradients (Rahbek 2005, McCain 2009). The question, of course, is what drives that variation? Despite a growing number of ecological and evolutionary hypotheses to explain elevational diversity gradients (Sanders 2002, Colwell et al. 2004, Smith et al. 2007, Wiens et al. 2007, Kozak and Wiens 2010, and citations therein), the causes remain poorly understood. One promising approach to infer the underlying processes shaping spatial variation in community composition is the use of phylogenetic tools (Cavender-Bares et al. 2009). Modern approaches build on the earlier use of taxonomic similarity to understand the assembly of communities (Elton 1946, Simberloff 1970). For example, if species within the same genus are more functionally and ecologically similar to one another than distantly related species, observed genus-to-species ratios that are higher than expected (i.e. when compared to a null model) might indicate that competition structures communities. The availability of well-sampled phylogenies has allowed the development of a framework (Webb 2000, Webb et al. 364

2002) which combines the approach of Elton (1946) with the information now available from phylogenetic trees. The framework allows inferring the potential mechanisms that underlie community phylogenetic structure, i.e. phylogenetic relationships among species coexisting within a community. After the actual phylogenetic structure of local communities is assessed, it is compared with structure of communities randomly assembled (i.e. following a specific null model) from the larger, regional species pool. Webb et al. (2002) argued that if the species occurring in a local community are clustered in the phylogeny (i.e. more phylogenetically related than in the null model communities) then the underlying cause of structure is likely to be environmental filtering on shared physiological tolerances, assuming that niches are conserved (Webb et al. 2002, Losos 2008). Alternatively, when species are overdispersed in the phylogeny (i.e. species are less related than in the null model communities) then either interspecific competition or trait convergence is implicated as the structuring force. A lack of a phylogenetic structuring suggests that neutral processes shape the community (Kembel and Hubbell 2006). However, it is worth noting as a caveat that other processes have also been proposed to lead to patterns similar to clustering/overdispersion (e.g. density-dependent

interactions, facilitation during succession; Cavender-Bares et al. 2009). A growing number of studies have used community phylogenetic approaches to better understand spatial variation in community composition (Stevens 2006, Emerson and Gillespie 2008, Graham and Fine 2008, Algar et al. 2009, Cavender-Bares et al. 2009, Vamosi et al. 2009). Yet, only two studies, to our knowledge, have tested whether the phylogenetic structure of local communities might vary along elevational gradients or whether the drivers of diversity along elevational diversity gradients can be inferred by employing a community phylogenetics perspective (Bryant et al. 2008, Graham et al. 2009). Bryant et al. (2008) examined the phylogenetic structure of microbial and plant communities at five sites along a single elevational gradient in the Rocky Mountains, USA, and found that the microbial communities tended to be phylogenetically clustered throughout the entire elevational gradient, but the plant communities were overdispersed at higher elevations. Graham et al. (2009) examined the phylogenetic structure of 189 hummingbird communities in the Andes in Ecuador and found that communities were overdispersed in lowlands, suggesting an important role of interspecific competition. The community phylogenetics approach hinges on an important assumption that closely related species share similar traits and functions. This assumption has been called phylogenetic conservatism, niche conservatism, or evolutionary stability (Losos 2008). Importantly, neither the Bryant et al. (2008) nor the Graham et al. (2009) study tested for phylogenetic conservatism. Moreover, the Bryant et al. (2008) and the Graham et al. (2009) focused on single elevational gradients such that the generality of the patterns they documented is hard to assess. In this study, we examine patterns of ant species density and community phylogenetic structure along three elevational gradients. We tested two predictions: 1) communities at high-elevation sites would be phylogenetically clustered, as would be expected if traits are conserved and only closely related species of a subset of lineages possessed the traits which allowed them to persist at cold, highelevation sites, and 2) communities at low-elevation sites would be phylogenetically overdispersed in the phylogeny, as would be expected if interspecific competition rather than environmental filtering shaped the composition of local communities. Finally, we assessed whether the elevational pattern in phylogenetic structure is sufficient to explain patterns in species density, or whether environmental gradients have effects on species density above and beyond the effects of phylogenetic structure.

Methods The data We obtained data on the identities and occurrences of species within local communities from two published studies (Chiricahua Mountains, USA: Andersen 1997; Vorarlberg Mountains, Austria: Glaser 2006) and our own work (southern Appalachian Mountains, USA: Sanders et al. 2007). Importantly, each of the datasets consists of samples from local communities along extensive elevational gradients (Table 1); the data are not interpolated ranges or derived from niche models. For detailed information on geography of montane systems and sampled sites, see Supplementary materials Appendix 1. Constructing phylogenies We constructed three phylogenies, one for each of the montane systems, based on published genus-level phylogenies (Brady et al. 2006, Moreau et al. 2006). We adopted the molecular datasets from these studies from the TreeBase database <www.treebase.org>. Nine of the 175 species considered here lacked species- and genus-level molecular data. In these few cases (5% of all species in this study), species were substituted with closely related taxa with relationships derived from Boltonâ&#x20AC;&#x2122;s (2003) classification. We extended the molecular dataset using 80 additional sequences (using the same genes as in the Brady et al. (2006) and Moreau et al. (2006) studies) available for particular species in GenBank in order to incorporate within-genus variability and to resolve some of the genuslevel polytomies (especially in the genera Pheidole and Camponotus). These additional sequences, their GenBank codes, as well as the substituted taxa are listed in the Supplementary materials Appendix 2. We aligned the edited sequences in MAFFT, ver. 6 (Katoh et al. 2002). To reconstruct the phylogenies, we employed a maximum likelihood approach with topology constraint in PAUP 4.0 (Swofford 1993). The tree topology, on which molecular data were forced, corresponded with the genus-level phylogeny of Bolton (2003), Brady et al. (2006), and Moreau et al. (2006). We estimated branch lengths on the basis of substitution rates in a combined molecular dataset. For more details, see Supplementary materials. Assessing phylogenetic structure of communities Prior to examining the phylogenetic structure of communities, we tested for niche conservatism/phylogenetic conservatism. Based on our understanding of the natural

Table 1. Location, number of sites, and elevational span sampled for the community data used in the analyses. The entire elevational extent of Chiracahua Mts: 1100 2900 m, Vorarlberg Mts: 350 3000 m, Smoky Mts: 250 2000 m. More information on geography of the montane systems and sites sampled is given in the Supplementary material Appendix 1. Author

Montane system

Location

No. of sites

Elevation range (m)

Andersen (1997) Glaser (2006) Sanders et al. (2007)

Chiricahua Mts Vorarlberg Smoky Mts

Arizona, USA Austria Tennessee/N Carolina, USA

9 sites 18 sites 29 sites

1400 2600 400 2100 379 1828

365

history of ants and previous studies, we tested for conservatism in three traits: habitat associations (woodlands, shrublands, meadows and grasslands), nest site (in soil, under rocks, mounds and ground nests, rotting wood, canopy and trees), and worker size (measured as the Weber’s length of thorax). Each of these three suites of traits represents a significant axis of the ecological niche of ants and can be related to interspecific competition (worker size, nest site) or environmental tolerance (habitat association, nest site). We employed the random tree-length distribution algorithm for discrete traits (Cubo et al. 2005) that randomly permutes taxa (and their character values) along the phylogeny, while holding the topology as well as the branch lengths constant. Each character is mapped on the phylogeny through the maximum parsimony procedure. Afterwards, the number of character steps along the actual phylogeny is contrasted with the distribution of the number of steps in the 10 000 randomly constructed phylogenies. In the case of continuous characters, we used squared-length parsimony (Cubo et al. 2005). We performed these analyses in Mesquite 2.7 (Maddison and Maddison 2002). Our results were not affected by the bias potentially introduced by phylogenetic signal in occurrence frequencies (i.e. closely related species appear in many communities as present) (Kembel 2009) because we employed a null model that takes into consideration the prevalence of species in the communities and samples them accordingly (see below, Gotelli 2000). After testing for niche/phylogenetic conservatism, we estimated the phylogenetic structure of each community from the three montane systems using two indices: mean phylogenetic distance (MPD) and mean nearest neighbor distance (MNND; Webb et al. 2002). MPD is an estimate of the average phylogenetic relatedness (on basis of branch lengths) between all possible pairs of taxa in a local community. MNND, in contrast, is an estimate of the mean phylogenetic relatedness between each taxon in a local community and its nearest relative. We then calculated standardized NRI and NTI indices. The NRI and NTI describe the difference between average phylogenetic distances (MPD and MNND, respectively) in the observed and randomly generated null communities, standardized by the standard deviation of phylogenetic distances in the null communities (Webb et al. 2008). We used R 2.8 to calculate NRI and NTI (Kembel et al. 2009). Since values of NRI and NTI were highly correlated (r 0.837, pB0.001), we report only NRI values in subsequent analyses. To assess whether the observed NRI values differed significantly from zero, we compared them to NRI values of null communities generated by Gotelli’s swap algorithm (Gotelli 2000); i.e. the occurrence matrix is randomized holding the number of species per sample and the frequency of occurrence of each species across samples constant. The phylogeny used for the calculation of NRI was not fully resolved and branch-length estimates were not available for all of the taxa. Therefore, we examined the impact of the phylogeny’s resolution (defined as branch lengths availability) on NRI by estimating three different distances: species-level distances, genus-level distances, and simple Grafen’s (1989) distances based on the tree topology. We then calculated new NRI values using these 366

distances for each of the three montane systems and mutually compared them. Environmental variables To examine the relationship between ant species density and climate, we extracted information on annual precipitation and annual mean temperature for each community in each montane region from the WorldClim v1.4 database (<www.worldclim.org>; Hijmans et al. 2005) using ArcView GIS (ver. 3.2, Esri 1992 2000; ESRI, Redlands, CA). WorldClim data pose some problems, especially in montane systems, because the resolution of the data is 1 km2. Considerable variation in temperature can occur within one square kilometer, especially in montane systems. As a check of the potential bias of using WorldClim data, we examined whether mean annual temperature data obtained from WorldClim were correlated with mean annual temperature data obtained from measurements of temperature from dataloggers arranged along the elevational gradient in the Smokies. Ideally, we would have measured temperature and precipitation data in each of the montane systems. However, because the WorldClim temperature data were correlated with the measured temperature data (r 0.998, pB0.001), and measured climate data were unavailable for two of the three gradients, we instead use WorldClim data. We also note that such an approach is common to other studies of elevational diversity gradients (McCain 2009) such that our work should be directly comparable. We chose these focal environmental variables because they are often strongly correlated with ant species density (Kaspari et al. 2000, Sanders et al. 2007, Dunn et al. 2009). Analyses We related elevation and the climate variables (mean annual temperature, annual precipitation) to species density (the number of species occurring in a local community) and to phylogenetic structure (NRI) of local communities using linear mixed-effect models. In the model, identity of a montane system was treated as a random effect, and elevation, mean annual temperature, annual precipitation, and all their combinations as explanatory variables. We used the maximum likelihood procedure to fit each model, and we compared those models via Bayesian information criterion (BIC). As temperature appeared to be the best predictor of both species density and community phylogenetic structure, we conducted further analyses to tease apart the mechanisms linking these variables. The effects of temperature on NRI and species density could operate in one of two ways. First, temperature could influence phylogenetic structure and phylogenetic structure could in turn influence species density. In this scenario, species density patterns are simply a consequence of species of a few clades possessing the traits necessary to persist in harsh conditions, such as the cold. Or second, temperature could influence species density via mechanisms independent of phylogenetic structure. Any of a variety of effects are possible, including effects on speciation rates (Davies et al. 2004) or abundance mediated

effects on extinction (Willig et al. 2003). To distinguish between these possibilities, we assessed whether the effect of phylogenetic structure (estimated as NRI) on species density was significant even when the effect of temperature was already included as an explanatory variable in the model predicting species density. This outcome would indicate that phylogenetic structure has a direct effect on species density. Alternatively, if phylogenetic structure is not related to species density after temperature is included in the model, the result implies that phylogenetic structure may influence patterns of species density, but is insufficient as a complete explanation for them. All the analyses were conducted in R 2.8 (Pinheiro and Bates 2000). We note that spatial autocorrelation within montane systems can inflate type I errors in statistical tests. However, because interpreting the coefficients from spatial regression can be challenging at best (Bini et al. 2009), we do not use spatial regression techniques in these analyses (e.g. SAM Rangel et al. 2006) and instead rely on BIC and R2 values as estimates of goodness of fit.

Results The three elevational gradients we considered consisted of 56 local ant communities with 175 ant species from seven subfamilies (Supplementary material Appendix 3). Along each of the gradients, ant species density decreased with elevation (Fig. 1). Bayesian information criterion indicated that the best environmental predictor of ant species density was annual mean temperature (BIC 364.04, positive relationship; Table 2). Only slightly less plausible were the models including elevation (DBIC 0.11; Table 2) and temperature precipitation (DBIC 0.64; Table 2). We found strong evidence for niche conservatism for each of the three traits we examined. The evolutionary stability of niches (represented by habitat associations, nest site, and worker size) was consistent among the montane

Species density

(a)

(b)

(c)

30 15

10 10

5 500

Net relatedness index

systems for each of the examined traits (10 000 randomizations, p B0.05) (Table 4). In other words, not only were the traits examined here phylogenetically conserved, they were conserved everywhere. NRI was correlated with temperature (BIC 338.95, R2 0.36; Table 3); the correlation was negative in each of the three montane systems (Fig. 2). This would be expected if environmental stress (due to lower temperatures at higher elevations) acted as a filter on lineages at high elevations and competition structured communities at low elevations. Besides temperature, the next best model of NRI with only a minor difference in BIC comprised elevation (DBIC 0.68; Table 3). Low-elevation communities tended to be significantly overdispersed (4 sites), whereas communities at higher elevations tended to be significantly clustered (7 sites) (Fig. 2). Ant species density was significantly and positively correlated both with temperature (p B0.001, R2 0.55) and NRI (p 0.002, R2 0.17) in independent models (above). However, once temperature had been added in the model of species density, the contribution of NRI became insignificant (p 0.76, R2NRI B0.01). Conversely, even if the model of species density already comprised NRI, the effect of temperature remained significant (p B0.001, R2temp 0.44). These outcomes suggest that both the species density and community phylogenetic structure are mutually independent products of temperature variation. The estimates of NRI are robust to phylogenetic resolution for the Smoky Mountains and Vorarlberg ant communities (Table 5). NRI for the communities from the Chiricahua Mountains varied with phylogenetic resolution, however, perhaps because there were only 9 sites sampled in the Chiricahuas and few genera were monotypic (as opposed to the Smoky Mountains and Vorarlberg) such that it was possible for polytomies within genera to have a greater effect. It could be argued that the species level phylogeny for the Smoky Mountains and Vorarlberg (due to its lower resolution) approaches the genus-level phylogeny; thus, resulting in a tight correlation between the respective NRI values. To avoid this artifact and examine

1000

1500

5 500

1000 1500 2000

1400

1800

2200

2600

1400

1800

2200

2600

1.0 4 2

0.0

–1

0 –2

–1.0

–3 500

1000

1500

500

1000 1500 2000 Elevation (m)

Figure 1. Community characteristics for (a) Smoky Mountains, (b) Vorarlberg, (c) Chiricahua Mountains. Species density (i.e. number of species in a local community) and net relatedness index (NRI) are plotted against elevation. Each point is a site sampled for ants.

367

Table 2. Models of ant species density. The most parsimonious model was identified via Bayesian information criterion (BIC). Abbreviations refer to temperature (Temp), precipitation (Precip), and elevation (Elev).

Null model Elevation Temperature Precipitation Temp Precip Temp Elev Precip Elev Precip Elev Temp

BIC

logLik

403.551 364.151 364.039 393.806 364.681 367.173 365.948 368.110

195.765 174.061 174.005 188.885 172.322 173.568 172.956 172.033

* 51 51 51 50 50 50 49

* 0.546 0.547 0.221 0.574 0.554 0.564 0.578

the robustness of NRI thoroughly, we additionally correlated genus-level NRIs and topology derived NRIs for Smoky Mts and Vorarlberg. Still, the NRIs were highly correlated (Smoky Mts: F1, 26 335.400, p B0.001, r 0.962; Vorarlberg: F1, 16 43.620, pB0.001, r 0.846) such that the incomplete phylogenetic resolution for a handful of species in this study is likely to have only marginal effects on our broad results. In addition, provided that the polytomies in some of the genera (e.g. Pheidole, Camponotus; Pie and Traniello 2007) represent rapid diversification events, additional resolution in these taxa should not affect the NRI estimates.

Discussion We found that, across three elevational gradients in North America and Europe, ant species density is positively related to temperature. Such a result is not surprising. Numerous other studies have documented that species density is positively correlated with temperature (Hawkins et al. 2003). Temperature is often correlated with diversity, not just in ants but in many taxa, though the mechanisms that link temperature to diversity have been a topic of much discussion (Clarke and Gaston 2006, Storch et al. 2006, Hawkins et al. 2007, Hessen et al. 2007). Our results indicate that at high elevations and in cooler conditions, there are fewer species than in warmer, low-elevation sites. Those high-elevation species tend to come from fewer lineages than would be expected by chance. Only a few lineages appear to have traits that allow them to persist in colder, high elevation conditions. Table 3. Models of community phylogenetic structure (represented by NRI values). The best model was identified via Bayesian information criterion (BIC). Abbreviations refer to temperature (Temp), precipitation (Precip), and elevation (Elev).

Null model Elevation Temperature Precipitation Temp Precip Temp Elev Precip Elev Precip Elev Temp

368

BIC

logLik

359.149 339.633 338.953 347.040 341.926 342.618 342.360 345.321

173.564 161.802 161.462 165.505 160.945 161.291 161.162 160.639

* 51 51 51 50 50 50 49

* 0.348 0.356 0.254 0.368 0.360 0.363 0.375

Table 4. Examination of the phylogenetic signal in relevant life history traits. If the number of parsimony character steps or ‘‘squaredlength of the character’’ in case of continuous traits, i.e. worker size, (Nsteps) is smaller than the lower borderline of the confidence interval (95% CI), the trait is phylogenetically conservative.

Smoky Mts

Vorarlberg Mts

Chiricahua Mts

Nsteps

95% CI

nest site habitat worker size nest site habitat worker size

18 6 7.9 25 16 10.7

(18; 24) (7; 9) (17.8; 175.7) (37; 44) (19; 27) (11.2; 233.9)

nest site habitat worker size

25 24 50.1

(27; 32) (25; 33) (55.7; 292.6)

By taking an explicit community phylogenetics approach, our work suggests that the factors that structure these ant communities may vary in predictable ways along environmental gradients. In 1960, Fischer wrote ‘‘much of the killing in high latitudes is done by the less selective inorganic forces . . . In the tropics on the other hand, the physical environment is more benign to most organisms, and the highly selective interorganic struggle for existence is more apparent.’’ Our results support Fischer’s assertion, but put it in the context of phylogenetic niche conservatism. Because traits of ants, like those of many taxa, are phylogenetically conserved, much of the ‘‘killing’’ at high elevations is not random phylogenetically, leaving species of those relatively few clades that have evolved the ability to deal with cold climates. In lowlands, competition appears to structure the communities rather than habitat filtering. Such mechanisms have long been suggested for both elevational and latitudinal gradients, but rarely tested (Bryant et al. 2008, Graham et al. 2009), especially among montane systems. A question separate from whether environmental conditions affect phylogenetic structure in consistent ways (they do), is whether the patterns in phylogenetic structure are sufficient to account for patterns in species density or other measures of biological diversity. We found that while phylogenetic structure was correlated with ant species density, it was a poorer predictor of species density than was temperature. In other words, temperature appears to affect species density over and above its effects on phylogenetic structure. Bringing these results together, it seems that temperature may influence ant diversity both via its effects on phylogenetic structure (in essence a function of niche conservatism) as well as via other mechanisms, whether due to effects on abundance and ultimately extinction rates or some other link (Willig et al. 2003, Davies et al. 2004, Sanders et al. 2007). Finally, we need to consider the possibility that phylogenetic structure both influences and is influenced by species density. It is interesting to consider just which ant taxa appear to have traits that allow them to pass through such an environmental filter in colder, high-elevation sites. Across these three disparate montane systems, species from the genera Formica, Myrmica, Temnothorax appear to be the most common at high-elevation sites (Supplementary material Appendix 3). Some of these genera possess traits

(a)

(b)

(c)

NRI

–2

15 0

5 10 Temperature (°C)

15 0

Figure 2. The relationship between phylogenetic community structure (NRI) and mean annual temperature (F1, 51 24.33, p B0.001, R2 0.36). The fitted relationship is depicted separately for (a) Smoky Mts, (b) Vorarlberg, (c) Chiricahua Mts. Significantly overdispersed/clustered communities are depicted as white points, and communities approaching random phylogenetic structure as black points.

that might facilitate overwintering (Kaspari and Vargo 1995, Geraghty et al. 2007). For example, some species of the genus Formica produce thatch mounds that enhance their freeze tolerance (Erpenbeck and Kirchner 1983, Heinze 1992). For other genera it is less clear what traits have allowed them to colonize and persist in high-elevation sites, though given that living at these sites appears difficult, it seems safe to assume that all of the relatively few lineages that persist there have evolved specific traits of one sort or another associated with cold tolerance. Overdispersion in the phylogenies at low elevations is congruent with the notion that interspecific competition shapes the composition of local ant communities. Though interspecific competition has long been given primacy as the structuring agent in ant communities (Ho¨ lldobler and Wilson 1990, Andersen 1992, Deslippe and Savolainen 1995, Parr et al. 2005), our results suggest that competition may structure ant communities, but conditionally so, only in more favorable climatic conditions (Retana and Cerda´ 2000). This is not to say that competition is unimportant in cold or other extreme

conditions, only that it is secondary to the effects of environmental or habitat filtering. The overall trend was clustering at high elevations, but overdispersion at low elevation sites. But of course there was variation around this general trend: several of the individual ant communities examined that did not display significant phylogenetic structure were found at high elevation. One possible explanation for lack of phylogenetic structure at a few high elevation sites is that those sites included species that were poorly resolved on our phylogeny and so were biased in some way. However, the robustness of NRI across different phylogeny resolutions makes this seem unlikely. An alternative explanation is that the relative extent of phylogenetic structure varies due to factors additional to those considered here, such as additional climatic or environmental variables (such as the composition of plant communities) or due to local variation in climate not captured at the scale at which we sampled the climate. In addition, both environmental filtering and competition might simultaneously occur and obscure one another in the overall phylogenetic structure of a community (Helmus

Table 5. The results indicate that both the NRIs derived from genus-level phylogeny and from phylogeny without branch length estimates are highly correlated (besides Chiricahua Mts) with NRIs based on our phylogenies (Supplementary material Appendix 2). Please, see the text for discussion of the results. Genus level

Smoky Mts Vorarlberg Mts Chiricahua Mts

Topology

3893 1219 3.744

26 16 7

0.996 0.999 0.505

B 0.001 B 0.001 0.094

281.8 41.29 3.529

26 16 7

0.955 0.838 0.490

B0.001 B0.001 0.102

369

et al. 2007). Gradients in phylogenetic structure with environmental gradients are unlikely to ever be absolute. Rather, they reflect tendencies from which individual communities can deviate in many ways. Our work adds to a growing number of studies that have examined phylogenetic structure (Emerson and Gillespie 2008, Graham and Fine 2008, Algar et al. 2009, CavenderBares et al. 2009, Vamosi et al. 2009). To our knowledge, however, only two studies to date have examined how phylogenetic structure varies along environmental gradients and neither has actually tested whether the niche conservatism assumed in filtering models existed. In the Andes, hummingbird communities studied by Graham et al. (2009) tended to be phylogenetically clustered at high elevation and overdispersed at low elevations, which agrees with the results we document here for ants. However, in another study conducted in the Rocky Mountains in Colorado, microbial communities were phylogenetically clustered at all elevations, whereas plant communities tended to be more overdispersed with increasing elevation (Bryant et al. 2008). One possibility for these contrasting results among taxa and studies is that different forces might act to shape the structure of communities composed of different taxa. After all, microbes, plants, hummingbirds, and ants all perceive their environment and interact with it in different ways; one taxon’s extreme climate is another taxon’s favorable climate. While the peaks of elevational gradients represent the coldest conditions in which ants are found (Ho¨ lldobler and Wilson 1990), they are not even close to the most extreme climates encountered by microbes, which can be found even at the center of snowflakes (Black 2008). In sum, our results imply that the interplay between interspecific interactions, trait evolution, and temperature shapes the distribution of species among three gradients. We can also infer the critical temperature at which the importance of competitive interactions as a structuring agent fades and habitat filtering (stress) begins to dominate. At least for the temperate ecosystems considered here that temperature is 108C (indicated by a zero value of NRI). It is worth noting that this is the temperature in recent global analyses (Dunn et al. 2009) at which ant diversity drops dramatically, suggesting that the barriers to overwintering, harvesting sufficient food, or some combination thereof, are overcome by only a few lineages below this temperature. More consideration of why only a subset of ant lineages do well in conditions cooler than this threshold will be fundamental to understanding patterns in ant distribution, but also to understanding how ant communities may change as global temperatures change.

Acknowledgements We thank Alan N. Andersen, Jessica Bryant, Andy Purvis, Michael D. Weiser, and anonymous reviewers for providing advice that greatly improved the manuscript. R. R. Dunn and N. J. Sanders were supported by a DOE-NICCR grant, DOE-PER grant DE-FG02-08ER64510 and NASA award NNX09AK22G. A. Machac and M. Janda were supported by a Czech Academy of Science grant (KJB612230701) and Czech Ministry of Education Grants (LC06073, 6007665801, ME0908), the Czech Grant Agency (206/08/H044, 206/09/

370

0115, P505/10/0673) and by the J. W. Fulbright and the Marie Curie Fellowships to MJ.

References Algar, A. C. et al. 2009. Evolutionary constraints on regional faunas: whom, but not how many. Ecol. Lett. 12: 57 65. Andersen, A. N. 1992. Regulation of momentary diversity by dominant species in exceptionally rich ant communities of the Australian seasonal tropics. Am. Nat. 140: 401 420. Andersen, A. N. 1997. Functional groups and patterns of organization in North American ant communities: a comparison with Australia. J. Biogeogr. 24: 433 460. Bini, L. M. et al. 2009. Coefficient shifts in geographical ecology: an empirical evaluation of spatial and non-spatial regression. Ecography 32: 193 204. Black, J. G. 2008. Microbiology: principles and explorations. Wiley. Bolton, B. 2003. Synopsis and classification of Formicidae. Mem. Am. Entomol. Inst. 71: 1 370. Brady, S. G. et al. 2006. Evaluating alternative hypotheses for the early evolution and diversification of ants. Proc. Natl Acad. Sci. USA 103: 18172 18177. Bryant, J. A. et al. 2008. Microbes on mountainsides: contrasting elevational patterns of bacterial and plant diversity. Proc. Natl Acad. Sci. USA 105: 11505 11511. Cavender-Bares, J. et al. 2009. The merging of community ecology and phylogenetic biology. Ecol. Lett. 12: 693 715. Clarke, A. and Gaston, K. J. 2006. Climate, energy and diversity. Proc. R. Soc. B 273: 2257 2266. Colwell, R. K. et al. 2004. The mid-domain effect and species richness patterns: what have we learned so far? Am. Nat. 163: E1 E23. Cubo, J. et al. 2005. Phylogenetic signal in bone microstructure of sauropsids. Syst. Biol. 54: 562 574. Davies, T. J. et al. 2004. Environmental energy and evolutionary rates in flowering plants. Proc. R. Soc. B 271: 2195 2200. Deslippe, R. J. and Savolainen, R. 1995. Mechanisms of competition in a guild of Formicine ants. Oikos 72: 67 73. Dunn, R. R. et al. 2009. Climatic drivers of hemispheric asymmetry in global patterns of ant species richness. Ecol. Lett. 12: 324 333. Elton, C. 1946. Competition and the structure of ecological communities. J. Anim. Ecol. 1: 54 68. Emerson, B. C. and Gillespie, R. G. 2008. Phylogenetic analysis of community assembly and structure over space and time. Trends Ecol. Evol. 23: 619 630. Erpenbeck, A. and Kirchner, W. 1983. Aspects of cold resistance of the red wood ant Formica polyctena (Hymenoptera, Formicidae). J. Appl. Entomol. 96: 271 281. Fischer, A. G. 1960. Latitudinal variations in organic diversity. Evolution 14: 64 81. Geraghty, M. J. et al. 2007. Body size, colony size, and range size in ants (Hymenoptera: Formicidae): are patterns along elevational and latitudinal gradients consistent with Bergmann’s Rule? Myrmecol. News 10: 51 58. Glaser, F. 2006. Biogeography, diversity and vertical distribution of ants (Hymenoptera: Formicidae) in Vorarlberg, Austria. Myrmecol. News 8: 263 270. Gotelli, N. J. 2000. Null model analysis of species co-occurrence patterns. Ecology 81: 2606 2621. Gotelli, N. J. and Colwell, R. K. 2001. Quantifying biodiversity: procedures and pitfalls in the measurement and comparison of species richness. Ecol. Lett. 4: 379 391. Grafen, A. 1989. The phylogenetic regression. Phil. Trans. R. Soc. B 326: 119 157.

Graham, C. H. and Fine, P. V. A. 2008. Phylogenetic beta diversity: linking ecological and evolutionary processes across space in time. Ecol. Lett. 11: 1265 1277. Graham, C. H. et al. 2009. Phylogenetic structure in tropical hummingbird communities. Proc. Natl Acad. Sci. USA 106: 19673 19678. Hawkins, B. A. et al. 2003. Productivity and history as predictors of the latitudinal diversity gradient of terrestrial birds. Ecology 84: 1608 1623. Hawkins, B. A. et al. 2007. Metabolic theory and diversity gradients: where do we go from here? Ecology 88: 1898 1902. Heinze, J. 1992. Life-histories of sub-arctic ants. In: 2nd Meeting on Circumpolar Ecosystems in Winter. Arctic Inst. N. Am., pp. 354 358. Helmus, M. R. et al. 2007. Separating the determinants of phylogenetic community structure. Ecol. Lett. 10: 917 925. Hessen, D. O. et al. 2007. Energy input and zooplankton species richness. Ecography 30: 749 758. Hijmans, R. J. et al. 2005. Very high resolution interpolated climate surfaces for global land areas. Int. J. Climatol. 25: 1965 1978. Ho¨ lldobler, B. and Wilson, E. O. 1990. The ants. Belknap Press of Harvard Univ. Press. Kaspari, M. and Vargo, E. L. 1995. Colony size as a buffer against seasonality Bergmann’s rule in social insects. Am. Nat. 145: 610 632. Kaspari, M. et al. 2000. Energy, density, and constraints to species richness: ant assemblages along a productivity gradient. Am. Nat. 155: 280 293. Katoh, K. et al. 2002. MAFFT: a novel method for rapid multiple sequence alignment based on fast Fourier transform. Nucl. Acids Res. 30: 3059 3066. Kembel, S. W. 2009. Disentangling niche and neutral influences on community assembly: assessing the performance of community phylogenetic structure tests. Ecol. Lett. 12: 949 960. Kembel, S. W. and Hubbell, S. P. 2006. The phylogenetic structure of a neotropical forest tree community. Ecology 87: S86 S99. Kembel, S. W. et al. 2009. ‘R’ tools for integrating phylogenies and ecology. R package ver. 0.7, <http://cran.r-project.org/ web/packages/picante/index.html>. Kozak, K. H. and Wiens, J. J. 2010. Niche conservatism drives elevational diversity patterns in Appalachian salamanders. Am. Nat. 176: 40 54. Losos, J. B. 2008. Phylogenetic niche conservatism, phylogenetic signal and the relationship between phylogenetic relatedness and ecological similarity among species. Ecol. Lett. 11: 995 1003. Maddison, W. P. and Maddison, D. R. 2002. Mesquite: a modular system for evolutionary analysis. Version 0.992. <http://mesquiteproject.org>. McCain, C. M. 2009. Global analysis of bird elevational diversity. Global. Ecol. Biogeogr. 18: 346 360. Moreau, C. S. et al. 2006. Phylogeny of the ants: diversification in the age of angiosperms. Science 312: 101 104.

Parr, C. L. et al. 2005. Constraint and competition in assemblages: a cross-continental and modeling approach for ants. Am. Nat. 165: 481 494. Pie, M. R. and Traniello, J. F. A. 2007. Morphological evolution in a hyperdiverse clade: the ant genus Pheidole. J. Zool. 1: 99 109. Pinheiro, J. C. and Bates, D. M. 2000. Mixed-effects models in S and S-PLUS. Springer. Rahbek, C. 2005. The role of spatial scale and the perception of large-scale species-richness patterns. Ecol. Lett. 8: 224 239. Rangel, T. et al. 2006. Towards an integrated computational tool for spatial analysis in macroecology and biogeography. Global. Ecol. Biogeogr. 15: 321 327. Retana, J. and Cerda´, X. 2000. Patterns of diversity and composition of Mediterranean ground ant communities tracking spatial and temporal variability in the thermal environment. Oecologia 123: 436 444. Sanders, N. J. 2002. Elevational gradients in ant species richness: area, geometry, and Rapoport’s rule. Ecography 25: 25 32. Sanders, N. J. et al. 2007. Temperature, but not productivity or geometry, predicts elevational diversity gradients in ants across spatial grains. Global. Ecol. Biogeogr. 16: 640 649. Simberloff, D. S. 1970. Taxonomic diversity of island biotas. Evolution 24: 23 47. Smith, S. A. et al. 2007. A phylogenetic perspective on elevational species richness patterns in Middle American treefrogs: why so few species in lowland tropical rainforests? Evolution 61: 1188 1207. Stevens, R. D. 2006. Historical processes enhance patterns of diversity along latitudinal gradients. Proc. R. Soc. B 273: 2283 2289. Storch, D. et al. 2006. Energy, range dynamics and global species richness patterns: reconciling mid-domain effects and environmental determinants of avian diversity. Ecol. Lett. 9: 1308 1320. Swofford, D. L. 1993. PAUP a computer program for phylogenetic inference using maximum parsimony. J. Genet. Physiol. 102: A9 A9. Vamosi, S. M. et al. 2009. Emerging patterns in the comparative analysis of phylogenetic community structure. Mol. Ecol. 18: 572 592. Webb, C. O. 2000. Exploring the phylogenetic structure of ecological communities: an example for rain forest trees. Am. Nat. 156: 145 155. Webb, C. O. et al. 2002. Phylogenies and community ecology. Annu. Rev. Ecol. Syst. 33: 475 505. Webb, C. O. et al. 2008. Phylocom: software for the analysis of phylogenetic community structure and trait evolution. Bioinformatics 24: 2098 2100. Wiens, J. J. et al. 2007. Phylogenetic history underlies elevational biodiversity patterns in tropical salamanders. Proc. R. Soc. B 274: 919 928. Willig, M. R. et al. 2003. Latitudinal gradients of biodiversity: pattern, process, scale, and synthesis. Annu. Rev. Ecol. Evol. Syst. 34: 273 309.

Download the Supplementary material as file E6629 from <www.oikos.ekol.lu.se/appendix>.

371

ECOGRAPHY 27: 29 /40, 2004

Interannual changes in folivory and bird insectivory along a natural productivity gradient in northern Patagonian forests C. Noemi Mazı´a, Thomas Kitzberger and Enrique J. Chaneton

Mazıá, C. N., Kitzberger, T. and Chaneton, E. J. 2004. Interannual changes in folivory and bird insectivory along a natural productivity gradient in northern Patagonian forests. / Ecography 27: 29 /40. Trophic regulation models suggest that the magnitude of herbivory and predation (topdown forces) should vary predictably with habitat productivity. Theory also indicates that temporal abiotic variation and within-trophic level heterogeneity both affect trophic dynamics, but few studies addressed how these factors interact over broad-scale environmental gradients. Here we document herbivory from leaf-feeding insects along a natural rainfall/productivity gradient in Nothofagus pumilio forests of northern Patagonia, Argentina, and evaluate the impact of insectivorous birds on foliar damage experienced by tree saplings at each end of the gradient. The study ran over three years (1997 /2000) comprising a severe drought (1998 /1999), which allowed us to test how climatic events alter top-down forces. Foliar damage tended to increase towards the xeric, least productive forests. However, we found a predictable change of insect guild prevalence across the forest gradient. Leaf miners accounted for the greater damage recorded in xeric sites, whereas leaf chewers dominated in the more humid and productive forests. Interannual folivory patterns depended strongly on the feeding guild and forest site. Whereas leaf-miner damage decreased during the drought in xeric sites, chewer damage increased after the drought in the wettest site. Excluding birds did not affect leaf damage from miners, but generally increased chewer herbivory on hydric and xeric forest saplings. Indirect effects elicited by bird exclusion became most significant after the drought, when total folivory levels were higher. Thus, interannual abiotic heterogeneity markedly influenced the amount of folivory and strength of top-down control observed across the forest gradient. Moreover, our results suggest that spatial turnovers between major feeding guilds may need be considered to predict the dynamics of insect herbivory along environmental gradients. C. N. Mazıá (cmazia@mail.agro.uba.ar), Dept de Produccioń Vegetal, Fac. de Agronomıá, Univ. de Buenos Aires, Av. San Martıń 4453, 1417 Buenos Aires, Argentina. / T. Kitzberger, Dept de Ecologıá, Univ. Nacional del Comahue, Quintral 1250, 8400 Bariloche, Rıó Negro, Argentina. / E. J. Chaneton, IFEVA /CONICET, Fac. de Agronomıá,Univ. de Buenos Aires, Av. San Martıń 4453, 1417 Buenos Aires, Argentina.

The search for patterns in trophic dynamics is central for understanding how food webs are structured and react to external perturbations. Phytophagous insects are stereotyped as primary consumers whose trophic relations can be strongly affected by environmental heterogeneity in both space and time (Hunter and Price 1992, Hartley and Jones 1997). Nonetheless, the extent to

which insect herbivory patterns reflect the simultaneous impact of spatio-temporal forms of environmental variation has received surprisingly little attention (Floyd 1996, Ritchie 2000, Forkner and Hunter 2000). Multiyear studies conducted at contrasting sites along major habitat gradients represent a powerful approach to examine herbivory dynamics in response to biotic and

Accepted 14 July 2003 Copyright # ECOGRAPHY 2004 ISSN 0906-7590 ECOGRAPHY 27:1 (2004)

abiotic sources of heterogeneity (Hunter and Price 1992, Brown et al. 2001). Large-scale, steep habitat gradients constitute natural scenarios where insect herbivory may drastically vary in response to either plant resources (Price 1992, Herms and Mattson 1992), natural enemies (Fernandes and Price 1992, Preszler and Boecklen 1996, Alonso 1999), or their interaction (Koptur 1985, Hacker and Bertness 1995, Denno et al. 2002). Inspired by food chain theory (sensu Fretwell 1987), mainly the exploitation ecosystems hypothesis (Oksanen et al. 1981, Oksanen and Oksanen 2000), several recent studies have investigated whether insect herbivory changes predictably along productivity and resource gradients (Stiling and Rossi 1997, Fraser and Grime 1998, Uriarte and Schmitz 1998, Denno et al. 2002). However, models based on simple energetic constraints have had varied success in predicting invertebrate herbivory (SchaÂ¨dler et al. 2003), and have been criticised for neglecting the role of environmental variability and within-trophic level heterogeneity in food web dynamics (Strong 1992, Hunter and Price 1992, Persson et al. 1996, Leibold et al. 1997). Temporal changes in abiotic conditions can markedly affect insect relations with both their food plants and natural enemies (Larsson 1989, Ritchie 2000). Year-toyear climatic variation and extreme events, e.g. droughts, are known to alter plant quality to herbivorous insects (Ayres 1993, Koricheva et al. 1998). Insect performance may be also directly affected by unusual weather conditions, which often account for transient peaks (Mattson and Haack 1987, Ritchie 2000) or declines (Shure et al. 1998) in herbivory. In addition, insects with different feeding modes, such as endophytic vs exophytic leaf feeders, display specific demographic or behavioural responses to changes in resource availability (Price 1992). Thus insect feeding guilds may exhibit differential susceptibility to various plant and stress factors (Larsson 1989, Koricheva et al. 1998, Shure et al. 1998), as well as to being killed by different natural enemies (Hawkins et al. 1997, Moon and Stiling 2002). This within-trophic level diversity may confound aggregate herbivory patterns, as feeding guilds might replace each other across habitat gradients, while also showing divergent responses to climate variation. There is ample evidence that natural enemies can limit the abundance of phytophagous insects and their impacts on host plants (e.g. Fernandes and Price 1992, Marquis and Whelan 1994, Hacker and Bertness 1995, Turchin et al. 1999). Simple food chain models predict that the magnitude of predation and strength of topdown control on herbivory increase with habitat productivity (Oksanen et al. 1981, Fretwell 1987). Although this idea has gained some support from invertebrate assemblages (Fraser and Grime 1997, 1998, Stiling and Rossi 1997, Denno et al. 2002), few studies so far addressed how insect guild heterogeneity may influence 30

trophic exploitation under changing productivity scenarios (see Sipura 1999, Ritchie 2000, Moon and Stiling 2002). Indeed, models allowing for species turnover within trophic levels predict a wider array of herbivory and predation responses to increased productivity (Leibold et al. 1997). The template/trophic cascade model proposed by Hunter and Price (1992) further suggests that temporal variability may be critical to the realised impact of top-down forces along habitat gradients. Yet, this issue has remained largely unexplored due to the short-term nature of most empirical studies (Ritchie 2000). We studied the spatial and temporal patterns of insect folivory in Nothofagus pumilio forests extending over a broad-scale rainfall/productivity gradient in northern Patagonia, Argentina. These native forests occur between the Andes and the Patagonian plains along a steep, west-to-east gradient of decreasing precipitation associated with distance from the Andean divide (Veblen et al. 1996). The forest gradient is characterised by varying levels of productivity decreasing eastward, as shown by remote sensing techniques based on satellite imagery (Paruelo et al. 2001). In addition, Patagonian forests occupy a climatically sensitive zone with large interannual variability in rainfall and temperature (Villalba and Veblen 1998), partly driven by the influence of El NinË&#x153;o Southern Oscillation on southern South America (Pittock 1980, Aceituno 1988). This setting offers a natural template where to examine how spatially and temporally changing environmental conditions may together influence trophic dynamics of herbivorous insects feeding on the same host plant species. The aims of this work were, first, to determine whether predictable patterns of herbivory and predation occur along a natural rainfall/productivity gradient in northern Patagonian forests, and second, to assess the interannual variation in strength of these top-down forces at contrasting forest habitats. We expected herbivory rates from various leaf-feeding insects and top-down effects from insectivorous birds, a major predator guild in the system (Grigera 1982, Jaksic and Feinsinger 1991), to change consistently across forests located at varying distances from the Andes. We also predicted spatial patterns of folivory and predation would be altered by year-to-year climatic variation in a feeding-guild specific way. Our 3-yr study comprised a severe drought period that affected interactions among trees, insects and birds. Repeated observations and experiments allowed us to document the impact of this climatic event on trophic dynamics along the forest gradient. We specifically addressed the following questions: 1) How does foliar damage from different insect feeding guilds vary over the rainfall/productivity gradient? 2) Are spatial patterns of herbivory consistent among years? 3) How does exclusion of insectivorous birds affect folivory levels at opposite ends of the forest gradient? 4) Does top-down ECOGRAPHY 27:1 (2004)

control from avian insectivores change in climatically contrasting years?

Methods Study system and field sites The study was conducted during 1998 /2000 along a west-to-east transect crossing Nahuel Huapi National Park (40845?S, 718 /728W), near Bariloche, Argentina. In this region the Andean cordillera sets an effective barrier to the westerlies, resulting in a pronounced eastward rainshadow. At the latitude of study, and over a distance of only 60 km, annual precipitation declines from /3000 mm near the divide to B/800 mm on the eastern foothills (Barros et al. 1983, Veblen et al. 1996). Above ca 1000 m a.s.l., native forests are dominated by the broad-leaved deciduous Nothofagus pumilio (‘‘lenga’’), a tall canopy tree forming extensive monospecific stands from the Valdivian rainforests in Chile through the drier forest zones in Argentina. Forest soils are poorly developed Andisols originated from volcanic ashes; depth of recent ash deposits varies from 2 m in the west to B/0.10 m in the east (Veblen et al. 1996). In March 1998, we selected N. pumilio stands at three contrasting positions along the west-to-east forest gradient, between 1050 and 1200 m altitude. Two sites were chosen at Paso Cordoba (40836?S, 71805?W), ca 60 km east of the Andean divide, where total rainfall is ca 1000 mm yr 1. The intermediate gradient position was represented by two sites located 35 /40 km away from the Andes, one in Cerro Bayo (40847?S, 71835?W) and the other in Cerro Lopez (41806?S, 71833?W), both with ca 1800 mm yr 1 rainfall. Finally, we selected one site at Paso Puyehue (40837?S, 71850?W) on the border between Chile and Argentina, with ca 3000 mm yr 1 rainfall. Limited accessibility and logistics prevented us from establishing another site on the wettest end of the gradient. The three locations on the forest gradient will be hereafter referred to as ‘‘xeric’’, ‘‘mesic’’ and ‘‘hydric’’. The selected forest zones differ considerably with regard of above-ground primary production as estimated by the Normalized Difference Vegetation Index (NDVI) derived from NOAA/AVHRR satellite imagery (Table 1; see Paruelo et al. 2001, Jobba´gy et al. 2002). For the study transect, NDVI measurements for the period 1995 /2000 (M. Garbulsky, Lab. of Regional Analysis, Fac. of Agronomy, Univ. Buenos Aires) show an eastward decline in annual productivity, relative to the hydric forest near the Andes, of 6% and 29% for mesic and xeric forests, respectively (Table 1). The study forests differed in annual tree growth rates estimated by ring-growth analysis (F /4.6, pB/0.025, n /8 trees/site, Table 1) and length of growing season. Leaf flushing in N. pumilio occurs between late October and early December, ECOGRAPHY 27:1 (2004)

depending on stand location (Veblen et al. 1996). Budbreak starts 3 /4 weeks earlier in the eastern xeric forests. Foliage colour changes during March /May, but leaf fall occurs a few weeks earlier in the wettest sites near the Andes. This results in foliage duration being ca 6-weeks longer in the eastern xeric forests. Drought conditions can produce premature leaf abscission especially in xeric forest sites (Veblen et al. 1996). The climate of northwestern Patagonia is characterized by cold and wet winters, and mild but dry summers (Fig. 1). Most precipitation falls as rain and snow during autumn and winter (March /September), well before the main growing season (November /February). Environmental conditions varied markedly during the 3-yr study period (Fig. 1). The 1997 /1998 season was humid (March /February: 940 mm) relative to the last 15 yr (mean9/SD /7769/178 mm). In contrast, the 1998 / 1999 season was extremely dry (357 mm). The 1998 winter was the warmest of the last 15 yr, and the ensuing summer /autumn period was also very warm and dry (Fig. 1). Total precipitation in 1999 /2000 was about twice (683 mm) that for the previous year, although the spring was again fairly dry (58% below average) and warm ( /1.78C), being followed by a humid summer (45% above average).

Tree foliage sampling Field sampling was conducted in March 1998, 1999 and 2000, towards the end of each of three growing seasons (Fig. 1, hereafter referred to as 1998 through 2000). In each site, we haphazardly selected 10 N. pumilio adult trees (dbh, mean9/SD [n /50]: 49.29/17.7 cm) from within a 1-ha forest stand. Trees were permanently tagged for use in subsequent years. On each sampling date, we collected one fully insolated (northly oriented) twig from each tree at a height of 5.5 m, using an extensible pole cutter. Loss of several tree tags in Cerro Lopez precluded resampling of this mesic site in March 2000. Herbivory was quantified for 30 leaves per tree as percentage leaf area damaged by insects, by placing a clear 1-mm2 plastic grid over the leaf and recording the area of intact and damaged (or missing) tissue. The number of damaged leaves in a sample was used to measure damage frequency per tree. We recorded foliar damage from different insect guilds, including chewers, miners, skeletonizers, and gallers (Gentili and Gentili 1988). Leaf-chewing insects included exophytic caterpillars (Lepidoptera) and weevils (Coleoptera, Curculionidae) (e.g. Bauerle et al. 1997). The most conspicuous chewers in the system are the larvae of Ormiscodes cinnamomea (Saturniidae: Hemileucinae), a native moth undergoing periodical outbreaks (Veblen et al. 1996). Larvae hatch in December from overwintering eggs and 31

Table 1. Characterisation of Nothofagus pumilio forests at three contrasting locations along the west-to-east rainfall/productivity gradient in northern Patagonia, Argentina. Data shown are means9/SE for each location. Site location on forest gradient

Annual NDVI* Tree-ring width (mm yr 1)$ Foliage traits Leaf size (cm2) Toughness (g mm 2) Water content (%) Nitrogen (%) Phosphorus (%)

Hydric

Mesic

Xeric

0.6919/0.016 2.69/0.5a

0.6489/0.008 1.49/0.3b

0.4949/0.018 1.29/0.2b

3.69/1.2a 40.49/1.2a 61.49/1.6a 2.149/0.05 0.279/0.004a

3.29/1.0b 40.69/1.0a 57.09/1.1b 2.079/0.04 0.189/0.005b

2.59/0.9c 67.19/0.8b 52.39/0.9c 2.079/0.07 0.199/0.009b

*NDVI /Normalised Difference Vegetation Index obtained from NOAA/NASA satellite Pathfinder AVHRR land database (James and Kalluri 1994). Values represent 5-yr (1995 /2000) averages of annual NDVI for an 8 /8-km scene centrered on each forest site. The NDVI estimates the fraction of photosynthetically active radiation absorbed by vegetation and derives from satellite measurements of red and near infrared light reflectance (see Paruelo et al. 2001, Jobba´gy et al. 2002). $Tree ring-growth data (1997 /2000, December 2001) were analyzed by repeated-measures ANOVA (site /year: F4,36 /2.18, pB/ 0.10), and leaf nutrient contents by one-way ANOVA. Measures of leaf size, toughness and % water represent 3-yr averages (1998 / 2000) analysed by two-way ANOVAs (see text for full statistics). Superscript letters indicate significant differences among sites (pB/ 0.05).

founding factor in comparative field studies (Hunter 1992, Mopper and Simberloff 1995), were not deemed to bias our herbivory measures across the gradient. Here, damage levels represent cumulative folivory within each growing season, reflecting both insect abundance and consumption. To further characterise differences among forest sites we recorded several foliar traits on tagged trees. Mean leaf size (area, cm2), toughness (g mm 2) and water content (% fresh mass) were measured in early March for 30 fresh leaves/tree in each site and year. Toughness was measured as the weight needed to punch a hole through the leaf using a 3 mm-diameter steel rod. Foliar N and P contents (% dry mass) were measured for five composite samples in each site; element concentrations were determined colorimetrically using an Alpkem IV autoanalyser.

Fig. 1. Seasonal and interannual variation in temperature (solid line) and rainfall (shaded bars) during the three-year study in northern Patagonia (INTA Bariloche, Rı´o Negro, Argentina). The open bars in both panels show the long-term means /SD (1984 /2000) for each season from autumn through summer. The arrows indicate the timing of leaf damage sampling.

develop through several instars feeding on tree foliage until February /March. Leaf mines in N. pumilio are produced by Heterobathmia spp. (Lepidoptera, Heterobathmiidae). Adult miners oviposit during November and early December and by late January (mid summer) larval feeding has been completed. At the time of sampling, leaves in all forest stands were fully expanded, but not senescing, and most foliar damage had already taken place (Bauerle et al. 1997, unpubl.). Thus, intersite differences in leaf phenology, a potentially con32

Bird exclusion experiments The effect of excluding birds on insect folivory rates was evaluated during the 1998 /1999 and 1999 /2000 seasons at two N. pumilio sites located on the xeric (Chall-Huaco Valley; 41812?S, 71812?W) and hydric (Paso Puyehue) ends of the gradient. The commonest insectivores in the system are the thorn-tailed rayadito (Aphrastura spinicauda , Furnaridae) and the white-crested elaenia (Elaenia albiceps, Tyranidae). Both are medium-sized species (body length :/15 cm) that feed preferentially on canopy insects (Grigera 1982) and are widely distributed across Andean forests (Grigera et al. 1996, Deferrari et al. 2001). In late spring 1998 (December), we selected in each forest site 20 N. pumilio saplings ranging 1 /1.5 m in height (2 /3 cm basal diameter). Bird exclusion cages were established on 10 randomly chosen saplings, while ECOGRAPHY 27:1 (2004)

the other 10 plants were left uncaged. Saplings within a site were at least 2 m apart from each other, being scattered over 2 ha of undisturbed forest. Cages were 2 m (height) /1 m (diameter) in size, and were constructed of wire-mesh wrapping (hole size: 2.5 cm) supported by 2 m-high iron rods. This mesh size effectively excluded insectivorous birds, while still allowing free access to other insect natural enemies such as spiders and parasitoids. The cages had a minor effect on light levels (PAR reduction B/5%) and had no measurable impact on air temperatures. This is not surprising given that the experiments were conducted under closed-canopy microhabitat conditions. Also, the cages were large enough to avoid any physical interference with sapling growth during the time frame of study. The bird exclusion experiments were started shortly after leaf flushing so that initial foliar damage on saplings was negligible. In early March 1999 and 2000, before foliage started changing colour, we collected 30 leaves per sapling to assess the percentage leaf area damaged by chewing and mining insects. By this late time of summer, foliar damage produced by both feeding guilds had already levelled off. We took special care to sample leaves from the central zone of saplings in order to minimise any ‘‘edge’’ effect on herbivory measures.

were arcsin-square-root transformed to reduce variance heterogeneity.

Results Foliage characteristics Mean leaf size increased westward from xeric, through mesic to hydric forest trees (F2,126 /26.1, pB/ 0.0001, Table 1), but did not vary among years (year and site / year, both pB/0.20). Leaf toughness was higher in xeric than in mesic or hydric sites (F2,126 /286.7, pB/ 0.0001) and remained so throughout the study, although toughness levels declined in all three sites during 1999 (year: F2,126 /153.3, pB/ 0.0001, site /year: pB/0.11). Leaf water content increased towards the hydric site (Table 1; F2,126 /13.8, pB/ 0.0001). Not surprisingly, leaf water varied among years (F2,126 /15.9, pB/ 0.0001), mainly because it decreased in xeric sites with the 1999 drought (site /year: F4,126 /6.7, pB/ 0.0001). Foliar N showed no significant difference across forest sites (F2,11 /0.5, pB/ 0.63), while P content was higher in the hydric forest (F2,11 /60.4, pB/ 0.0001, see Table 1).

Patterns of insect herbivory Statistical analysis We performed factorial ANOVA to test for differences in herbivory among sites and years. The analyses were conducted using the General Linear Models procedure of SAS with type-III sum of squares (Anon. 1996). Means and error measures were calculated by pooling all trees sampled in each gradient position within a year. This design considered individual trees as the basic unit of analysis and, therefore, it tested for statistical differences among forest ‘‘sites’’ without implying a causal relationship between herbivory and any particular habitat factor. The site /year interaction was strongly significant for most damage measures, indicating the occurrence of temporally changing patterns of herbivory across forest sites. Post-hoc pairwise comparisons were performed to examine differences among sites within years, using Bonferroni tests adjusting for a fixed number of comparisons (Day and Quinn 1989). A three-way, split-plot ANOVA was used to evaluate the effect of excluding birds on leaf damage for xeric and hydric forest saplings. The site and year of study entered the analysis as unreplicated main plots and the cage treatment as replicated subplot. Because ANOVA revealed strong cage /year effects on herbivory levels, we compared cage effects between forest sites within each year using Bonferroni tests adjusting for k /6 pairwise tests/year (Day and Quinn 1989). Percent damage data ECOGRAPHY 27:1 (2004)

Total insect damage changed significantly among sites and years (Table 2). The frequency of damaged leaves per tree ranged 30 /70%, being higher in xeric and mesic sites, except for 2000 (Fig. 2). Damage frequency was overall significantly greater in 2000 after the extended drought. Herbivory levels represented, on average, B/14% of the leaf area. However, leaf area damage varied as much as 3.3 /4.6-fold across sites within any given year, and up to 5-fold among years within a given site. The highest foliar damage occurred in the xeric forest in 1998 and 2000 (Fig. 3), but no significant difference was found in 1999 (site /year: pB/ 0.001). After the drought (2000), leaf area damage increased at both ends of the gradient (Fig. 3). We observed little interannual variation in folivory levels within mesic sites. Leaf miners and chewers accounted for /80% of total insect damage. The amount of herbivory from both feeding guilds varied significantly among forest sites and with the study year (Table 2, Fig. 3). Leaf area damage by miners was much greater in the xeric forest, except for 1999, when damage levels dropped 50% relative to 1998/ 2000 (site /year: pB/ 0.001). Damage frequency by leaf miners showed a slightly different picture, although it also varied widely (10 /50%) across sites and years. Mining insects consistently damaged a larger fraction of available leaves in the xeric and mesic sites than in the hydric one (Fig. 2). For mining insects, damage frequency approximates the rate of successful mines initiated by ovipositing adults, without reference to 33

Table 2. ANOVA results for leaf damage levels from various insect feeding guilds on Nothofagus pumilio at contrasting sites on the west-to-east, rainfall/productivity gradient in northern Patagonia, Argentina, during 1998 /2000. Significant effects are shown in bold type. Site

Leaf area damage (%) Miners Chewers Skeletonizers Gallers All folivores Frequency of damaged leaves (%) Miners Chewers All folivores

Site /year*

Year

pB/

p B/

89.93 2.02 5.69 1.31 59.13

0.0001 0.137 0.004 0.27 0.0001

7.24 17.99 1.42 0.64 28.44

0.0011 0.0001 0.25 0.42 0.0001

5.12 2.62 4.48 0.33 5.33

0.001 0.038 0.002 0.71 0.001

25.45 8.14 11.13

0.0001 0.001 0.0001

2.73 14.50 15.62

0.069 0.0001 0.0001

2.00 4.39 2.84

0.098 0.003 0.027

*Statistical analyses, with type-III SS, included 3 sites /3 yr, except for gallers (2000 excluded due to zero values). Error term DF / 126 (for gallers, DF /89).

larval feeding performance (Mopper and Simberloff 1995). Damage from leaf-chewing insects did not consistently vary across sites, but changed among years (Table 2). Damage frequency by chewers increased over the study period in mesic and hydric sites (site /year: pB/0.01), from 20% in 1998 to 45% in 2000 (Fig. 2). Yet, the leaf area damaged by chewers increased after the drought (2000) only in the hydric forest (Fig. 3, site /year: p B/ 0.05). Insect skeletonizers contributed a small fraction to total folivory during this study. However, skeletonizers also showed a significant peak after the drought, but in the xeric forest (Fig. 3). Leaf-galling insects showed no trend across sites or years, with damage levels remaining B/0.5%. Overall, during the three-year study, chewing insects produced most foliar damage in mesic and hydric sites (Fig. 4), whereas leaf miners accounted for the greater damage recorded in xeric sites (two-way ANOVA on proportional leaf-miner damage, site: F2,126 /74.3, pB/ 0.001, site /year: p / 0.10).

Effects of bird exclusion Total folivory levels in experimental saplings were in the range of those found on adult trees within the same season, and generally increased from 1999 (2.5 /4%) to 2000 (4 /9%). Total leaf damage was strikingly similar in hydric and xeric forest saplings, irrespective of year and cage treatment (Table 3). This reflected a compensatory shift in proportional damage caused by chewing and mining insects in each forest site (Fig. 5). Leaf-chewer herbivory on saplings was significantly higher in the hydric site, especially during 2000 (site /year: pB/ 0.015). In contrast, leaf-miner damage was higher on the xeric forest saplings, and did not significantly change between years (see Table 3). 34

Bird exclusion promoted a highly significant increase in leaf-chewer damage levels, which was equivalent for the xeric and hydric forest sites (Fig. 5, cage /site: p / 0.10). However, cage effects on chewer damage and total folivory rates significantly differed between study years (Table 3). The treatment effect became most apparent in 2000, when chewer folivory levels in either forest site were about twice as high for caged saplings as for those open to bird access (Fig. 5). Excluding birds did not consistently affect the amount of leaf-miner herbivory on saplings. However, there was a marginal three-way interaction (see Table 3), suggesting a weak cage effect on leaf miners for the xeric site that changed direction between years (Fig. 5).

Discussion Many studies have documented the trophic relations of endophytic or exophytic insects along natural habitat gradients (Koptur 1985, Fernandes and Price 1992, Hacker and Bertness 1995, Preszler and Boecklen 1996, Uriarte and Schmitz 1998, Alonso 1999, Denno et al. 2002). In addition, a number of studies focused on the temporal variation of insect herbivory (Faeth 1985, Shure et al. 1998) and of top-down factors acting on phytophagous insects (Floyd 1996, Turchin et al. 1999, Ritchie 2000). In this work, we explicitly linked both sources of abiotic heterogeneity, and thus clearly showed that folivory patterns over broad-scale environmental gradients can be highly variable across years, even for insects feeding on the same plant species. Furthermore, we found that spatial patterns of herbivory and predation not only were affected by a severe drought event, but also depended strongly on the focal insect feeding guild. These results are compatible with Hunter and Priceâ&#x20AC;&#x2122;s (1992) conceptual model emphasising the importance of temporal variability and guild differentiation for ECOGRAPHY 27:1 (2004)

Fig. 3. Leaf area damage (mean /1 SE) by different insect guilds in Nothofagus pumilio along the west-to-east, rainfall/ productivity gradient. Means represent all trees sampled in each gradient position within a year. Horizontal lines connect years not differing in foliar damage (p /0.05). Different letters above error bars denote significant differences among sites within a year (Bonferroni test, p B/ 0.0066).

be viewed as two major ecosystem drivers (Veblen et al. 1996, Paruelo et al. 2001, see also Price 1992). It must however be noted that several ‘‘bottom-up’’ factors also changed predictably with precipitation geographical distance from the Andes. These comprised various measured leaf traits (Table 1), soil depth, canopy openness, length of growing season and foliar phenology, all of which may directly or indirectly affect forest insects (see Hunter 1992, Shure and Wilson 1993, Mopper and Simberloff 1995, Shure et al. 1998). Therefore, the dynamics of insect folivory reported here should not be interpreted as being the sole consequence of a specific site factor.

Spatial patterns of insect herbivory

Fig. 2. Frequency of leaf damage (mean /1 SE) in Nothofagus pumilio along a west-to-east rainfall/productivity gradient in northern Patagonia. Means represent all trees sampled for each gradient position within a year. Horizontal lines connect years not differing in foliar damage (p /0.05). Different letters above error bars denote significant differences among sites within a year (Bonferroni test, pB/ 0.0066).

understanding trophic regulation along habitat gradients (see also Persson et al. 1996, Leibold et al. 1997). We studied insect herbivory along a natural, composite forest gradient, where rainfall and productivity can ECOGRAPHY 27:1 (2004)

Overall folivory levels tended to increase towards the xeric, least productive end of the forest gradient. More importantly, we found a predictable change in feeding guild prevalence across N. pumilio forests (Fig. 4). Mining insects caused most foliar damage in xeric sites, whilst leaf chewers prevailed in mesic/hydric sites. The same pattern emerged from the cage experiments, eventhough in general damage levels for (uncaged) saplings were lower (2 /5%) than those for mature trees (2 /14%). Such a dominance replacement between insects with different feeding habit may reflect alternate adaptations to cope with natural variation in plant resources and abiotic factors (Price 1992). This result agrees with resource manipulation experiments that show contrasting feeding-guild specific responses to 35

Fig. 4. Changes in proportional leaf damage produced by mining and chewing insects on Nothofagus pumilio at contrasting sites along the west-to-east, rainfall/productivity gradient in northern Patagonia. Data are mean damage levels over three years; vertical bars show max./min. values for each site.

changes in host-plant quality or stress (Koricheva et al. 1998, Ritchie 2000, Moran and Scheidler 2002). Unfortunately, however, most field studies describing natural dynamics of insect herbivory have focused on a single feeding guild or species, which has limited our ability to account for community-wide patterns of herbivory in terrestrial environments (Hunter and Price 1992). Moreover, food chain models (e.g. Oksanen et al. 1981, Fretwell 1987) inevitably fail to accommodate such spatial complexities, because they do not allow for differential susceptibility to various controlling factors among herbivorous consumers (Polis and Strong 1996, Leibold et al. 1997). The observed shift in folivory mode was driven by increased leaf-miner activity in xeric sites. This is

consistent with the notion that mining insects may be well protected from dessication (Connor and Taverner 1997). It has been proposed that the distribution of endophytic insects may not reflect habitat productivity, but selection for oviposition sites (Price 1992). Although in our case the frequency of mined leaves did not vary as much between mesic and xeric sites, the leaf area damaged by miners was always greater in the xeric forest (Figs 2 and 3). This suggests that large-scale variation in miner herbivory could result from differences in feeding performance as well as ovipositional preferences (cf. Mopper and Simberloff 1995). Trees showing heavy miner damage in xeric sites had smaller and tougher leaves than least damaged conspecifics in wetter sites (Table 1). Yet, toughness may not work as an effective barrier to miners, since colonisation occurs early in the season on newly flushed, softer foliage (Maz覺織a unpubl.). Leaf phenology (Hunter 1992, Mopper and Simberloff 1995) and premature abscission (Bultman and Faeth 1986, Connor and Taverner 1997) have been indicated as factors influencing oviposition patterns and larval performance in mining insects. It is thus possible that early flushing and delayed abscission in xeric sites contributed to determine overall differences in miner herbivory across the forest gradient. Folivory by chewing insects did not differ among sites in two out of three study years. In contrast to mining insects, leaf chewers appeared to be primarily influenced by temporal effects (Table 2). From a bottom-up perspective (Price 1992), and given the contrasting habitats and vegetation traits associated with the forest gradient, the lack of spatial consistency in leaf-chewer herbivory is puzzling. Whether this resulted from compensatory changes between chewer abundance and consumption rates across forest habitats is unknown. When significant spatial differences for leaf chewers did emerge after the drought, highest damage levels occurred in the most productive forest (Figs 3 and 5), where trees had large and softer leaves, with a relatively high nutritional quality compared to trees in xeric sites (see Table 1).

Table 3. Three-way, split-plot ANOVA for the effect of excluding insectivorous birds on levels of leaf area damage experienced by Nothofagus pumilio saplings in hydric and xeric sites during 1998 /1999 and 1999 /2000. Significant effects are shown in bold type. Source

Site Year Site /year Subplot (site /year) Cage Cage /site Cage /year Cage /site /year Residual

Leaf chewers

Leaf miners

All folivores

pB/

p B/

1 1 1 36 1 1 1 1 36

74.63 98.90 14.29 2.10 113.94 5.21 29.58 0.33 2.62

35.56 47.13 6.81

0.0001 0.0001 0.013

11.81 2.45 1.04

0.0015 0.13 0.31

0.59 0.0001 0.48

0.0001 0.17 0.002 0.73

0.58 0.36 4.45 4.31

0.46 0.56 0.042 0.046

1.68 177.71 2.89 5.75 136.68 2.63 63.37 4.27 4.04

0.29 30.91 0.50

43.42 1.98 11.27 0.12

53.75 10.96 4.66 4.47 0.88 0.54 6.76 6.54 1.52

33.86 0.65 15.70 1.06

0.0001 0.43 0.0003 0.31

ECOGRAPHY 27:1 (2004)

Fig. 5. Effect of bird exclusion on levels of insect folivory on Nothofagus pumilio saplings in hydric and xeric forest sites. Bars show the mean /SE of 10 saplings/site. Different letters above error bars indicate significant differences within a year (Bonferroni test, pB/ 0.0083).

Temporal dynamics of herbivory Folivory levels varied markedly among years and, in general, were highest after the extended drought period. Nonetheless, dominant feeding guilds exhibited contrasting, site-specific temporal dynamics (Figs 2 and 3). Whereas leaf-miner damage in xeric sites dropped during the 1999 drought, to recover afterwards, leaf-chewing insects accounted for the increased damage recorded in the hydric forest after the drought. Interannual folivory changes on adult trees corresponded with those found on xeric and hydric forest saplings, which experienced a two-fold increase between 1999 and 2000 in miner and chewer damage, respectively. Shure et al. (1998) reported qualitatively similar responses to drought from leaf ECOGRAPHY 27:1 (2004)

skeletonisers and strip-feeding lepidopterans in an upland oak forest. Both direct and plant-mediated responses of foliar insects to abiotic conditions may have influenced yearto-year changes in herbivory (Larsson 1989, Ayres 1993, Ritchie 2000). External factors causing severe plant stress have been shown to negatively affect leaf miners in other forest communities (Shure et al. 1998). The decline in leaf area consumed by miners in 1999 suggests that drought conditions reduced their feeding performance and/or survival, perhaps by forcing early leaf abscission during the summer (Veblen et al. 1996; see Bultman and Faeth 1986, Mopper and Simberloff 1995). Whatever the mechanism, the rapid recovery of leaf 37

miners between 1999 and 2000 indicated that any drought effect was transient. On the other hand, temporal changes in chewer herbivory were consistent with the common observation that droughts often correlate with lagged peaks of insect damage in temperate forests (Mattson and Haack 1987, Larsson 1989, Price 1992). Low moisture levels and elevated temperatures prevailing from autumn 1998 through spring 1999 may have enhanced leaf-chewer activity in 2000 (Figs 2 and 3). Alteration of insect demographic rates and/or plant nutritional quality have both been invoked to explain such delayed peaks in herbivory (Mattson and Haack 1987, Larsson 1989, Ayres 1993). That the outburst of chewers damage had taken place only in the wettest forest suggests that local site factors modulated insect responses to climate fluctuation.

Top-down control on insect folivory Cage experiments indicated that insectivorous birds may exert an important regulatory force on tree folivory in these Patagonian forests. Indirect effects elicited by predator exclusion leading to increased herbivore damage on food plants provide evidence for the existence of trophic cascades (Schmitz et al. 2000), but such effects have rarely been documented for phytophagous insects on temperate forest trees (Marquis and Whelan 1994, Forkner and Hunter 2000). Notice that this study was not deviced to identify the precise mechanism behind the observed effects of excluding avian predators on insect herbivory. Thus, it must be recognised that differences in folivory between cage treatments may not need result from direct predation effects on insect abundance (see Marquis and Whelan 1994), but could also arise from nonlethal changes in herbivore behaviour (e.g., reduced feeding time, increased migration rate) associated with predation risk (Schmitz 1998). More critically, however, our results suggest that the actual strength of top-down effects from insectivorous birds changed depending on the leaf-feeding guild, and between climatically contrasting years. Bird exclusion produced weak and variable effects on leaf miners, but increased chewers damage by over 50%, especially after the drought (Fig. 5). These findings stress the importance of focusing on different herbivore guilds when looking at the relative impact of top-down forces along productivity/resource gradients (Sipura 1999, Moon and Stiling 2002, Moran and Scheidler 2002). It has been proposed that leaf miners may be more susceptible to parasitoid attack than to predator-induced mortality (Hawkins et al. 1997). The presence of such ‘‘intermediate’’ invertebrate consumers may compound the interpretation of herbivory patterns in vertebrate exclusion trials (Tscharntke 1997). Conceivably, inconsistent cage effects as observed for leaf miners in this 38

study (Table 3, Fig. 5) might result from such ‘‘hidden’’ trophic complexities (Polis and Strong 1996). We recognise that other taxa not measured here, e.g. spiders, may be important predators of foliar insects, while also serving as alternative food to birds. Nevertheless, our results indicate that invertebrate predators, if abundant at all, did not fundamentally modify the effect of excluding insectivorous birds on leaf-chewer herbivory. The elevated chewer damage on caged saplings suggests that bird predation could alter the herbivore guild structure across the forest gradient. We hypothesise that, by suppressing leaf-chewing insects, birds increase the prevalence of leaf miners in xeric forest sites (see Figs 4 and 5). The exploitation ecosystems hypothesis predicts that top-down control on herbivorous consumers should increase along habitat productivity gradients (Oksanen et al. 1981, Oksanen and Oksanen 2000). Thus, we expected bird exclusion would produce a greater increase in folivory on hydric than on xeric forest saplings. But our experiments revealed no significant cage /site interaction on folivory rates (Table 3), indicating that the net strength of top-down effects from insectivorous birds did not vary among forest habitats. Consequently, post-drought differences in herbivory experienced by saplings at contrasting productivity sites (see Fig. 5) could not be simply attributed to bird insectivory. Source-sink models developed by Oksanen (1990) show that trophic exploitation from highly mobile predators such as birds may actually erase local differences in herbivore abundance across spatially heterogeneous landscapes. Furthermore, our work, together with resource-addition field studies (Ritchie 2000, Forkner and Hunter 2000), suggests that bird /insect herbivore interactions along productivity gradients may not conform with models assuming temporally constant environments. There is much evidence that top-down controls on insect herbivory can be highly variable in time (e.g. Floyd 1996, Turchin et al. 1999) and may sometimes be rather weak (Ritchie 2000, Forkner and Hunter 2000). Our results support this proposal by showing that climate fluctuation and top-down factors may affect herbivory in a strongly non-additive way (Table 3). While bird exclusion did not alter folivory patterns in the drought year, it produced significant effects in 2000 when overall damage levels were higher (Fig. 5). It appears that avian activity on saplings increased once the drought period receded, as insectivorous birds probably tracked down local changes in insect abundance (Deferrari et al. 2001), or were attracted by foraging cues derived from greater foliage damage (Marquis and Whelan 1994, Floyd 1996, Sipura 1999). The temporal herbivory patterns reported in this study show that environmental effects associated with extreme climatic events may rapidly propagate through interactions ECOGRAPHY 27:1 (2004)

linking adjacent trophic levels. In this way, temporal abiotic heterogeneity may not only constrain the role of top-down forces in food web regulation (Floyd 1996, Ritchie 2000), but may also override the influence of ‘‘static’’ habitat gradients on herbivory. We conclude that a full understanding of herbivory patterns cannot be attained without consideration of the changing environmental context in which interactions are embedded (Brown et al. 2001). Our work supports the contention that biotic and abiotic sources of heterogeneity interact to control herbivory in terrestrial ecosystems (Hunter and Price 1992). We have shown that annual climatic variation differentially affected endophytic vs exophytic insects across a forest productivity gradient. Thus, accounting for spatial turnovers between major feeding guilds may be crucial to predict insect herbivory along environmental gradients. Acknowledgements / We are grateful to K. Heineman, N. Tercero, B. Guarnaschelli and G. Amico for field help, and N. Suarez for laboratory assisstance. M. Garbulsky (LART-UBA) kindly provided the NDVI satellite data. We also thank T. Whitham, M. Oesterheld, J. Gowda, M. Omacini and S. Faeth for discussion and/or comments on the manuscript. The study was supported by the Administracio´n de Parques Nacionales, U.S. National Science Foundation (Grant SBR-9421881 to T. Veblen) and Agencia Nacional de Promocio´n Cientı´fica y Tecnolo´gica (BID 802/OC-AR-PICT 2268 to T. Kitzberger).

References Aceituno, P. 1988. On the functioning of the Southern Oscillation in the South American sector. Part 1. Surface climate. / Monthly Weather Rev. 116: 505 /524. Alonso, C. 1999. Variation in herbivory by Yponomeuta mahalebella on its only host plant Prunus mahaleb along an elevational gradient. / Ecol. Entomol. 24: 371 /379. Anon. 1996. SAS System for Windows, release 6.12. / SAS Inst., Cary, NC, USA. Ayres, M. 1993. Plant defense, herbivory, and climate change. / In: Kareiva, P., Kingsolver, J. and Huey, R. (eds), Biotic interactions and global change. Sinauer, pp. 75 /94. Barros, V. et al. 1983. Cartas de precipitacioń de la zona oeste de las provincias de Rıó Negro y Neuqueń. / Internal Rep., Fac. Cs. Agricult., Univ. Nacl. Comahue, Neuqueń, Argentina. Bauerle, P., Rutherford, P. and Lanfranco, D. 1997. Defoliadores de roble (Nothofagus obliqua ), raulı´ (N. alpina ), coihue (N. dombeyi ) y lenga (N. pumilio ). / Bosque 18: 97 /107. Brown, J. H. et al. 2001. Complex system interactions and the dynamics of ecological systems: long-term experiments. / Science 293: 643 /649. Bultman, T. L. and Faeth, S. H. 1986. Selective oviposition by a leaf miner in response to temporal variation in abscission. / Oecologia 69: 117 /120. Connor, E. F. and Taverner, M. P. 1997. The evolution and adaptive significance of the leaf-mining habit. / Oikos 79: 6 /25. Day, R. W. and Quinn, G. P. 1989. Comparisons of treatments after an analysis of variance in ecology. / Ecol. Monogr. 59: 431 /463. Deferrari, G. et al. 2001. Changes in Nothofagus pumilio forest biodiversity during the forest management cycle. 2. Birds. / Biodiv. Conserv. 10: 2093 /2108. ECOGRAPHY 27:1 (2004)

Denno, R. F. et al. 2002. Bottom-up forces mediate naturalenemy impact in a phytophagous insect community. / Ecology 83: 1443 /1458. Faeth, S. H. 1985. Quantitative defense theory and patterns of feeding by oak insects. / Oecologia 68: 34 /40. Fernandes, G. W. and Price, P. W. 1992. The adaptive significance of insect gall distribution: survivorship of species in xeric and mesic habitats. / Oecologia 90: 14 /20. Floyd, T. 1996. Top-down impacts on creosotebush herbivores in a spatially and temporally complex environment. / Ecology 77: 1544 /1555. Forkner, R. E. and Hunter, M. D. 2000. What goes up must come down? Nutrient addition and predation pressure on oak herbivores. / Ecology 81: 1588 /1600. Fraser, L. H. and Grime, J. P. 1997. Primary productivity and trophic dynamics investigated in a North Derbyshire, UK, dale. / Oikos 80: 499 /508. Fraser, L. H. and Grime, J. P. 1998. Top-down control and its effects on the biomass and composition of three grasses at high and low soil fertility in outdoor microcosmos. / Oecologia 113: 239 /246. Fretwell, S. D. 1987. Food chain dynamics: the central theory of ecology? / Oikos 50: 291 /301. Gentili, M. and Gentili, P. 1988. Lista comentada de los insectos asociados a las especies sudamericanas del geńero Nothofagus. / Monogr. Acad. Nac. Cs. Ex. Fis. y Nat. 4: 85 /106. Grigera, D. 1982. Ecologıá alimentaria de algunas passeriformes insectı´voras frecuentes en los alrededores de S.C. de Bariloche. / Ecol. Argentina 7: 67 /84. Grigera, D., Ubeda, C. and Reca, A. 1996. Estado de conservacioń de las aves del Parque y Reserva Nacional Nahuel Huapi. / Hornero 14: 1 /13. Hacker, S. D. and Bertness, M. D. 1995. A herbivore paradox: why salt marsh aphids live on poor-quality plants. / Am. Nat. 145: 192 /210. Hartley, S. E. and Jones, C. G. 1997. Plant chemistry and herbivory, or why the world is green. / In: Crawley, M. J. (ed.), Plant ecology. Blackwell, pp. 284 /324. Hawkins, B. A., Cornell, H. V. and Hochberg, M. E. 1997. Predators, parasitoids, and pathogens as mortality agents in phytophagous insect populations. / Ecology 78: 2145 /2152. Herms, D. A. and Mattson, W. J. 1992. The dilemma of plants: to grow or defend. / Quart. Rev. Biol. 67: 283 /335. Hunter, M. D. 1992. A variable insect /plant interaction: the relationship between tree budburst phenology and opulation levels of insect herbivores among trees. / Ecol. Entomol. 15: 401 /408. Hunter, M. D. and Price, P. W. 1992. Playing chutes and ladders: heterogeneity and the relative roles of bottom-up and topdown forces in natural communities. / Ecology 73: 724 / 732. Jaksic, F. and Feinsinger, P. 1991. Bird assemblages in temperate forests of North and South America: a comparison of diversity, dynamics, guild structure, and resource use. / Rev. Chil. Hist. Nat. 64: 491 /510. James, M. E. and Kalluri, S. N. V. 1994. The pathfinder AVHRR land data set: an improved coarse resolution data set for terrestrial monitoring. / Int. J. Rem. Sen. 15: 3347 / 3363. Jobba´gy, E. G., Sala, O. E. and Paruelo, J. M. 2002. Patterns and controls of primary production in the Patagonian steppe: a remote sensing approach. / Ecology 83: 307 /319. Koptur, S. 1985. Alternative defenses against herbivores in Inga (Fabaceae: Mimosoideae) over an elevational gradient. / Ecology 66: 1639 /1650. Koricheva, J., Laarson, S. and Haukioja, E. 1998. Insect performance on experimentally stressed woody plants: a meta-analysis. / Annu. Rev. Entomol. 43: 195 /216. Larsson, S. 1989. Stressful times for the plant stress /insect performance hypothesis. / Oikos 56: 277 /283. Leibold, M. A. et al. 1997. Species turnover and the regulation of trophic structure. / Annu. Rev. Ecol. Syst. 28: 467 /494.

Marquis, R. J. and Whelan, C. J. 1994. Insectivorous birds increase growth of white oak through consumption of leafchewing insects. / Ecology 75: 2007 /2014. Mattson, W. J. and Haack, R. A. 1987. The role of drought in outbreaks of plant-eating insects. / Bioscience 37: 110 /118. Moon, D. C. and Stiling, P. 2002. The influence of species identity and herbivore feeding mode on top-down and bottom-up effects in a salt marsh system. / Oecologia 133: 243 /253. Mopper, S. and Simberloff, D. 1995. Differential herbivory in an oak population: the role of plant phenology and insect performance. / Ecology 76: 1233 /1241. Moran, M. D. and Scheidler, A. R. 2002. Effects of nutrients and predators on an old-field food chain: interactions of top-down and bottom-up processes. / Oikos 98: 116 /124. Oksanen, L. and Oksanen, T. 2000. The logic and realism of the hypothesis of exploitation ecosystems. / Am. Nat. 155: 703 /723. Oksanen, L. et al. 1981. Exploitation ecosystems in gradients of primary productivity. / Am. Nat. 118: 240 /261. Oksanen, T. 1990. Exploitation ecosystems in heterogeneous environments. / Evol. Ecol. 4: 220 /234. Paruelo, J. M., Jobba´gy, E. G. and Sala, O. E. 2001. Current distribution of ecosystem functional types in temperate South America. / Ecosystems 4: 683 /698. Persson, L. et al. 1996. Productivity and consumer regulation / concepts, patterns, and mechanisms. / In: Polis, G. and Winemiller, K. O. (eds), Food webs: integration of patterns and dynamics. Chapman and Hall, pp. 396 /434. Pittock, A. B. 1980. Patterns of climate variation in Argentina and Chile. I. Precipitation, 1931 /1960. / Monthly Weather Rev. 108: 1347 /1361. Polis, G. A. and Strong, D. R. 1996. Food-web complexity and community dynamics. / Am. Nat. 147: 813 /846. Preszler, R. W. and Boecklen, W. J. 1996. The influence of elevation on tri-trophic interactions: opposing gradients of top-down and bottom-up effects on a leaf-mining moth. / Ecoscience 3: 75 /80. Price, P. W. 1992. Plant resources as the mechanistic basis for insect herbivore population dynamics. / In: Hunter, M. D., Ohgushi, T. and Price, P. W. (eds), Effects of resource distribution on animal-plant interactions. Academic Press, pp. 139 /173.

Ritchie, M. E. 2000. Nitrogen limitation and trophic vs abiotic influences on insect herbivores in a temperate grassland. / Ecology 81: 1601 /1612. Scha¨dler, M. et al. 2003. Does the Fretwell /Oksanen model apply to invertebrates? / Oikos 100: 203 /207. Schmitz, O. J. 1998. Direct and indirect effects of predation and predation risk in old-field interaction webs. / Am. Nat. 151: 327 /342. Schmitz, O. J., Hamba¨ck, P. A. and Beckerman, A. P. 2000. Trophic cascades in terrestrial systems: a review of the effects of carnivore removals on plants. / Am. Nat. 155: 141 /153. Shure, D. J. and Wilson, L. A. 1993. Patch-size effects on plant phenolics in successional openings of the southern Appalachians. / Ecology 74: 55 /67. Shure, D. J., Mooreside, P. D. and Ogle, S. M. 1998. Rainfall effects on plant-herbivore processes in an upland oak forest. / Ecology 79: 604 /617. Sipura, M. 1999. Tritrophic interactions: willows, herbivorous insects and insectivorous birds. / Oecologia 121: 537 /545. Stiling, P. and Rossi, A. M. 1997. Experimental manipulations of top-down and bottom-up factors in a tri-trophic system. / Ecology 78: 1602 /1606. Strong, D. R. 1992. Are trophic cascades all wet? Differentiation and donor-control in speciose ecosystems. / Ecology 73: 747 /754. Tscharntke, T. 1997. Vertebrate effects an plant /invertebrate food webs. / In: Gange, A. C. and Brown, V. K. (eds), Multitrophic interactions in terrestrial systems. Blackwell, pp. 277 /297. Turchin, P., Taylor, A. D. and Reeve, J. D. 1999. Dynamical role of predators in population cycles of a forest insect: an experimental test. / Science 285: 1068 /1071. Uriarte, M. and Schmitz, O. J. 1998. Trophic control across a natural productivity gradient with sap-feeding herbivores. / Oikos 82: 552 /560. Veblen, T. T. et al. 1996. Ecology of southern Chilean and Argentinean Nothofagus forest. / In: Veblen, T. T., Hill, R. and Read, J. (eds), The ecology and biogeography of Nothofagus forests. Yale Univ. Press, pp. 293 /353. Villalba, R. and Veblen, T. T. 1998. Influences of large-scale climatic variability on episodic tree mortality in northern Patagonia. / Ecology 79: 2624 /2640.

ECOGRAPHY 27:1 (2004)

ECOGRAPHY 2 2 548-566. Copenhagen 1999

The altitudinal gradient of vascular plant richness in Aurland, western Norway Arvid Odland and H. J. B. Birks

Odland, A. and Birks, H. J. B. 1999. The altitudinal gradient of vascular plant richness in Aurland, western Norway. - Ecography 22: 548-566. Differences in vascular plant species richness along the altitudinal gradient in the Aurland area of western Norway have been investigated. Based on field surveys, as complete lists as possible of all vascular plants have been compiled for each 100 m altitudinal band, from sea level to the highest mountain (1764 m). For each of the 18 altitudinal bands, climatic data have been estimated. A total of 444 vascular plant species were recorded. Highest species richness (263 species) occurred in the 600-700 m band, whereas the uppermost band had only 10 species. There are minor differences in species number between the altitudinal bands < 1000 m. Partial least squares regression shows that species richness for the overall altitudinal gradient is well predicted by mean July and January temperatures and mean annual precipitation. Species turnover is highest in the 100-200 m, 600-700 m, and 1400-1500 m altitudinal bands. In terms of the gradient in summer temperature, the study supports the generally assumed linear relationship between July temperature and the number of vascular plant species between 700 and 1500 m corresponding with a mean July temperature range of 7-11°C. The study shows a decrease of ca 30 vascular plant species with a 1°C decrease in mean July temperature, and that the “climatic vascular plant limit” is here estimated to occur at a mean July temperature of 2.4”C. Above 1500 and below 700 m, species number is lower than expected based on summer temperature conditions alone. The 700-800 m band represents the highest floristic difference compared to the other bands. Ordination and classification analyses of the floristic compositional data of all the bands highlight the 600-800 and 1500-1600 m altitudinal bands as the major biotic boundaries along the gradient. No major discontinuity in species richness, composition, or turnover was consistently found, however, at the 1100- 1200 m band representing the forest-limit ecotone in Aurland. A . Odland (arvid.odland@hit.no), Inst. of Environmental Studies, Telemark College, N-3800 BO, Norway. - H. J. B. Birks, Botanical Inst., Univ. of Bergen, Alligaten 41, N-5007 Bergen, Norway.

It is widely observed that patterns of species richness decrease with decreasing growing-season temperatures (Rannie 1986), a pattern that applies to both latitudinal (Connor and McCoy 1979, Engelskjarn and Skifte 1995, Murray 1997) and altitudinal gradients (Jarrgensen 1932, Ozenda 1988, Rahbek 1995). The mechanistic reasons for this correlation remain, however, a matter of ongoing debate (cf. Rodhe 1992, 1997, Rosenzweig 1992, Gaston 1996, Rosenzweig and Sandlin 1997). In addition to changes in species richness, there is also a qualitative compositional change with an increase in

species of open habitats and a decrease in forest species with increasing altitude or latitude (Morisset et al. 1983, Engelskjarn and Skifte 1995). Consequently a major change or turnover in species composition is expected along latitudinal and altitudinal gradients (van Steenis 1984). However, few detailed data sets have been analysed to detect quantitative changes in species richness, composition, and turnover along altitudinal gradients, particularly in temperate areas. Recently, possible changes in alpine plant diversity as a result of global warming have been investigated (e.g.

548

ECOGRAPHY 22:5 (1999)

Grabherr et al. 1994, 1995). It is assumed that alpine vegetation zones on mountain slopes will show an upward migration of 400-600 m (Peters and Darling 1985, Holten and Carey 1992, Holten 1993, cf. Woodward 1993). Alpine tundra on moderately low mountains is thus in danger of extinction because of encroachment by today’s forests. In mountain areas a reduction in species richness may thus occur. This hypothesis assumes, however, that there is a close correlation between species richness and temperature, especially near and above the forest limit. In mountain areas, the vegetation is often divided into vertical altitudinal zones or belts. In all such zonations the forest limit is regarded as the main criterion for separation between the Boreal zone and the Alpine zone (Ahti et al. 1968, Moen 1998). Consequently one could assume that at the forest-limit ecotone there should be major changes in floristic composition and that there may also be changes in species richness. A vegetation zone is often assumed to be characterized by more or less homogenous vegetation types delimited by relatively narrow boundaries. Such delimitations involve the classical continuum/discontinuity controversy (Austin and Smith 1989, Minchin 1989, Austin 1990, Kent et al. 1997). The discontinuity hypothesis proposes that when species distributions are plotted along some gradient or gradient-complex whose rate of change is constant, groups of species exist that are replaced by other groups of species along the gradient (Whittaker 1956, Whittaker and Niering 1975, Austin 1990). Within each group most species have similar distributions, and the end of one group may coincide with the beginning of another. Although it has been implied that there are “critical altitudes”, i.e. points where vegetation and/or some environmental variables change abruptly (Hamilton 1975), the pattern has rarely been demonstrated with quantitative data (Kitayama 1992, Auerbach and Shmida 1993). The individualistic hypothesis proposes, on the other hand, that the centres and boundaries of species distribution are scattered individually along the environmental gradient (Whittaker 1956, Shipley and Keddy 1987). Different approaches and methods have been applied to describe or test variation in vascular plant richness and composition along altitudinal gradients in different floristic areas, climates, and vegetational types. These include: 1) total number of species (e.g. Shmida and Wilson 1985, Ozenda 1988); 2) distribution of tree species (e.g. Mark and Sanderson 1962, Mark 1963, Ahti et al. 1968, Hamilton 1975, Hamilton and Perrott 1981, Austin et al. 1983, Druitt et al. 1990, Kitayama 1992); 3) number of species with lower and upper distribution limits (species turnover) (e.g. van Steenis 1984, Shmida and Wilson 1985, Vazquez and Givnish 1998); 4) similarity or dissimilarity indices (e.g. Beak 1969, Hamilton 1975, Hamilton and Perrott 1981, ECOGRAPHY 2 2 5 (1999)

Baruch 1984, Kirkpatrick and Brown 1987, Ohsawa 1991, Itow 1991, Vazquez and Givnish 1998); 5) ordination and classification techniques that consider the overall floristic composition (e.g. Ogden and Powell 1979, Baruch 1984, Kirkpatrick and Brown 1987, Druitt et al. 1990, Kitayama 1992, Boyce 1998); 6) Monte Carlo simulations to determine the expected distributions of upslope and downslope distribution boundaries (e.g. Auerbach and Shmida 1993); and 7) various estimates of beta-diversity coefficients based on presence-absence (e.g. Wilson and Shmida 1984) or quantitative data (e.g. Wilson and Mohler 1983). In this study we have restricted our analyses to approaches 1, 3, and 5, namely species richness, species turnover, and species composition along altitudinal gradients. Several of the other approaches are not appropriate in our study given the limited number (9) of tree species, the relatively small number (18) of altitudinal bands, and the large number of species (444) compared to the number of altitudinal bands. Patterns in animal species richness and abundance along altitudinal gradients have also been studied using a variety of approaches and techniques (e.g. Graves 1985, 1988, Lawton et al. 1987, McCoy 1990, Rahbek 1997, Patterson et al. 1998), whereas altitudinal gradients in species richness have received rigorous theoretical analyses by Colwell and Hurtt (1994). Rahbek (1995) has reviewed the literature dealing with the altitudinal gradients of species richness (both plants and animals), and discussed whether there are any uniform patterns. Most of the papers reviewed were found to support the general view that species richness declines with altitude, but that this decline is not necessarily monotonic. Rahbek emphasized that the sampling method and the size of area sampled may strongly influence the results, thereby making comparisons and generalisations difficult. He concluded that “we have to accept the unsatisfactory realization that we do not know whether a general relationship exists between species richness and elevation, or whether an universal explanation can be given”. This paper considers the differences in vascular plant species richness and composition along the 0- 1764 m altitudinal gradient in Aurland, western Norway. It addresses the following questions: 1) How does species richness vary with altitude? 2) Are there any major changes or discontinuities in floristic composition and turnover? If so, do these coincide with major changes in the vegetation, for example the forest-limit ecotone? 3) How well is species richness predicted, in a statistical sense, by simple climatic variables and by the area of the different altitudinal bands? 4) How well is species richness at and above the forest-limit predicted by temperature? 5) What are the implications to species richness at these altitudes if air temperatures increase in the future? 549

Investigation area

sisting of Precambrian intrusive igneous rocks, mostly gneiss and arnphibolite, which are much poorer in The floristic data were obtained from the Aurland river calcium than the underlying rocks (Rye and Faugli catchment, which is situated in an inner branch of the 1994). Sognefjord, ca 140 km from the coast of western NorDue to the favourable climate and bedrock, Aurland way (Fig. 1). The catchment has a total area of 762 is one of the richest catchments for vascular plant km2, and stretches from sea level to 1764 m elevation. species in western Norway (Odland 1991, 1994). In the Only ca 4% of the catchment area lies > 1500 m, and lowlands, a thermophilous flora and rich deciduous the area > 1700 m is very small (0.1%). Most of the forests occur. Up to ca 800 m species-rich Alnus incuna catchment (ca 60%) is situated between 900 and 1400 forests and meadows are common. The Betulu m, 6% lies between 600 and 900 m, 3% between 300 and pubescens forest limit is generally situated at 1150 m 600 m, and 3% < 300 m (Table 1). Major parts of the elevation. In the low-alpine zone (1 150-1400 m), area < 100 m have been cultivated. The main Au- mountain heaths and grasslands are frequent. Complete rlandsdalen valley, and its tributaries, are deeply cut flora lists and other characteristics of the Aurland flora into the highland plateau (Faugli 1994a). The valley and vegetation are given in Odland (1990, 1994). Desides are steep, but most have a closed vegetation cover. tails of the Aurland catchment and its geology, landThe steep topography, especially between 100 and 700 scape, hydrology, and land-use are given in various m, restricts the formation of mires and wetlands. papers in Faugli (1994b). The bedrock consists of three main formations (Bergstrgm 1975, Rye and Faugli 1994). The lowest layer comprises a granitic basement exposed in the lower parts of the main valley. This is covered by Material and methods Cambro-Silurian meta-sediments, mainly mica schist. We divided the area into 18 altitudinal 100 m bands. Their thickness varies from a few metres to several Records of all vascular plant species were compiled for hundred metres, but they are exposed from sea level up each band (cf. Ogden and Powell 1979). The floristic to ca 1500 m. The meta-sediments are located on both data were collected during extensive field surveys by flanks of the lower end of the Aurland valley, here AO, over ca 40 fieldwork days, within the Aurland river reaching sea level. They rarely reach > 1400-1500 m catchment during the summers of 1988- 1993. A Thommen altimeter and barometer (with an accualtitude. The Jotun nappe overlies the Cambro-Silurian 10 m) was used regularly to verify the sediments. The nappe has a varied composition, con- racy of

550

7030' .,.........,, ;.' $.

Fig. 1. The Aurland study area and (insert) its position in southern Norway (from Faugli 1994b). ECOGRAPHY 2215 (1999)

M 00 N m m m t- t- 00 00 IA d t- d o mmddmdwIAdm-mwIANwm-

"mNr4"m"N----

I-r-o-IAMdbNNb-bo-00

---N"

altitudinal position of all plant records. The altimeter was calibrated at known altitudes whenever possible during each day's field work. Floristic inventories were carried out in different parts of the area, and for each visit flora lists were made. Species lists were compiled for each 100 m interval (cf. Wilson and Shmida 1984, Shmida and Wilson 1985, Mirek 1990, Auerbach and Shmida 1993). Species previously recorded in recent years by other investigations (e.g. available flora lists and herbarium records) were also included. Available floristic data used are given in Odland (1990). One cannot, however, assume that all species in all bands have been found. Vascular plant taxonomy follows Lid and Lid (1994). Climatological investigations have shown that southsouthwest-facing (225") slopes are ecologically the most favourable in northern temperate areas (Dargie 1984), and species generally have different vertical distributions on different aspects. For example, it has been shown that the birch forest-limit in the eastern parts of western Norway generally lies 100-150 m higher on south-facing slopes compared to north-facing slopes, whereas the difference between south- and west-facing slopes is mostly < 60 m (Ve 1940, Odland 1996). In order to minimise the different distribution patterns of species resulting from different aspects, we stratified our sampling area to include only south- or west-facing slopes. The main aim of our study is to investigate the altitudinal distribution patterns in relation to regional climate. The distributions of the species are therefore assumed to be continuous between their lowest and highest altitudinal records. This is because species may be locally absent from intermediate altitudes because of local topographic or edaphic factors (cf. Wilson and Shmida 1984), which may therefore locally reduce their potential climatic distribution. Species with sporadic and rare occurrences along the main paths and roads, streams (mostly alpine plants), and by the coast (halophytes) are omitted because their occurrences are unlikely to be determined by regional climate. There are no temperature stations within the study area, and therefore temperature conditions within each altitudinal band must be interpolated from neighbouring stations. On the basis of data from 6 stations (all within a radius of ca 60 km) in the inner parts of Sognefjord (cf. Aune 1993) the mean July temperature (reduced to sea level) is 15.0+0.4"C, and the mean January temperature is - 2.0 1.9"C. Lapse rates of 0.57"C for mean July temperatures and of 0.44"C for mean January temperatures for a 100 m change in altitude were applied (Laaksonen 1976). As pointed out by Utaaker (1987) from a neighbouring valley, we can expect some small deviations from the interpolated values because of variations in local topography. There are three precipitation stations within the area, situated at 15, 647, and 810 m (cf. Ferrland 1993a). In

rRAPHY 22

1999)

55 1

the mountains there is no station, but according to the Norwegian annual precipitation map (Ferrland 1993b), the highest mountains in the study area receive an annual precipitation of ca 2000 mm. A second-order polynomial regression model (r2= 0.99) of the measured precipitation data from four altitudes was used to estimate annual precipitation for each 100 m altitudinal band (Table 1). A second-order polynomial provided the minimal adequate statistical model for the available precipitation data. Although we have estimated and subsequently used mean July and January temperatures and annual precipitation for each altitudinal band, we are not assuming that plant richness or species distributions are controlled directly by these parameters. Plants presumably respond to parameters like the duration of the growing season, maximum or minimum temperatures, total summer warmth, or annual accumulated temperatures > 5°C (e.g. Woodward 1987, Dahl 1998). These are, in general, highly correlated to simpler parameters like mean July temperature. Absolute minimum temperatures are probably also important in influencing plant distribution (Woodward 1987, 1988, Dahl 1998), and we use mean January temperature as a surrogate for minimum temperature (Dahl 1998). We use total annual precipitation as a crude index of moisture availability (Woodward 1987). As area is often an important predictor of species richness (e.g. Connor and McCoy 1979, Williamson 1981), an attempt was made to estimate the area of the 18 altitudinal bands. The catchment area is ca 762 km2 and relatively large parts consist of lakes (especially between 1400 and 1500 m) and glaciers (Fig. 1). The areal estimates were based on averages of several transects stretching from sea level to the mountain summits (see Table 1). Differences in species composition and species number between each altitudinal band were compared here in three different ways: 1) differences in species numbers; 2) species turnover sensu van Steenis (1984), i.e. counting the number of species that begin or end their ranges in each altitudinal band; and 3) comparisons between the bands based on their total floristic composition. The abundance or frequency of the species was ignored, and only presencelabsence data were used. “Species diversity” or species richness is based here on direct species number which, according to Peet (1974), is the simplest, most practical, and least ambiguous measure of species diversity. In addition total species richness in each band was predicted from the climatic and areal data for the study area. Variation in species number was examined initially by simple scatter plots (Fig. 2) of species richness in each altitudinal band plotted against altitude, mean July temperature, mean January temperature, annual precipitation, and the ‘YO area of the altitudinal bands. A LOESS scatter-plot smoother (Cleveland 1979, 552

Trexler and Travis 1993) was fitted to the plots with a span of 0.35 to highlight the major trends in each plot in the absence of any a priori statistical model. In these and all other plots involving altitude, the altitude plotted is the mean for each altitudinal band, namely 50, 150, ..., 1750 m. The statistical power of mean July temperature as a predictor of species richness over the entire altitudinal gradient was assessed by a linear regression between species number and mean July temperature. The apparent standard error or root mean square error (RMSE) and the coefficient of determination (r2) between the observed species richness and the estimated species richness based on the regression model were calculated. As RMSE is invariably under-estimated and 3 over-estimated when based solely on the data used in the regression (Martens and Naes 1989, Birks 1995), the cross-validation technique of “leave-one-out” jack-knifing (Martens and Naes 1989, ter Braak and Juggins 1993), was used to derive more realistic estimates of RMSE and r2. In this cross-validation, the regression was applied to the data containing n observations n times, but with one observation left out in turn. The regression equation was thus based on (n - 1) observations and was then applied to predict the species number of the one observation omitted from the regression, thereby giving a predicted value for that observation. By subtracting this predicted value from the observed value, a more realistic prediction error for that observation can be obtained. These errors were then accumulated to derive a leave-one-out RMSE of prediction (RMSEP). The coefficient of determination (r2) can also be estimated between the observed and the predicted richness values (Birks 1995). The predictive power of the three climatic variables (Table 1) for species richness over the entire altitudinal gradient was evaluated using partial least squares (PLS) regression (Martens and Naes 1989, Birks 1995, 1996). PLS is a regression technique that guards against multicollinearity among the predictor variables by selecting a small number of uncorrelated orthogonal components that are selected to maximise the covariance with the response valuable, in this case species richness. The appropriate number of PLS components was estimated by leave-one-out cross-validation as being the number of components that gives the lowest RMSEP and/or the highest r2 in cross-validation. As area is often a good predictor of species richness (e.g. Connor and McCoy 1979, Williamson 1981, McGuiness 1984), the statistical power of the ‘YOarea of the different altitudinal bands was assessed by a linear regression between species number and the ‘YOarea with cross-validation to derive a leave-one-out RMSEP and r2. The ability of the three climatic variables (Table 1) plus ‘YO area to predict species richness was evaluated using PLS and leave-one-out cross-validation to estimate the RMSEP and r2 for the most parsimonious ECOGRAPHY 225 (1999)

pLS model. All PLS computations were done using the program CALIBRATE (ver. 0.81) by S. Juggins and C. J. F. ter Braak. In an attempt to test if there are any statistically significant discontinuities along the altitudinal gradient at Aurland, we applied the biotic boundary technique of McCoy et al. (1986) to the Aurland data. This technique determines the floristic similarities between altitudinal bands and detects if any statistically significant boundaries occur between pairs of adjacent bands

on the basis of the observed similarities in floristic composition. Initially the number of species expected to be in common between bands under a random allocation model was computed (Connor and Simberloff 1978, Raup and Crick 1979). Each of the 444 species found along the altitudinal gradient was randomly assigned to T altitudinal bands where T is the number of bands in which the species is actually found. This was repeated 100 times and the number of species in common between all bands was recorded for each simula-

500

1000

1500

5.0

2000

-8.0

-6.0

I -4.0

800

-2.0

Jan. mean temperature (â&#x20AC;&#x153;C)

Fig. 2. Number of vascular plant species within 100 m altitudinal bands plotted against a) altitude, b) mean July temperature, c) mean January temperature, d) annual precipitation, and (e) percentage area of altitudinal band. The fitted lines are LOESS smoothers (span = 0.35). ECOGRAPHY 2 2 5 (1999)

9.0

11.0 13.0 15.0

July mean temperature (â&#x20AC;&#x153;c)

Altitude (m)

-10.0

7.0

900 1200 1500 1800 2 00

Annual precipitation (mm)

Band area (%) 553

tion. From the 100 simulations an estimate was derived of the expected number of species in common and their associated variance by chance alone. The expected numbers were then compared with the observed number of species in common between altitudinal bands to determine if the observed number of species in common is significantly ( a = 0.05) greater ( + ) or less ( - ) than expected by chance alone. A matrix of all statistically significant positive ( + ) and negative ( - ) similarities was then constructed between all pairs of altitudinal bands. In these simulations a species could only be assigned to the same number of bands that the species actually occurs in, in other words the row sums of the simulated species occurrence x band matrices were constrained to match the observed row sums. No constraints were imposed on the number of species occurring in a band, in other words the row and column sums were not constrained (cf. Brunaldi 1980, Snijders 1991, Manly 1995, Sanderson et al. 1998). Potential statistically significant boundaries between the altitudinal bands were then located using the matrix of statistically significant coefficients between the altitudinal bands (McCoy et al. 1986). The matrix of floristic similarities between bands was examined for clusters of bands between which the floristic differences are greater or smaller than expected by chance. Two types of boundaries between bands can be identified, depending on how the similarity matrix is viewed (McCoy et al. 1986). If interest centres on identifying bands that are very different from one another in overall floristic similarity, a boundary would be identified where there are lower-than-expected similarities ( - ) between groups of bands on both sides of the boundary, socalled “strong” boundaries (sensu McCoy et al. 1986). For a strong boundary, breaks in a series of - similarities within the similarity matrix are of prime concern. If interest is on grouping bands that are similar to one another, a boundary would appear as a break in a series of greater-than-expected similarity values ( ), a so-called “weak” boundary (sensu McCoy et al. 1986). similarities within the Here a break in a series of similarity matrix is the criterion for identifying a weak boundary. McCoy et al. (1986) call this type of boundary “weak” because a high degree of similarity can potentially exist between bands on either side of the boundary. All possible boundaries between pairs of adjacent bands were considered and Dweakand Dsfrong values were calculated between all pairs of adjacent bands. The largest Dstrongor mean Dweak values were identified. The statistical significance of the best single boundary highlighted by high Dweakor Dstrongvalues (McCoy et al. 1986) was assessed by G-tests of independence (Sokal and Rohlf 1995) in which the distribution and - (scaled by the proportion of bands of represented) between group A x B and groups A x A and B x B combined were compared, where group A contains all the bands to one side of the boundary and

554

group B contains all the bands to the other side of the boundary. Secondary boundaries can be identified by bimodality in the plots of Dstrongand Dweakvalues between pairs of adjacent bands. McCoy et al. (1986), Botts and Cowell (1988), and Botts and McCoy (1993) give further details of the computations involved and several ecological examples. The boundary analyses were implemented by the program BOUNDARY, extensively modified by H. J. B. B. and J. M . Line from the original program of McCoy et al. (1986). Detrended correspondence analysis (DCA) (Hill and Gauch 1980) was applied to the floristic data to assess the floristic turnover or compositional change along the altitudinal gradient estimated as standard deviation (SD) units of turnover (detrending by segments, nonlinear rescaling). As the floristic turnover is > 2-3 SD, unimodal-based ordinations (ter Braak and Prentice 1988) and their related divisive classification techniques were then used to detect any major patterns in floristic composition within the data that might help identify potential floristic discontinuities and changes in floristic composition along the altitudinal gradient. Correspondence analysis (CA) was used to ordinate the floristic data without any altitudinal constraint, whereas constrained or canonical correspondence analysis (CCA) was used to ordinate the floristic data constrained by altitude (ter Braak 1987). Because of the potential influence of the different areas of the altitude bands on floristic composition, partial CA and CCA (ter Braak and Prentice 1988) were also made in which the effect of area was partialled out as a covariable prior to the ordination analysis. All these analyses were made using the program CANOCO 3.12a (ter Braak 1988, 1990) with strict convergence criteria and no down-weighting of rare species. The divisive classificatory procedures of two-way-indicator species analysis (TWINSPAN, Hill 1979) that is based on an initial CA ordination of the data and constrained indicator species analysis (COINSPAN, Carleton et al. 1996) that is based on an initial CCA ordination of the data, in this case constrained by altitude, were used to identify floristic discontinuities within the floristic data. The analyses were made using the programs TWINSPAN 2.2a and COINSPAN 0.56X with strict convergence criteria.

Results Variation in species richness along the altitudinal gradient Variation in the total number of vascular plant species between each altitudinal band is shown in Fig. 2a (cf. Table 1). Within the Aurland data there is little variation in richness from sea level, with 236 species, to 900- 1000 m with 227 species. The highest species numECOGRAPHY 2 2 5 (1999)

Fig. 3. Number of vascular plant species within 100 m altitudinal bands plotted against altitude. Circles are the results from Aurland. Solid circles denote bands above the birch forest limit. Data from the Jotunheimen mountains in southern Norway are shown as asterisks (data from Jsrgensen 1932). To the right, the number of species in relation to altitude in similar investigations in the Alps. Squares give the results from Rube1 (191 1) and triangles the results from Raunkiaer (1908).

350 -

300

250

c .LI! tn

.-a0, a,

200

150

+?Y

100

i I

lo00

JOTUNHEIMEN

2000

3000

4Ooo

Altitude (m) ber (263) is recorded between 600 and 700 m. Above 600 m there is a gradual decrease up to 1500 m. Above 1500 m there is a slightly steeper decrease in species richness. The relationship between species richness and the three climatic variables in Table 1 (Figs 2b-2d) show little change in species richness at high mean July (1 1°C or more) and January temperature ( - 5°C or more) temperatures or at low annual precipitation (1000 mm or less). At lower temperatures and higher precipitation, species richness decreases with decreasing July and January temperatures and increasing annual precipitation (Figs 2b-2d). There is no clear relationship between species richness and the % area of the altitudinal bands (Fig. 2e). The patterns of species richness in relation to altitude from Aurland are compared in Fig. 3 with comparable floristic data sampled within altitudinal bands from the Jotunheimen mountains in southern Norway (Jmgensen 1932) and from the Alps (Raunkirer 1908, Rube1 1911).

Species turnover along the altitudinal gradient Figure 4 shows the number of species that have their altitudinal distribution limits within the different altitudinal bands. Upper-limit species turnover has one major peak, at 1500 m. The highest numbers of plants with their lower distribution limits are found within the 100-200 m, 600-700 m, 700-800 m, and 800-900 m bands. Total species turnover has three “peaks”, with ECOGRAPHY 2 2 5 (1999)

major species turnovers between 100 and 200 m, between 600 and 700 m, 700 and 800 m, 800 and 900 m, and between 1400 and 1500 m. Particularly low turnover rates occur between 200 and 600 m and between 1200 and 1300 m. The total number of species in each band is also shown in relation to the forest-limit ecotone on Fig. 4.

Predicting species richness from climatic variables The apparent root mean square error (RMSE) and r2 for the regression of species richness in relation to mean July temperature are 40.42 and 0.73, respectively, whereas the root mean square error of prediction (RMSEP) and r2 based on cross-validation are 46.89 and 0.65, respectively (p = 0.01). These results and a plot of predicted species richness against observed species richness (Fig. 5a) show that mean July temperature by itself is a relatively poor predictor of species richness over the full altitudinal extent of Aurland. It over-estimates species richness at the low and high ends of the observed richness gradient and under-estimates in the range of 120-250 species. The three climatic variables (mean July and January temperatures and total annual precipitation), when considered together in PLS, provide a good statistical prediction of species richness in a 2-component PLS model (RMSE = 9.56, apparent r2 = 0.99, RMSEP = 12.03, r2 = 0.98, p = 0.01) (Fig. 5b). In this model there is a small bias in the predictions with a tendency to over-estimate by 10-15 species at the species-poor and the species-rich ends of the richness gradient.

555

The area of the altitudinal band is a very poor predictor of species richness (Fig. 5c) (RMSE = 77.44, apparent r2=0.24, p=O.54). If area is included as a predictor along with the three climatic variables in a 3-component PLS model (RMSE = 9.65, apparent r2 = 0.99, RMSEP = 12.04, r2 = 0.98, p = 0.01), the resulting predictions (Fig. 5d) are, not surprisingly, very similar 80-

, I

to those from the PLS model using the three climatic variables only (Fig. 5b). Area of the altitudinal bands is thus a very poor predictor species richness in the data set. If the effects of area are partialled out as a covariable first, the resulting predictive models for species richness are similar to, or slightly better than, those in which the effects of area are ignored or included as a predictor, with small increases (ca 5%) in r2 and small decreases (ca 2%) in RMSEP.

70 -

60-

Species number as a function of mean July temperature within the 700-1500 m range

5040 -

Species richness along part of the altitudinal gradient can be expressed directly as a function of mean July temperature. Figure 5a indicates that for the 7 altitudinal bands between 700 and 1500 m there is a good relationship between observed and predicted species richness, and that this relationship is strongly correlated with mean July temperatures (Fig. 2b). Regression analysis between the number of vascular plant species (NA)for the bands within the interval 700-1500 m and mean July temperature (Tj,l,) gives a significant regression model (Fig. 6) with the following equation:

3020 10-

6050-

403020 -

NA= 30.58TJ,,,, - 7.47 (n = 7, rz = 0.97, p < 0.001) 10-

40-

30-

20-

10-

=; 250-

(1)

200-

This suggests that there is a general decrease in the number of vascular plant species of ca 30 species per 1째C decrease in mean July temperature in our study area and that no species of vascular plants occurs when mean July temperatures are < 2.4"C. In Fig. 6 the floristic data from Jotunheimen (Jerrgensen 1932, see Fig. 3) are also plotted as a function of mean July temperature based on climatic data from the nearest meterological station (VBgimo, cf. Aune 1993). A regression analysis of the data, ranging from 1500 to 2300 m, gives the following equation: N,

= 29.11TJ,,,-65.6

(n=9, $=0.99, p<O.OOl)

(2)

150-

This also suggests a general decrease in vascular plant richness of ca 30 species per 1째C decrease in mean July temperature.

loo500-

250

500

750

I loo0

Upper altitude of band (m) Fig. 4. Variation in vascular plant species richness and species turnover along the Aurland altitudinal gradient. The turnover of upper- and lower-limits of species, total turnover, and total species richness are plotted against the upper altitude of the altitudinal bands. The vertical line shows the position of the forest-limit ecotone.

556

Floristic discontinuities between the altitudinal bands The matrix of statistically significant similarities between all pairs of altitudinal bands is summarised in Table 2. There are 62 observed similarities that are significantly greater than the expected similarity under ECOGRAPHY 2 2 5 (1999)

-H v)

200-

150-

d 0 5

100

E a

50-

0 1. 50

1 0 0 1 5 0 2 0 0 2 5 0 3 0 0

I 50

Observed species richness

I 100

I I I 1 5 0 2 0 0 2 5 0 :

Observed species richness

(d) xa

200-

150-

100

100-

a 504

oy 0

1 0 0 1 5 0 M o 2 5 x a

Observed species richness

50-

I/ 0

100

1 5 0 2 0 0

250

Observed species richness

Fig. 5. a) Predicted number of vascular plant species based on a linear regression of species richness and mean July temperature, b) predicted number of vascular plant species based on partial least squares regression of species richness and mean January temperature and July temperature and annual total precipitation, c) predicted number of vascular plant species based on a linear regression of species richness and percentage area of the altitudinal bands, and d) predicted number of vascular plant species based on the three climatic variables and percentage area of the altitudinal bands, all plotted against observed species richness.

a random allocation model and 66 observed values that are significantly less than expected. Twenty-five are not significantly different from random expectation. The highest Dstrong and Dweakvalues (McCoy et al. 1986) occur between the 600-700 and 700-800 m bands (0.534 and 0.537, respectively) (Fig. 7). Both values are statistically significant according to a Gtest of independence (p < 0.001). With the exception of the high Dstrongvalues between the 600-700 and ECOGRAPHY 2 2 5 (1999)

700-800 m bands (0.483) and between the 400-500 and 500-600 m bands (0.419) (Fig. 7), none of the other DStrOng values ( - 0.194-0.297) are statistically significant (p > 0.05). The only other significant Dweak values (p < 0.05) are between the 400-500 and 500600 m bands (0.465) and the 600-700 and 700-800 m bands (0.517) (Fig. 7). The results thus suggest statistically significant differences in the number of species in common above and below 400 and 800 m. 557

3oo 250

.-2

I I I I I I

mmv, lI2IZZZ++++

! l

'0 150

v,mm

I I I I I I IZZZ+++++

I l l l l l l $ ? + + + + + + 0.0

2.5

5.0

7.5

10.0

12.5

15.0

July mean temperature ("c) Fig. 6. Number of vascular plant species in relation to mean July temperature along two altitudinal gradients. The regression lines for Aurland (NA), based on the altitudinal bands between 700 and 1500 m a d . (solid circles) are compared to the relationship found by regression analysis of data from the Jotunheimen mountains (N,) (asterisks) (Jargensen 1932). The open circles represent Aurland data not included in the N, regression model.

f 2

I I I I I I$?+++++

m d

I I I I I I$?$?++++

2 I 1 I I ll$?2+++

I I I I I

Differences in floristic composition between the altitudinal bands

I2+++

I I I I I%?++

The first DCA axis (eigenvalue 0.48, variance explained = 37.6% of the total inertia) has a long gradient length of 4.39 SD units, compared to DCA axis 2 (eigenvalue = 0.08, variance explained = 5.9% of the total inertia, gradient length = 1.14 SD). In view of this long gradient length in the first gradient of variation in the floristic data, unimodal-based methods of ordination (CA) and constrained ordination (CCA) and classification (TWINSPAN, COINSPAN) were used to detect patterns of variation in the floristic composition between the altitudinal bands. The first CA axis (eigenvalue = 0.48, 37.6% of the total inertia) is highly correlated with altitude (r = 0.99). When the CA axis 1 scores are plotted in relation to altitude (Fig. Sa), there is a change from negative to positive CA scores between the 600-700 and 700-800 m bands and a marked change in the CA scores (as shown by the LOESS smoother on Fig. 8a) between the 1400-1500 and 1500-1600 m altitudinal bands. The second CA axis (eigenvalue=0.22, 17.1% of the total inertia) is poorly correlated with altitude (r = - 0.07). When the effects of altitudinal band area are partialled out as a covariable in a partial CA, axis 1 scores (eigenvalue = 0.29, 29% of the total inertia) remain highly correlated with altitude (r = 0.91). The resulting plot of partial CA axis 1 scores (Fig. 8b) shows the first change from negative to positive scores between the 400-500 and 500-600 m bands, little variation in scores until the 1400-1500 m band and a major change between the 1400- 1500 and 1500- 1600 m bands. 558

m 00

I I I

I22++

I I25?+++

22++++ W vl

2++++

++++ +++ m N

ECOGRAPHY 2 2 5 (1999)

â&#x20AC;&#x2122;

-0.2 Partial CA =A

TWINSPAN

scores gradually increase to the 900- 1000 m band, drop to the 1400-1500 m band, and increase sharply between the 1400-1500 and 1500-1600 m bands. A TWINSPAN analysis places the first major dichotomy in the floristic data between the 1000-1100 and 1100-1200 m bands, with Calluna uulgaris as the indicator species for the lower altitude group. Secondlevel divisions are between the 600-700 and 700-800 m bands (with Andromeda polifolia as the indicator species for the upper altitude group) and between the 1400-1500 and 1500-1600 m bands (with Alchemilla alpina as the indicator species for the group containing the 1100-1500 m bands) (Fig. 7). A constrained indicator species analysis (COINSPAN) with altitude as the only environmental variable positions the first major dichotomy between the 1400-1500 and the 1500-1600 m bands (Fig. 7) and identifies Anthoxanthum odoratum as the indicator species for the 0-1500 m group of bands. A second division is placed between the 900- 1000 and the 1000-1100 m bands with Agrostis stolonifera as the indicator species for the 0-1000 m group of bands. Many of the numerical methods used to identify possible boundaries or areas of marked differences in floristic composition between the altitudinal bands suggest major changes between the 400-500 and 500600 m bands or the 600-700 and the 700-800 m bands (Fig. 7) and between the 1400-1500 and 15001600 m bands (Fig. 7). Only the TWINSPAN and COINSPAN results identify differences in floristic composition between the 900- 1000 and 1000- 1100 m bands (COINSPAN) or between the 1000-1100 and 1100-1200 m bands (TWINSPAN). These are the bands at or near the forest-limit ecotone.

Discussion Interpretation of the altitudinal gradient

When the floristic data are constrained by altitude in a CCA, axis 1 (eigenvalue=0.47, 37% of the total inertia) is highly correlated with altitude (r = 0.99). The CCA axis 1 scores change from negative to positive between the 600-700 and 700-800 m bands (Fig. 8c) and there is a change in the gradient of scores between the 1400-1500 and 1500-1600 m bands. CCA axis 2 is similar to CA axis 2 with an eigenvalue of 0.22 representing 17.2% of the total inertia. In a partial CCA with YOarea of the bands partialled out as a covariable, the first CCA axis has an eigenvalue of 0.27, captures 27% of the total inertia, and is highly correlated to altitude (r = 0.99). The partial CCA axis 1 scores (Fig. 8d) change from negative to positive between the 400-500 and 500-600 m bands. The ECOGRAPHY 225 (1999)

We have tried to minimise subjectivity and bias as far as possible in the data collection. The data sampling has been stratified by the selection of only south- and west-facing slopes. Each altitudinal belt investigated is relatively large, and includes major differences in topographic and edaphic conditions. The diversity measured here can be described as beta diversity (sensu Whittaker 1972, 1975), since we are considering compositional differentiation along an altitudinal gradient that represents a complex environmental gradient. Beta diversity is defined as the â&#x20AC;&#x153;extent of species replacement or biotic change along environmental gradientsâ&#x20AC;? (Whittaker 1972). Theoretically, all species are assumed to have more or less well defined upper and lower distributional limits along the altitudinal gradient (unimodal response) (Whittaker 1975, Wilson and Mohler 1983). 559

Altitudinal limitations and edaphic-topographic differences within the area investigated should be considered when the results of our investigation are discussed. The altitudinal gradient, and thus the climatic gradient, is limited. One should not expect therefore that all the species have climatic distributional limits within the area. The mountains are not high enough to include the upper limits of all species present, and probably not all the potential high-alpine plants. We assume that the reason why the close linear relationship between species number and mean July temperature occurs within the range 700-1500 m is due to the general terrain steepness. Below 700 m there is relatively little variation in biotopes, and moist forests and meadows, aquatic biotopes, and mires are almost completely missing. The generally poor relationship between area and species number, especially between 200 and 700 m, indicates that area per se is a poor predictor of species richness in the present study. Environmental variation within each band is probably much more important than area itself. In order to minimise the effects of differences in topographic and edaphic factors between the bands, species are assumed to have a continuous distribution between their uppermost and lowermost records. Therefore the data cannot be used to assess the effects of topographic and edaphic factors on plant diversity and distribution.

”-

Several different approaches have been used to study the influence of altitude on floristic composition, and this makes it difficult to compare results from one area to another. It has been proposed that the results depend, to some extent at least, on the methodology used (Auerbach and Shmida 1993). If the goal of the investigation is to obtain a representative or a near total floristic list for a particular area, all different biotopes must be studied, and much time must be spent on data collection. The altitudinal change in vegetation composition has also been shown to be strongly affected by substrate in addition to the general climatic gradients (Mooney et al. 1962, Ogden and Powell 1979). The results of any investigation should therefore take into account topographic, geologic, and climatic factors (cf. Boyce 1998). Altitude may influence plant growth indirectly through strong correlations and interactions with temperature, humidity, snow duration, bedrock, and soil, as well as topographic factors. Whittaker (1956) differentiated these “secondary gradients” from “primary gradients” of climatic factors.

The forest-limit ecotone The ecological and floristic importance of the forest limit in phytogeographical zonations has remained unquestioned, even though its value as an important

4 -2.0

500

1033

1500

2wo

-2

5M)

loo0

1503

2wo

Altitude (m)

Aitiiude (m)

8.0

I I

‘8 0

1.0

-0.0--

-1.0

-2.0

joo0

Altitude (m)

560

-2.0 2wo

,d, Altitude (m)

2wo

Fig. 8. a) Correspondence analysis (CA) axis 1 scores, b) partial CA axis 1 scores, c) canonical correspondence analysis (CCA) axis 1 scores, and d) partial CCA axis 1 scores for the 18 altitudinal bands plotted against altitude (mid-altitude of each bands). The fitted lines are LOESS smoothers (span = 0.35). The dashed lines distinguish between positive (above the line) and negative (below) scores. ECOGRAPHY 22:5 (1999)

ecological boundary has rarely been verified. An interesting question is therefore whether the presence or absence of birch forests in Aurland is associated with any abrupt discontinuity in floristic richness or composition. According to Westman (1983), Kirkpatrick and Brown (1987), and Kirkpatrick et al. (1996), concomitant turnover of plant species with distance is often associated with abrupt changes in climatic or other environmental conditions. However, where a particular life-form or a particular dominant species reaches its environmental limits there may often be a rapid turnover of species (cf. Minchin 1989, Austin 1990, Kent et al. 1997). The classical example of this sort of boundary is the climatic treeline or forest limit (Dahl 1986, 1998, Grace 1989, Stevens and Fox 1991, Hofgaard 1997). The 1100-1200 m band in Aurland can be described as the forest-limit ecotone since in most investigated sites (south- and west-facing slopes), the climatic forestlimit is situated within this interval (Odland 1990). On steep, south-west-facing slopes the forest reaches maximally 1150 m, whereas on west-facing slopes it reaches 1100 m. With the turnover measures used here, the 1100-1200 m band does not represent any major change or turnover in richness (Fig. 4). The CA, CCA, TWINSPAN, COINSPAN and biotic boundaries analyses which are based on the total floristic composition of the different bands, give different results (Figs 7 and 8) but only TWINSPAN and COINSPAN suggest a major floristic transition between 1100 and 1200 m. As can be seen from Fig. 4, there is no major peak in species turnover at the forest-limit ecotone. Auerbach and Shmida (1993) found (in Israel) that species borders were occasionally aggregated within individual altitudinal bands, with a coincidence of aggregated downslope and upslope termini, indicating that an abrupt ecotone occurred around 1200 m. This altitude marked the transition from maquis vegetation to an open deciduous forest. In Malesia, van Steenis (1984) suggested, on the basis of species turnover data, that there are critical altitudes where the floristic composition showed abrupt changes at, for example, 1000, 1500, 2400, and 4000 m altitude. These changes were used to demarcate 5 broad vegetational zones (tropical, submontane, montane, subalpine, and alpine zones). Frahm and Gradstein (19915 found tha; peaks in bryophyte turnover coincided with the major tropical rain forest belts in Columbia, Peru, Borneo, and Papua New Guinea. Our results from Aurland suggest that the forest limit is not associated with any major floristic change, either in species turnover or in species richness, but is associated with some change in overall species composition. Hofgaard (1997) similarly found little difference in beta diversity or floristic composition 300 m above or below the tree line in the central Norwegian mountains. ECOGRAPHY 2 2 5 (1999)

A possible explanation for the lack of any major floristic change at the forest limit is historical. During the last 9000 yr of the Holocene, forest limits have been 200-300 m higher than today in western and central Norway (Birks et al. unpubl.), as evidenced by macrofossils of Betula pubescens and Pinus sylvestris preserved in lake sediments. Forest limit appears to have descended to present-day altitudes comparatively recently, at the time of the so-called â&#x20AC;&#x153;Little Ice Ageâ&#x20AC;? of 1700- 1800 AD. All the available palaeoecological data from western and central Norway suggest that the low-alpine zone between ca 1100 and 1400 m is, in reality, sub-alpine birch woodland but lacking any B. pubescens trees. The floristic composition of dwarfshrubs, grasses, sedges, herbs, and pteridophytes in the low-alpine zone is very similar to the flora 200-300 m below the present forest limit. Moen (1998) lists 17 species with their uppermost limits in the low-alpine zone in Norway as a whole, many of which are primarily woodland plants (e.g. Stellaria nemorum, Gymnocarpium dryopteris, Geum rivale, Cicerbita alpina, Linnaea borealis) or are trees (e.g. Pinus sylvestris, Sorbus aucuparia). It is not known if these species limits are in equilibrium with climate (cf. Boyce 1998, Leathwick 1998) as several species are long-lived perennials that can persist vegetatively for many years or even centuries in unfavourable habitats or climates (e.g. Gorham 1957, Billings and Mooney 1968, Bliss 1971, Summefield 1972, Pigott and Huntley 1980, 1981, Crawford 1989, Molau 1993, 1997). Seed or fruit production by some species may be dependent on particularly favourable or extreme climatic conditions (e.g. Pigott 1970, 1975, Pigott and Huntley 1981, Molau 1997) whereas growth may be relatively unaffected (Callaghan and Emanuelsson 1985, Callaghan et al. 1996). Such species may persist vegetatively (Callaghan et al. 1996) and may only set seed in extreme summers (cf. Murray 1987). There is also a possibility that the geographical ranges of some species are not complete due to lags in dispersal and sporadic seed or fruit production (Pigott 1992), particularly in response to the rapid climatic changes in the last 200-500 yr.

Changes within the alpine zone Within the alpine zones, the major change in species composition and turnover (Figs 4 and 7) occurs at ca 1400-1500 m, the transition from the low-alpine to the mid-alpine zone in this part of Norway. In an analysis of the entire south Norwegian mountain flora, Dahl (1986) showed a major decrease in richness and turnover at ca 1400 m, with 138 species having their altitudinal limit between 1200 and 1400 m whereas only 63 had their limits between 1400 and 1600 m. In the Aurland area many species (92) do not extend above 1400 m or 1500 m including Betula nana, Juniperus 561

communis, the willows S a l k glauca, S. phylicifolia, S. lanata, S. reticulatu, S. myrsinites, the dwarf-shrubs Phyllodoce caerulea, Arctostaphylos alpinus, and A . ucaursi, the tall-herbs Aconitum septentrionale, Stellaria nemorum, and Angelica archangelica, and a wide range of herbs and pteridophytes including Anthyllis vulneraria, Lotus corniculatus, Pinguicula vulgaris, Carex norvegica ssp. norvegica, C. capillaris, C. saxatilis, C. atrofusca, C. atrata, C. brunnescens, Juncus castaneus, J . Jiliformis, Trichophorum cespitosum ssp. cespitosum, Myosotis decumbens, Epilobium IactiJlorum, Equisetum palustre, Gentianella campestris ssp. campestris, Veronica fruticans, Cystopteris fragilis, Eriophorum vaginatum, Astragalus alpinus, Selaginella selaginoides, Dryas octopetala, Erigeron borealis, Pedicularis lapponica, Nardus stricta, Viola bifora, Silene dioica, Gymno carpium dryopteris, Festuca rubra and Deschampsia cespitosa. The maximum number of upper altitudinal limits in any 100 m altitudinal band (Fig. 4) is between the 1400-1500 and 1500-1600 m bands (Table 1). The transition between the low-alpine zone and the midalpine zone thus represents not only a major change in vegetation type and structure, landscape processes, and disturbance regime but also a major discontinuity in floristic composition, turnover, and richness (Dahl 1986).

Changes within the forested zone Below 1100 m, the main peak in species turnover in terms of both upper and lower species limits (Fig. 4) and the main change in floristic composition as detected by CA and CCA (Fig. 8) and by the biotic boundary technique (Fig. 7) occur between 600 and 800 m, corresponding to a mean July temperature of ca 11-11.6"C and an annual precipitation of ca 1000 mm (Table 1). At a regional scale this altitude corresponds to the boundary between the middle and northern boreal zones in southern and western Norway (Moen 1998), even though there are few obvious changes in vegetation structure or composition. Many species attain their upper altitudinal limits in the Aurland area at 700-800 m. These include the trees Betula pendula, Salix caprea, and Ulmus glabra, woodland plants such as Carex digitata, Epilobium montanum, Circaea alpina, Geum urbanum, Impatiens noli-tangere, Vicia sylvatica, Viola mirabilis, and Pteridium aquilinum, grassland plants and species associated with meadows, fields, field margins, and open waste areas such as Artemisia vulgaris, Cardamine flexuosa, Carum carvi, Festuca pratensis, Dactylis glomerata, Lappula deflexa, Erophila verna, Galium aparine, G. verum, Lathyrus pratensis, Tanaceturn vulgare, Leucanthemum vulgare, Linaria vulgaris, Myosotis arvensis, Phleum pratense, Alopecurus geniculatus, Polygonum aviculare, Stellaria media, Tussilago farfara, Verbascum nigrum, V. thapsus, Vicia cracca, V. 562

sepium, and Veronica chamaedrys, and plants of dry rock outcrops and screes including Asplenium septentrionale, A. trichomanes, Origanum vulgare, Arabis glabra, Scleranthus annuus, Arabidopsis thaliana, Arenaria serpyllifolia, Draba incana, Lychnis viscaria, Potentilla argentea, Sedum acre, S. album, and Viola tricolor. Several species of damp habitats also disappear at this altitude, including Drosera rotundifolia, Caltha palustris, Cirsium palustre, and Phalaris arundinacea. A surprisingly large number of species have their lower altitudinal limits between 700 and 800 m. These include species commonly associated with alpine habitats such as Silene acaulis, Thalictrum alpinum, Antennaria alpina, Oxyria digyna, Potentilla crantzii, Juncus triglumis, Sedum villosum, Bartsia alpina, Draba norvegica, Carex atrata, C. norvegica ssp. norvegica, C. capillaris, C. vaginata, Saxijiraga cernua, S. stellaris, Saussurea alpina, Salix lanata, S. glauca, S. reticdata, S. lapponum, S. myrsinites, Pyrola norvegica, Gentiana nivalis, Veronica fruticans, and Astragalus alpinus. Several species of open and/or damp woodland also have their lower limits at ca 700-800 m, including Orthilia secunda, Blechnum spicant, Cornus suecica, Listera cordata, Athyrium distentifolium, Stellaria borealis, and Dryopteris carthusiana. In addition several plants of open grassland or heaths (e.g. Lycopodium clavatum, Luzula multiflora ssp. frigida, L. parvipora, Phyllodoce caerulea, Nardus stricta, Botrychium lunaria, Arctostaphylos uva-ursi) mires or ponds (e.g. Carex paucifora, Betula nana, Juncus Jiliformis, Andromeda polifolia, Ranunculus reptans, Vaccinium oxycoccus ssp. microcarpum, Trichophorum cespitosum ssp. cespitosum), and nutrient-enriched sites (e.g. Chenopodium bonus-henricus) do not descend <700-800 m in our study area. Despite the lack of any obvious change in physiognomy or tree dominants, the major floristic change in the Aurland area occurs between 600 and 800 m and not at the forest-limit ecotone at 1100-1200 m. The dominant trees are Pinus sylvestris that extends to 1000 m, Alnus incana and Sorbus aucuparia that reach to 1100 m, and Betula pubescens and Populus tremula that extend to 1200 m.

Comparisons with other areas The species richness-temperature relationship shown in Fig. 6 shows that at both high ( > 1500 m) and low altitudes ( < 900 m) the number of species recorded is lower than one would predict from the highly significant linear relationship between mean July temperature and species number for the interval 700-1500 m (Fig. 6). This deviation might, in part, be explained as a result of topographic and climatic factors and cultural influences within the study area. The discrepancy between the predicted and observed number of species for higher altitudes might be explained by the relatively low ECOGRAPHY 2 2 5 (1999)

mountains, as areas with mean July temperatures < 5.3”C are absent. Therefore, areas that could include the climatically potential range of high-alpine species and arctic plants are not available in Aurland. It is generally known that the altitude of the forest limit is strongly influenced by the height of the mountains (Dahl 1986). The forest limit may not reach its potential altitudinal limit, on the assumption that lapse rates from high mountains are applicable, in areas where the mountains are low, the so-called Massenerhebung effect (Dahl 1986, 1998). If we assume that vascular plants can only occur in areas where the mean July temperature is above 2.4”C (from eq. l), the mountains in Aurland should be at least 2300 m high for such a “climatic vascular plant limit” to be reached. The observed relationship between species richness and mean July temperature in Aurland is similar to the relationship based on data from the Jotunheimen mountains (Figs 3 and 6). Jerrgensen compiled flora lists for each 100 m altitudinal band based on the uppermost records of the vascular plants on the mountains he investigated. The Jotunheimen data predict that the “climatic vascular plant limit” there is 2.3”C. According to Jerrgensen (1932), only 5 vascular plant species have been found above 2300 m close to the summit of Galdhopiggen (2469 m), the highest mountain in Norway, and the highest localities with vascular plants are at 2370 m altitude, corresponding to a mean July temperature of ca 2.2”C. A general relationship between regional floristic richness and summer-temperature regime in the Arctic has been noted in many studies. Rannie (1986) found a statistically significant correlation between the number of vascular plant species and various summer warmth indices at 38 localities in the Canadian Arctic. The highest correlation (r = 0.97) was found between species diversity (NAC) and mean July temperature (TJuly). Within the gradient from 10 to 3°C of mean July temperature, the relation is described by the equation NAC = 24.2Tjuly- 29.1

(3)

Aurland and the Canadian Arctic are very different ecologically, as are the sampling methods of the two studies. Therefore it is interesting that the differences between eqs 1 and 3 are not very great. Equation 3 predicts that no vascular plant species should occur where the mean July temperature is < 1.2”C in the Canadian Arctic, whereas the “climatic vascular plant limit” in the central Norwegian mountains is estimated to be where the mean July temperature is 2.3-2.4”C. This discrepancy could be explained by the general climatic differences between Aurland (mildly oceanic) and the Canadian Arctic (continental) (see Tuhkanen 1984). A continental climate is characterised by a greater annual amplitude in temperature, with higher maximum temperatures during the summer even ECOGRAPHY 2 2 5 (1999)

though the mean July temperature is the same. Since there is not a linear relationship between temperature and plant metabolism (Dahl 1998), the physiological effect of a mean July temperature is different in a continental climate compared to a more oceanic climate. The distribution of plants is mainly determined by maximum or minimum temperatures (Dahl 1998), and therefore one might expect that the distribution limits of vascular plants to be associated with lower values for mean July temperatures in continental areas compared to more oceanic areas (Odland 1996).

Conclusions This study has provided some answers to the questions presented at the outset. 1) Vascular plant species richness decreases with altitude in the Aurland area but only between 600 and 1760 m. It is not known if similar patterns in richness occur in other plant groups, for example bryophytes or macro-lichens. 2) The major changes in floristic composition and turnover occur between 600 and 800 m and between 1400 and 1500 m. These changes are at about the same altitudes as the transition between the middle and northern boreal zones, and the low-alpine and mid-alpine zones, respectively. There is surprisingly little change in floristic richness composition or turnover at or near the forestlimit ecotone, perhaps because the low-alpine zone is very similar floristically to the sub-alpine areas 200- 300 m below the forest-limit ecotone except for the absence of Betula pubescens trees in the low-alpine zone. 3) Species richness over the 0-1760 m altitudinal range is well modelled by partial least square regression by a combination of mean July and mean January temperatures and annual precipitation. Area of the different altitudinal bands does not predict species richness. 4) Species richness over the 700- 1500 m elevational range is well modelled by mean July temperature both in the Aurland area and in the Jotunheimen mountains of central Norway. The resulting statistical models suggest that vascular plant richness decreases by ca 30 species per 1°C decrease in mean July temperature. 5) In theory, if future climate warms by 1-2°C in the summer and if the present-day distributions of species near and above the forest-limit are in equilibrium with climate, it is tempting to think that species richness may increase in the 700-1500 m range, perhaps by ca 15 species in each altitudinal band, representing an increase in richness of ca 8-12%. Saetersdal et al. (1998) predicted a latitudinal increase in species richness of ca 10-20Y0 over much of northern and central Fennoscandia under a future climate with 1-2°C warmer summers and 2-3°C warmer winters, and a 10% increase in annual precipitation. Saetersdal and Birks (1997) (see also Holten 1993) suggested that some plants confined to high altitudes today in Norway (e.g. Cussiope hyp563

noides. Ranunculus ducialis) mav decline in their distribution or even go locally extinct. In considering possible changes in species richness a t or above the forest limit, Woodward (1993) has emphasised, however, the importance of wind speed increasing with altitude on mountains. As plants of high altitudes tend t o be more wind-resistant than lowland species, it is unclear if lowland species will be able to expand altitudinally into windier habitats even if the temperature regime is suitI

able. As wind speed is itself highly correlated with altitude and hence negatively correlated with temperature and positively correlated with precipitation, it is extremely difficult to predict from observational data alone such as ours the likely changes in species richness t o future climate change. Carefully designed field manipulative experiments conducted over many years are urgently required to supplement long-term observations on species richness patterns in relation t o altitude. The Aurland area is well suited for such studies. Acknowledgements - We are grateful to Hilary Birks, John Hodgson and Carsten Rahbek for many, helpful, and penetrating comments and suggestions, to Beate H. Ingvartsen for help with diagrams, and to the Norwegian Water Resources and Energy Administration and the Olaf Grolle Olsens legat for financial support.

References Ahti, T., Hamet-Ahti, L. and Jalas, J. 1968. Vegetation zones and their sections in northwestern Europe. - Ann. Bot. Fenn. 5: 169-211. Auerbach, M. and Shmida, A. 1993. Vegetation change along an altitudinal gradient on Mt Hermon, Israel - no evidence for discrete communities. - J. Ecol. 81: 25-33. Aune, E. J. 1993. Temperaturnormaler. Normalperiode 19611990. - Det norske meteorol. inst. Rapport 2: 1-63. Austin, M. P. 1990. Community theory and competition in vegetation. - In: Grace, J. B. and Tilman, D. (eds), Perspectives on plant competition. Academic Press, pp. 215-238. Austin, M. P. and Smith, T. M. 1989. A new model for the continuum concept. - Vegetatio 83: 35-47. Austin, M. P., Cunningham, R. B. and Good, R. B. 1983. Altitudinal distribution of several Eucalypt species in relation to other environmental factors in southern New South Wales. - Aust. J. Ecol. 8: 169-180. Baruch, Z. 1984. Ordination and classification of vegetation along an altitudinal gradient in the Venezuelan paramos. Vegetatio 55: 115-126. Beak, E. B. 1969. Vegetational change along altitudinal gradients. - Science 165: 981-985. Bergstrsm, B. 1975. Deglasiasjonsforlspet i Aurlandsdalen og omridene omkring Vest-Norge. - Norges Geol. Unders. 317: 33-69. Billings, W. D. and Mooney, H. A. 1968. The ecology of arctic and alpine plants. - Biol. Rev. 43: 481-529. Birks, H. J. B. 1995. Quantitative palaeoenvironmental reconstructions. - In: Maddy, D. and Brew, J. S. (eds), Statistical modelling of Quaternary science data. Tech. Guide 5, Quaternary Res. Assoc., Cambridge, pp. 161-254. Birks, H. J. B. 1996. Statistical approaches to interpret diversity patterns in the Norwegian mountain flora. - Ecography 19: 332-340. Bliss, L. C. 1971. Arctic and alpine life cycles. - Annu. Rev. Ecol. Syst. 2: 405-438.

564

Botts. P. S. and Cowell. B. C. 1988. The distribution and abundance of herbaceous angiosperms in west-central Florida marshes. - Aq. Bot. 32: 225-238. Botts, P. S. and McCoy, E. D. 1993. Delineation of spatial boundaries in a wetland habitat. - Biodiv. Conserv. 2: 351-358. Boyce, R. L. 1998. Fuzzy set ordination along an elevation gradient on a mountain in Vermont, USA. - J. Veg. Sci. 9: 191-200. Brunaldi, R. A. 1980. Matrices of zeroes and ones with fixed row and column sum vectors. - Lin. Alg. Appl. 33: 159-231. Callaghan, T. V. and Emanuelsson, U. 1985. Population structure and processes of tundra plants and vegetation. - In: White, J. (ed.), The population structure of vegetation. Dr. Junk, pp. 399-439. Callaghan, T. V., Carlsson, B. A. and Svensson, B. M. 1996. Some apparently paradoxical aspects of the life cycles demography and population dynamics of plants from the subarctic Abisko area. - Ecol. Bull. 45: 133-143. Carleton, T. J., Stott, R. H. and Nieppola, J. 1996. Constrained indicator species analysis (COINSPAN): an extension of TWINSPAN. - J. Veg. Sci. 7: 125-130. Cleveland, W. S. 1979. Robust locally weighted regression and smoothing scatter plots. - J. Am. Stat. Ass. 74: 829-836. Colwell, R. K. and Hurtt, G. C. 1994. Nonbiological gradients in species richness and a spurious Rapport effect. - Am. Nat. 144: 570-595. Connor, E. F. and Simberloff, D. 1978. Species number and compositional similarity of the Galapagos flora and avifauna. - Ecol. Monogr. 48: 219-248. Connor, E. F. and McCoy, E. D. 1979. The statistics and biology of the species-area relationship. - Am. Nat. 113: 791-833. Crawford, R. M. M. 1989. Studies in plant survival: ecological case histories of plant adaption and diversity. - Blackwell. Dahl, E. 1986. Zonations in Arctic and Alpine tundra and Fellfield ecobiomes. - In: Polunin, N. (ed.), Ecosystem theory and application. John Wiley, pp. 35-62. Dahl, E. 1998. The phytogeography of northern Europe (British Isles, Fennoscandia and adjacent areas). - Cambridge Univ. Press. Dargie, T. C. D. 1984. On the integrated interpretation of indirect site ordinations: a case study using semi-arid vegetation in south-eastern Spain. - Vegetatio 55: 37-55. Druitt, D. G., Enright, N. J. and Ogden, J. 1990. Altitudinal zonation in the mountain forests of Mt Hauhungatahi, North Island, New Zealand. - J. Biogeogr. 17: 205-220. Engelskjsn, T. and Skifte, 0. 1995. The vascular plants of Troms, North Norway. - TROMURA 80: 1-227. Faugli, P. E. 1994a. The landscape of Aurland. - Norsk Geogr. Tidsskr. 48: 9- 11. Faugli, P. E (ed.) 1994b. The Aurland catchment area - the imDact of hvdrooower develoDment. - Norsk Geogr. Tiisskr. 48: f-84: Fsrland. E. J. 1993a. Nedbsrsnormaler. Normaberiode 19611990: - Det norske meteorol. inst. RapporC39: 1-63. Fsrland, E. J. 1993b. Det norske meteorologiske institutt. Arsnedber 1:2 mill, kartblad 3.1.1. - Nasjonalatlas for Norge, Statens kartverk. Frahm, J.-P. and Gradstein, S. R. 1991. An altitudinal zonation of tropical rain forests using bryophytes. - J. Biogeogr. 18: 669-678. Gaston, K. J. 1996. Biodiversity - latitudinal gradients. Prog. Phys. Geogr. 20: 466-476. Gorham, E. 1957. Developments of peatlands. - Quart. Rev. Biol. 32: 145-166. Grabherr, G., Gottfried, M. and Pauli, H. 1994. Climate effects on mountain plants. - Nature 369: 448. Grabherr, G. et al. 1995. Patterns and current changes in alpine plant diversity. - In: Chapin, F. S. and Korner, C. (eds), Arctic and Alpine biodiversity. Ecol. Stud. 113, Springer, pp. 167-181. I

ECOGRAPHY 2215 (1999)

Grace, J. 1989. Tree lines. - Phil. Trans. R. SOC.Lond. B. 324: 233-245. Graves, G. R. 1985. Elevational correlations of speciation and intraspecific geographic variation in plumage in Andean forest birds. - The Auk 102: 556-579. Graves, G. R. 1988. Linearity of geographic range and its possible effect on the population structure of Andean birds. - The Auk 105: 47-52. Hamilton, A. C. 1975. A quantitative analysis of altitudinal zonation in Uganda forests. - Vegetatio 30: 99-106. Hamilton, A. C. and Perrott, R. A. 1981. A study of altitudinal zonation in the montane forest belt of Mt. Elgon, KenyaiUganda. - Vegetatio 45: 107- 125. Hill, M. 0. 1979. TWINSPAN - a FORTRAN program for arranging multivariate data in an ordered two-way table by classification of individuals and attributes. - Cornell Univ., Ithaca. Hill, M. 0. and Gauch, H. G. 1980. Detrended correspondence analysis: an improved ordination technique. - Vegetatio 42: 47-58. Hofgaard, A. 1997. Inter-relationships between tree line position, species diversity, land use and climatic change in the central Scandes Mountains of Norway. - Global Ecol. Biogeogr. Lett. 6: 419-429. Holten, J. I. 1993. Potential effects of climatic change on the distribution of plant species, with emphasis on Norway. In: Holten, J. I., Paulsen, G. and Oechel, W. C. (eds), Impacts of climatic change on natural ecosystems with emphasis on boreal and arctic/alpine areas. Norwegian Inst. for Nat. Res., Trondheim, pp. 85-104. Holten, J. I. and Carey, P. D. 1992. Responses of climatic change on natural terrestrial ecosystems in Norway. NINA Forskningsrapport 29: 1-59. Itow, S. 1991. Species turnover and diversity patterns along an evergreen broad-leaved forest coenocline. - J. Veg. Sci. 2: 477-484. Jergensen, R. 1932. Karplantenes heydegrenser i Jotunheimen. - Nyt Mag. f. naturv. 72: 1-128. Kent, M. et al. 1997. Landscape and plant community boundaries in biogeography. 1 Prog. Phys. Geogr. 21: 315-353. Kirkpatrick, J. B. and Brown, M. J. 1987. The nature of the transition from sedgeland to alpine vegetation in southwest Tasmania. I. Altitudinal vegetation change on four mountains. - J. Biogeogr. 14: 539-549. Kirkpatrick, J. B. et al. 1996. Explaining a sharp transition from sedgeland to alpine vegetation on Mount Sprent, southwest Tasmania. - J. Veg. Sci. 7: 763-768. Kitayama, K. 1992. An altitudinal transect study of the vegetation on Mount Kinabalu, Borneo. - Vegetatio 102: 149- 171. Laaksonen, K. 1976. The dependence of mean air temperatures upon latitude and altitude in Fennoscandia (19211950). - Ann. Acad. Sci. Fenn. A 111 119: 1-18. Lawton, J. H., MacGarvin, M. and Heads, P. A. 1987. Effects of altitude on the abundance and species richness of insect herbivores on bracken. - J. Anim. Ecol. 5 6 147-160. Leathwick, J. R. 1998. Are New Zealand's Nothofagus species in equilibrium with their environment? - J. Veg. Sci. 9: 719-732. Lid, J. and Lid, D. T. 1994. Norsk flora. 6th ed. by R. Elven - Det norske samlaget, Oslo. Manly, B. F. J. 1995. A note on the analysis of species co-occurrences. - Ecology 76: 1109- 1115. Mark, A. F. 1963. Vegetation studies on Secretary Island, Fiordland. N. Z. J. Bot. 1: 188-202. Mark, A. F. and Sanderson, F. R. 1962. The altitudinal gradient in forest composition, structure, and regeneration in the Hollyford Valley, Fiordland. - Proc. N. Z. Ecol. SOC.9: 17-26. Martens, H. and Naes, T. 1989. Multivariate calibration. - J. Wiley. ~

ECOGRAPHY 22:s (1999)

McCoy, E. D. 1990. The distribution of insects along elevational gradients. - Oikos 58: 313-322. McCoy, E. D., Bell, S. S. and Walters, K. 1986. Identifying biotic boundaries along environmental gradients. - Ecology 67: 749-759. McGuiness, K. A. 1984. Equations and explanations in the study of species-area curves. - Biol. Rev. 59: 423-440. Minchin, P. R. 1989. Montane vegetation of the Mt. Field Massif, Tasmania: a test of some hypotheses about properties of community patterns. - Vegetatio 83: 97- 110. Mirek, Z. 1990. Classification of the altitudinal ranges of plant species in central Europe. - In: Bohn, U. and Neuhausl, R. (eds), Vegetation and flora of temperate zones. SPB Academic Publ., pp. 11-19. Moen, A. 1998. Nasjonalatlas for Norge: vegetasjon. - Statens kartverk, Hsnefoss. Molau, U. 1993. Relationships between flowering phenology and life history strategies in tundra plants. - Arct. Alp. Res. 25: 391-402. Molau, U. 1997. Age-related growth and reproduction in Diupensia lapponica, an arctic-alpine plant. - Nord. J. Bot. 17: 225-234. Mooney, H. A,, Andre, G. St. and Wright, R. D. 1962. Alpine and subalpine vegetation patterns in the White Mountains of California. - Am. Midl. Nat. 68: 257-273. Morisset, P., Payette, S. and Deshaye, J. 1983. The vascular flora of the Northern Quibec-Labrador Peninsula: phytogeographical structure with respect to the tree-line. Nordicana 47: 141-151. Murray, D. F. 1987. Breeding systems in the vascular flora of arctic North America. - In: Urbanska, K. M. (ed.), Differentiation patterns in higher plants. Academic Press, pp. 239-262. Murray, D. F. 1997. Regional and local vascular plant diversity in the Arctic. - Opera Bot. 132: 9-18. Odland, A. 1990. Endringer i flora og vegetasjon som felge av vannkraftutbyggingen i Aurlandsdalen. - NINA Forskningsrapport 15: 1-76. Odland, A. 1991. Klassifisering av vassdrag p& Vestlandet ut fra deres floristiske sammensetning. - NINA Forskningsrapport 16: 1-88. Odland, A. 1994. Characteristics of the Aurland flora and consequences of the regulation. - Norsk Geogr. Tidsskr. 48: 29-37. Odland, A. 1996. Differences in the vertical distribution pattern of Betula pubescens in Norway and its ecological significance. - Palaoklimaforschung 20: 43-59. Ogden, J. and Powell, J. A. 1979. A quantitative description of the forest vegetation on an altitudinal gradient in the Mount Field National Park, Tasmania, and a discussion of its history and dynamics. - Aust. J. Ecol. 4: 293-325. Ohsawa, M. 1991. Structural comparison of tropical montane rain forests along latitudinal and altitudinal gradients in south and east Asia. - Vegetatio 97: 1-10, Ozenda, P. 1988. Die Vegetation der Alpen im europaischen Gebirgsraum. - Gustav Fischer, Stuttgart. Patterson, B. D. et al. 1998. Contrasting patterns of elevational zonation of birds and mammals in the Andes of southeastern Peru. - J. Biogeogr. 25: 593-607. Peet. R. K. 1974. The measurement of soecies diversity. Annu. Rev. Ecol. Syst. 5: 285-307. Peters. R. H. and Darling, J. D. 1985. The greenhouse effect and nature reserves. C'BioScience 35: 707-717. Pigott, C. D. 1970. The response of plants to climate and climatic change. - In: Perring, F. H. (ed.), The flora of a changing Britain. Classey, London, pp. 32-44. Pigott, C. D. 1975. Experimental studies on the influence of climate on the geographical distribution of plants. Weather for 1975: 82-90. Pigott, C. D. 1992. Are the distribution of species determined by failure to seed set? - In: Marshall, C. and Grace, J. (eds), Fruit and seed production. Symp. SOC.Exp. Biol. 47, pp. 203-216.

565

Pigott, C. D. and Huntley, J. P. 1980. Factors controlling the distribution of Tiliu cordaltu at the northern limits of its geographical range 11. History in the north-west England. - New Phytol. 84: ,145-164. Pigott, C. D. and Huntley, J. P. 1981. Factors controlling the distribution of Tiliu cordutu at the northern limits of its geographical range 111. Nature and causes of seed sterility. - New Phytol. 87: 817-839. Rahbek, C. 1995. The elevational gradient of species richness: a uniform pattern? - Ecography 18: 200-205. Rahbek, C. 1997. The relationships among area, elevation, and regional species richness in neotropical birds. - Am. Nat. 149: 875-902. Rannie, W. F. 1986. Summer air temperature and number of vascular species in Arctic Canada. - Arctic 39: 133-137. Raunkirer, C. 1908. Livsformernes statistik. - Bot. Tidsskr. 29. Raup, D. M. and Crick, R. E. 1979. Measurement of faunal similarity in paleontology. - J. Paleontol. 53: 1213-1227. Rodhe, K. 1992. Latitudinal gradients in species diversity: the search for the primary cause. - Oikos 65: 514-527. Rodhe, K. 1997. The larger area of the tropics does not explain latitudinal gradients in species diversity. - Oikos 79: 169-172. Rosenzweig, M. L. 1992. Species diversity gradients: we know more and less than we thought. - J. Mammal. 73: 715730. Rosenzweig, M. L. and Sandlin, E. A. 1997. Species diversity and latitudes: listening to area's signal. - Oikos 80: 172176. Riibel, E. 1911. Pflanzengeographische Monographie des Berninagebietes. - Engler's bot. Jahrb. 47, Leipzig. Rye, N. and Faugli, P. E. 1994. Geology and geomorphology of the Aurland district. - Norsk Geogr. Tidsskr. 48: 13-22. Sanderson, J. G., Moulton, M. P. and Selfridge, R. G. 1998. Null matrices and the analysis of species co-occurrences. Oecologia 116: 275-283. Sretersdal, M. and Birks, H. J. B. 1997. A comparative ecological study of Norwegian mountain plants in relation to possible future climatic change. - J. Biogeogr. 24: 127152. Sztersdal, M., Birks, H. J. B. and Peglar, S. M. 1998. Predicting changes in Fennoscandian vascular-plant species as a result of future climatic change. - J. Biogeogr. 25: 111122. Shipley, B. and Keddy, P. A. 1987. The individualistic and community-unit concepts as falsifiable hypotheses. - Vegetatio .. . .. 69: . .. 47-55 . . Shmida, A. and Wilson, M. V. 1985. Biological determinants of species diversity. - J. Biogeogr. 12: 1-20. Snijders, T. A. B. 1991. Enumeration and simulation methods for 0-1 matrices with given marginals. - Psychometrica 56: 397-417. Sokal, R. R. and Rohlf, F. J. 1995. Biometry (3rd ed.). - W. H. Freeman. Stevens, G. C. and Fox, J. F. 1991. The causes of treeline. Annu. Rev. Ecol. Syst. 22: 177-191. Summerfield, R. J. 1972. Biological inertia - an example. - J. Eco~.60: 793-798.

566

ter Braak, C. J. F. 1987. Ordination. In: Jongman, R. H. G., ter Braak, C. J. F. and van Tongeren, 0. F. R. (eds), Data analysis in community and landscape ecology. Pudoc, Wageningen, The Netherlands, pp. 91- 173. ter Braak, C. J. F. 1988. CANOCO - a FORTRAN program for canonical community ordination by [partial] [detrended] [canonical] correspondence analysis, principal components analysis and redundance analysis (ver. 2.1). Report LWA-88-02, Agricult. Math. Group, Wageningen. ter Braak, C. J. F. 1990. Update notes: CANOCO ver. 3.1. Agricult. Math. Group, Wageningen. ter Braak, C. J. F. and Prentice, I. C. 1988. A theory of gradient analysis. - Adv. Ecol. Res. 18: 271-317. ter Braak, C . J. F. and Juggins, S. 1993. Weighted averaging partial least squares regression (WA-PLS): an improved method for reconstructing environmental variables from species assemblages. - Hydrobiologia 269/270: 485-502. Trexler, J. C. and Travis, J. 1993. Nontraditional regression analysis. - Ecology 74: 1629-1637. Tuhkanen, S. 1984. A circumboreal system of climatic-phytogeographical regions. - Acta Bot. Fenn. 127: 1-50. Utaaker, K. 1987. Landskapets innvirkning p i sommertemperaturen i indre Sogn. - Univ. Bergen, Meterol. Rep. Ser. 19844: 1-31. van Steenis, C. G. G. J. 1984. Floristic altitudinal zones in Malesia. - Bot. J. Linn. SOC.89: 289-292. Vazquez, G. J. A. and Givnish, T. J. 1998. Altitudinal gradients in tropical forest composition, structure, and diversity in the Sierra de Manantlan. - J. Ecol. 86: 999-1020. Ve, S. 1940. Skog og treslag i indre Sogn f r i Laerdal ti1 Fillefjell. - Medd. Vestlandets forstlige forssksstat. 23: 1-224. Westman, W. E. 1983. Xeric Mediterranean-type shrubland associations of Aka and Baja California and the community/continuum debate. - Vegetatio 52: 3- 19. Whittaker, R. H. 1956. Vegetation of the Great Smoky Mountains. - Ecol. Monogr. 26: 1-69. Whittaker, R. H. 1972. Evolution and measurement of species diversity. - Taxon 21: 213-251. Whittaker. R. H. 1975. Communities and ecosvstems. MacMillan. Whittaker. R. H. and Niering, W. A. 1975. Vegetation of the Santa Catalina mountain:,' Arizona. V. Biomass, production and diversity along the elevation gradient. - Ecology 56: 771-790. Williamson, M. 1981. Island populations. - Oxford Univ. Press. Wilson, M. V. and Mohler, C. L. 1983. Measuring compositional change along gradients. - Vegetatio 54: 129-141. Wilson, M. V. and Shmida, A. 1984. Measuring beta diversity with presence-absence data. - J. Ecol. 72: 1055-1064. Woodward, F. I. 1987. Climate and plant distribution. Cambridge Univ. Press. Woodward, F. I. 1988. Temperature and the distribution of plant species. - In: Long, S. P. and Woodward, F. I. (eds), Plant and temperature. Symp. SOC. Exp. Biol. 42, pp. 59-75. Woodward, F. I. 1993. The lowland-to-upland transition modelling plant responses to environmental change. Ecol. Appl. 3: 404-408. -

ECOGRAPHY 225 (1999)

ECOGRAPHY 27: 350 /360, 2004

Evaluating alternative data sets for ecological niche models of birds in the Andes Juan L. Parra, Catherine C. Graham and Juan F. Freile

Parra, J. L., Graham, C. C. and Freile, J. F. 2004. Evaluating alternative data sets for ecological niche models of birds in the Andes. / Ecography 27: 350 /360. Ecological niche modeling (ENM) is an effective tool for providing innovative insights to questions in evolution, ecology and conservation. As environmental datasets accumulate, modelers need to evaluate the relative merit of different types of data for ENM. We used three alternative environmental data sets: climatic data, remotesensing data (Normalized Difference Vegetation Index), and elevation data, to model the distribution of six bird species of the genus Grallaria in the Ecuadorian Andes. We assessed the performance of models created with each environmental data set and all possible combinations by comparing the geographic predictions of our models with detailed maps developed by expert ornithologists. Results varied depending on the specific measure of performance. Models including climate variables performed relatively well across most measures, whereas models using only NDVI performed poorly. Elevation based models were relatively good at predicting most sites of expected occurrence but showed a high over-prediction error. Combinations of data sets usually increased the performance of the models, but not significantly. Our results highlight the importance of including climatic variables in ENM and the simultaneous use of various data sets when possible. This strategy attenuates the effects of specific variables that decrease model performance. Remote-sensing data, such as NDVI, should be used with caution in topographically complex regions with heavy cloud-cover. Nonetheless, remote-sensing data have the potential to improve ENM. Finally, we suggest a priori designation of modeling purposes to define specific performance measures accordingly. J. L. Parra (juanp@uclink.berkeley.edu) and C. C. Graham, Museum of Vertebrate Zoology, 3101 VLSB, Univ. of California, Berkeley, CA 94720-3160 USA. / J. F. Freile, Fundacio´n Numashir para la Conservacio´n de Ecosistemas Amenazados, Casilla Postal 17-12-122, Quito, Ecuador.

Ecological niche models have been used to study issues in evolution (Peterson 2001, Hugall et al. 2002), ecology (Anderson et al. 2002), and conservation (Godown and Peterson 2000, Sa´nchez-Cordero and Martı´nez-Meyer 2000, Peterson and Robins 2003). These methods (e.g. BIOCLIM-Nix 1986, Busby 1991; GARP-Stockwell and Noble 1991; DOMAIN-Carpenter et al. 1993) combine geographic locations of a given species with spatial surfaces of environmental data to identify suitable parameters for a given species and then map this information to predict the species geographic distribution. Typically, interpolated climate data (e.g. Berry et al.

2002, Joseph and Stockwell 2002); or environmental data obtained through remote sensing (e.g. Fuentes et al. 2001, Oindo 2002, Zinner et al. 2002) are used to build models. To date, there has been no assessment of the relative performance of models created by these different datasets. In this article, we assess the utility of using climate, remotely-sensed Normalized Difference Vegetation Index (NDVI) data, elevation, and a combination of these variables to predict distributions for six bird species inhabiting the Ecuadorian Andes. Interpolated climate data are derived from direct measurements of climate at weather stations (New et

Accepted 29 December 2003 Copyright # ECOGRAPHY 2004 ISSN 0906-7590

350

ECOGRAPHY 27:3 (2004)

al. 1999). These data usually cover a considerable span of time (e.g. 30 yr) and are sensitive to error occurring during data processing and archiving. Coverage of weather stations is not uniform in space or time, and while there is information for some large areas with high resolution (e.g. Canada, Australia), information is still scarce for other areas (e.g. Amazonia). NDVI is a measure of the reflectance of earthâ&#x20AC;&#x2122;s surface vegetation (Lillesand and Kiefer 2000), and is representative of the leaf area index (Asrar et al. 1984) and vegetation productivity (Chong et al. 1993). While NDVI data has many positive attributes for modeling (i.e. global coverage at a 1.1 km2 resolution), it is susceptible to interference from moisture and/or background noise (Huete 1989). Elevation data is often used in models as a covariate of an environmental variable (Hutchinson 2000), or as a descriptor of a physical characteristic such as slope (Berry et al. 2002). Remote sensing programs such as the Shuttle Radar Topography Mission ( B/http://www.jpl.nasa.gov/srtm/ /) are further increasing the accuracy and resolution of digital elevation surfaces. Climate, NDVI, and elevation data sets are often related (Henricksen and Durkin 1986, Hutchinson 1998, Richard and Poccard 1998, Ichii et al. 2002, Oindo 2002); for example, in South America, â&#x20AC;&#x2DC;â&#x20AC;&#x2DC;greenerâ&#x20AC;&#x2122;â&#x20AC;&#x2122; areas have high rainfall and low temperatures (Ichii et al. 2002). Likewise, elevation shows a close relationship to temperature. Nevertheless, the nature of these relationships is variable in space and time (Ichii et al. 2002). Assuming a correlation between NDVI values and climate parameters (FjeldsaË&#x161; et al. 1997, Oindo 2002), we expect model performance by each data set to be similar in terms of the correctly predicted occurrences, in relation to 1) all expected occurrences (sensitivity) and 2) all predicted occurrences (positive predictive powerPPP). We also expect models based on each of these variables to have similar spatial predictions of species distributions. However, if data sets are not correlated, predictions should be different and possibly complementary, resulting in an overall improvement of model performance when datasets are combined. While climate and NDVI are similar in terms of the biotic and abiotic factors they measure, there are also several differences. These differences include temporal, spatial coverage, and sampling effort. NDVI data are available for short periods within the last 20 yr while climate data span up to 30 yr. In addition, at the time of this study, there was no global climate coverage available at a 1 km2 resolution, as with NDVI. NDVI data are collected such that a systematic reading is taken for each pixel. In a climate surface, pixel values are interpolated from weather station data. Hence, NDVI data should reflect spatial and temporal variation / for the space and time covered / at a higher resolution than climate data. Based on the consistent and recent coverage by remote ECOGRAPHY 27:3 (2004)

sensing, we predict that NDVI based models will be more representative of current distribution boundaries. Within the range map of a given species, NDVI based models should have less over-prediction (commission error), or in other words, higher specificity / higher probability of correctly predicting a cell as absent. Exclusion of old locality records which reflect available habitat in previous times that has recently been disturbed, should improve the performance of NDVI models. In order to assess the utility for modeling of climate, NDVI, and elevation data sets, we compared predicted versus expected distributions of six bird species in the Ecuadorian Andes. Predictive models were evaluated with expert knowledge based maps developed by Krabbe et al. (1998) for the highlands ( /1200 m a.s.l.) of Ecuador. The high levels of endemism and species richness characteristic of the Ecuadorian Andes, combined with a long history of anthropogenic use that has left many species threatened through habitat alteration, makes it an area of high conservation priority (Renjifo et al. 1997, Krabbe et al. 1998, Sierra et al. 2002). We hope this intent will encourage similar modeling studies in other countries and eventually at large biogeographic scales.

Methods: data sets Seven data sets were used to develop ecological niche models: 1) climate data set compiled and developed by C. Graham for Ecuador, 2) NDVI data set available from the â&#x20AC;&#x2DC;â&#x20AC;&#x2DC;Land Processes Distributed Active Archive Centerâ&#x20AC;&#x2122;â&#x20AC;&#x2122; (B/http://edcdaac.usgs.gov/1KM/1kmhomepage. html /), 3) elevation data set developed by Sierra (1999), 4) climate /elevation, 5) NDVI /elevation, 6) NDVI /climate, and 7) NDVI /climate /elevation. All grid layers were scaled to 1 km2 grid cell size. Climate surfaces for Ecuador were created using an interpolation technique based on thin plate smoothing splines (ANUSPLIN, Hutchinson 2000). Monthly climate surfaces were extrapolated for total precipitation and average temperature. Thin plate smoothing splines are a generalization of standard multi-variate linear regression in which the parametric model is replaced by a suitable smooth non-parametric function (Hutchinson 1995). We included three independent spline variables: longitude, latitude, and elevation to interpolate climate surfaces. Elevation was incorporated because temperature and rainfall are often highly correlated with elevation, and inclusion of elevation in the model reduced statistical error (Hutchinson 1991). We used a digital elevation model developed for Ecuador at a 0.16 km2 resolution by Sierra (1999). We included 264 rainfall points spread across Ecuador. The standard predictive errors ranged from 11.3% to 16.5%, which is 351

similar to errors in other topographically complex areas (Faith et al. 2001). The number of temperature data points was smaller than rainfall but the strong dependence of temperature on elevation enabled us to create accurate surfaces. We used data from 163 climate stations to create maximum and minimum monthly temperature surfaces. The standard predictive errors varied between 2.6% to 3.1% and 2.8% to 3.3% for maximum and minimum temperatures; respectively. Small errors such as these are typical for temperature surfaces (Hutchinson 1991, 2000). We used ANUCLIM ver. 5.1 package (Houlder et al. 2000), to create bioclimatic parameters that are biologically meaningful combinations of monthly climate variables (Nix 1986). These surfaces included annual mean temperature, annual total rainfall, monthly coefficients of variation for temperature and rainfall, and dry quarter-three consecutive months with the least total precipitation (grids available upon request from C. Graham). All climate grids were resampled to 1 km2 resolution. NDVI measurements used in this study were captured by the Advanced Very High Resolution Radiometer (AVHRR) carried by the National Oceanic and Atmospheric Administrationâ&#x20AC;&#x2122;s (NOAA) satellite. NDVI is calculated as the difference between the reflectance readings in the near infrared (channel 2) and visible (channel 1) light spectrum, and normalized over the sum of both readings: (Channel 2 /Channel 1)/(Channel 1 / Channel 2). Raw values range from /1.0 to 1.0; positive high values indicate highly vegetated areas (almost all visible spectrum is absorbed plus a high reflectance in the near infrared spectrum), and lower negative values non-vegetated features or cloud-covered areas (all visible spectrum is reflected; Holben 1986). Raw readings are processed according to international agreements ( B/http://edcdaac.usgs.gov/1KM/paper.html /) and final NDVI products are available online. Values are rescaled to a 0 /200 scale, where values B/100 are considered non-vegetated. We used NDVI data from two years, April 1992 to March 1993 and February 1995 to January 1996. Data are downloadable as maximum readings from 10-day composites (maximum NDVI

value obtained during a 10 day time lapse). To eliminate additional noise by cloud interference or soil reflectance, maximum monthly values out of the two years of data were used to calculate measures representative of productivity and seasonality. We calculated the following monthly layers based on monthly maximum values: overall maximum NDVI, overall minimum NDVI, mean annual NDVI, coefficient of variation, maximum and minimum quarters (three consecutive months with the maximum or minimum mean value respectively), annual seasonality (100 /[overall maximum /overall minimum]/overall maximum, Hurlbert and Haskell 2003), and dense and medium â&#x20AC;&#x2DC;â&#x20AC;&#x2DC;greennessâ&#x20AC;&#x2122;â&#x20AC;&#x2122; (number of months where each pixel had a value higher than 150 or between 109 and 150 respectively, Holben 1986). We tested for correlation among variables within and between data sets by plotting 1000 random points within Ecuador and extracting associated environmental values. Pearsonâ&#x20AC;&#x2122;s correlations were performed among all variables. We excluded variables that had a coefficient of correlation /0.7 (Green 1979) within any data set (i.e. among climate variables). When datasets were used in conjunction (e.g. climate and elevation) we included all variables even if they were correlated (Table 1). Four climatic variables (total annual rainfall, annual monthly mean temperature, coefficient of variation in monthly temperature and monthly rainfall), three NDVI derived variables (overall maximum NDVI, annual seasonality, and medium â&#x20AC;&#x2DC;â&#x20AC;&#x2DC;greennessâ&#x20AC;&#x2122;â&#x20AC;&#x2122;), and the elevation data were used as predictor variables.

Point localities We modeled the distribution of six species of antpittas (Grallaria quitensis, G. ruficapilla , G. rufula , G. nuchalis, G. squamigera , and G. hypoleuca , Passeriformes, Formicariidae) for which expert opinion distribution maps were published (Krabbe et al. 1998), and /40 different locality points were available for Ecuador (Freile 2000). Locality points were gathered from major collections and published records. Bird specimen data were gathered

Table 1. Pearsonâ&#x20AC;&#x2122;s correlation coefficients among variables used in modeling excercise. Variables within a dataset with a correlation coefficient /0.7 were not included in the analysis.

1 2 3 4 5 6 7 8 9

Environmental variables

Annual monthly mean temperature CV monthly temperature Total annual rainfall CV monthly rainfall Dry quarter Elevation Maximum NDVI Annual seasonality Medium greenness

1.00 0.38 0.43 0.24 0.27 0.99* 0.20 0.15 0.43

1.00 0.42 0.35 0.36 0.41 0.04 0.03 0.01

1.00 0.66 0.88* 0.45 0.12 0.16 0.15

1.00 0.87* 0.27 0.22 0.27 0.35

1.00 0.30 0.23 0.26 0.33

1.00 0.19 0.14 0.41

1.00 0.35 0.35

1.00 0.35

1.00

* Pearsonâ&#x20AC;&#x2122;s correlation coefficient /0.7.

352

ECOGRAPHY 27:3 (2004)

from the Academy of Natural Sciences of Philadelphia (ANSP), American Museum of Natural History (AMNH), British Museum of Natural History (BMNH), Carnegie Museum of Natural History (CMNH), Field Museum of Natural History (FMNH), Louisiana State Univ. Museum of Zoology (LSUMZ), Museo Ecuatoriano de Ciencias Naturales (MECN), Museo de Zoologıá, Pontificia Univ. Cato´lica del Ecuador (QCAZ), Museo de Zoologıá, Escuela Politećnica Nacional del Ecuador (EPN), U.S. National Museum of Natural History (NMNH), and the Western Foundation of Vertebrate Zoology (WFVZ). JFF verified identification and locality of published and museum records for birds. Points were checked for concordance between the reported and projected elevation correcting or removing points with an incongruence /500 m. No duplicate point localities were included in the models. We used a recent vegetation map developed by Sierra et al. (1999) to identify points within currently deforested areas and measure the change in NDVI based model performance when excluding those points from the analysis. We assumed deforested or heavily degraded areas were not suitable for any of the bird species included in the analyses (Freile 2000).

Model We used BIOCLIM as the modeling technique to simulate species distributions. BIOCLIM is a simple ‘‘envelope’’ or ‘‘profile matching’’ technique where a rule set, based on the variables included in the dataset is developed (Busby 1991). After a bioclimatic profile is obtained, the program determines all possible locations

on the grid with similar environmental characteristics to produce a map of the predicted potential distribution of a species in realized geographic space. The predicted distribution was based on the environmental envelope included between the 5th and 95th percentiles of the total environmental space; in other words, points that fell within the 5 tail percentiles of the total environmental space (10% of the total environmental space) were removed.

Model performance The performance of models should be measured relative to the particular research objective (Manel et al. 2001). In our case, the purpose was to closely predict geographic areas that were ‘‘suitable’’ or where we expected a species to occur, based on environmental correlates. In order to obtain a measure of the accuracy of the output, we compared the predictions from each dataset versus expert distribution maps limiting our analyses to the area encompassed within the proposed range map of each species (Fig. 1). We limited our comparison to cells within range maps (obtained from R. S. Ridgely as digital files) to create relatively objective and consistent confusion matrices (Appendix). Ecological niche models often over predict to areas where the organism is absent due to ecological (i.e. there is an extra dimension in the model, such as a competing species, that was not included) or historical reasons (i.e. a geographic barrier has impeded a species to colonize). This over prediction is difficult to evaluate, especially when areas of predicted occurrence are not close to current proposed ranges. By taking only areas within the range map, we restrict the

Fig. 1. Expert knowledge and predictive BIOCLIM distribution maps of Grallaria nuchalis in Ecuador. Only areas within the range map were included in the analysis. Black areas represent 5 /95 percentiles of BIOCLIM’s prediction.

ECOGRAPHY 27:3 (2004)

353

possible interpretations and assume that the absence of a species from any area within the range map is due to niche unsuitability. We used the predicted modelâ&#x20AC;&#x2122;s results within range maps to build a confusion matrix taking as true positives (a) the number of correctly predicted occurrence cells (within the range map), false positives (b) the number of grid cells predicted within the range but outside the expert map, false negatives (c) the number of grid cells incorrectly predicted as absent, and true negatives (d) as the cells that were correctly predicted absent. From the confusion matrix we derived: sensitivity (proportion of correctly predicted occurrence cells in relation to all expected cells, a/[a /c]), specificity (proportion of cells correctly predicted absent cells in relation to all absent cells, d/[b /d]), positive predictive power- PPP (proportion of correctly predicted cells in relation to all predicted occurrence cells, a/[a /b]), and Kappa (defined as the accuracy of the prediction in relation to that expected by chance alone, Farber and Kadmon 2003, [(a / d) / (((a / c)(a /b) / (b /d)(c /d))/N)]/[N / (((a / c)(a /b) /(b /d)(c /d))/N)], where N is the sum of all cases, Fielding and Bell 1997). To test for statistical differences in the average of each performance measure among models based on different data sets, we used analysis of variance (ANOVA) with a repeated measures design (Zar 1996). Data were transformed using arcsinĂ˘, which is the suggested transformation for proportions (Zar 1996). Non-parametric tests (Friedmanâ&#x20AC;&#x2122;s test) were performed when normality or homogeneity of variance was violated. If a significant ANOVA was obtained, we used Tukey post-hoc tests, or their non-parametric equivalent, to evaluate which datasets were different. We used paired T-tests to evaluate significant changes in Kappa values of NDVI models when excluding points in deforested areas. All data were checked for normality and homogeneity of variances. Statistical tests were performed using statgraphics plus 5.1 software, except for the Friedmanâ&#x20AC;&#x2122;s test, which were done by hand following Zar (1996).

Results Model performance Average performance of models based on different data sets was statistically different in terms of PPP (x2r(0.05,7,6) /31.142, pB/005). Models including elevation and climate performed better in terms of PPP, relative to models based only on elevation or NDVI (Fig. 2). Elevation based models had a higher sensitivity / predicted a higher proportion of the true area of occurrence / than any of the other models (F0.05(2)6,30 /55.337, pB/0.05, Tukey post hoc tests were all significant). Average specificity / the proportion of correctly predicted absence / was significantly 354

lower for both elevation and NDVI only based models (x2r(0.05,7,6) /33.5, pB/0.05), but not different from each other (q0.05,8,7 /0.378, p /0.05). In terms of Kappa values, the most accurate models were based on combined datasets of climate and elevation (x2r(0.05,7,6) / 21.571, pB/0.05). Nonetheless, differences among datasets were not significant except in relation to NDVI based models, which performed significantly worse (various q0.05,8,7 /4.17, pB/0.05) (Fig. 2). To better understand how Kappa values were related to the other performance measures assessed, we created pair-wise plots relating two specific performance measures for each species prediction with their respective Kappa value (Fig. 3). There was no consistent trend for all performance measures. When PPP was plotted against specificity or sensitivity, high PPP values often reflected high Kappa values, whereas there was no trend for specificity or sensitivity. When specificity was plotted against sensitivity, usually intermediate values of both measurements reflected high Kappa values, except for elevation based models which showed high Kappa values but low specificity. The overall accuracy (Kappa) of NDVI based models did not improve (maximum increase of Kappa /0.08, T0.05(1),5 / /0.083, p /0.05) when points in currently deforested areas (sensu Sierra et al. 1999) were left out. The only two species for which the Kappa values slightly increased (G. quitensis and G. ruficapilla ), were the ones where most points were removed from the model (38 and 48% of total points available, respectively).

Predicted areas Area predicted by elevation and NDVI based models within range maps was always larger than area predicted by any other model (Table 2). This trend was consistent when total predicted area was analyzed. On average, elevation based models predicted two times more area than climate based models (range 1.47 /3.01) when results only within range maps were analyzed, and 2.8 times more when total area was considered (range 1.41 / 6.68). NDVI based models showed a similar trend, predicting even total larger areas in some instances (Table 2).

Discussion One of the main goals of ecological niche modeling is to accurately predict the distribution of a species at the present time. Climate data is an obvious choice but of limited availability and quality in some areas. Remotely sensed data, such as NDVI are being increasingly used for this purpose (Fuentes et al. 2001, Berry et al. 2002, Zinner et al. 2002). Global coverage at ca 1-km ECOGRAPHY 27:3 (2004)

Fig. 2. Box plots summarizing results of measures of performance for each dataset used. Median values (line across box), range excluding outliers (error bars), interquartile range containing 50% of values (box), and outliers (circles) from results of six species for each performance measure evaluated across data sets (E /elevation, C /climate, N /NDVI). Untransformed values were used.

resolution and public accessibility are two characteristics that make the NDVI data set attractive. However, in the present study we found that models based solely on NDVI generally performed poorly in comparison to those created using climate and elevation data. Models based on our seven data sets varied in terms of the performance measures evaluated (PPP, specificity and sensitivity), specific agreement (Kappa), and area that they predicted. No obvious congruence or tendency was found across different measures of performance and we could not easily identify a set of environmental layers that consistently performed well across all performance measures. This result indicates that choice of environmental variables can be complex and the measure of performance should be specific to the type of errors, which should be minimized for a given application (Loiselle et al. 2003, Fielding and Bell 1997). Given that relative model performance was not consistent across evaluation criteria, it may be necessary to assign prior cost values to specific types of errors based on the particular objectives of the modeling procedure (Fielding and Bell 1997). For example, conECOGRAPHY 27:3 (2004)

servationists often prioritize models with high sensitivity and PPP because failure to predict actual areas of occurrence may be more costly than falsely predicting them (Fielding 2002, Loiselle et al. 2003). In essence, by maximizing these two measures, the probability of omission error / not including an area where a species occurs in the prediction / is decreased, and the accuracy of occurrence predictions is increased, as long as commission error (false occurrence predictions) is maintained. In our case, PPP did not vary significantly across most models; however, models built only with elevation data had higher sensitivity than models based on any other dataset. Interestingly, elevation based models achieve a high sensitivity (no areas of occurrence are omitted) by sacrificing specificity through a high overprediction. Hence, final decisions on which models are the most reliable, should take all types of error into account and how each error could compromise the specific research objective (Fig. 3). Application of ecological niche models in ecology and evolution has raised the importance of measuring and interpreting commission errors (Anderson et al. 355

Fig. 3. Bubble charts showing relationships between specific measures of performance (PPP, sensitivity and specificity) represented by the axes, and Kappa values, represented by the surface area of the circles.

2003) / errors of over-prediction. In principle, if environmental variables included in the model are appropriate, over-prediction can indicate that other other factors may be determining geographic distributions such as historical connectivity or competitive exclusion by closely related or ecologically similar taxa. Specificity, which measures the proportion of correctly predicted absent cells to all expected absent cells (Fielding 2002), was highest in climate-based models. Models able to differentiate suitable and non-suitable areas within the range map should be the most reliable in terms of over-predictions. Range maps are large-scale expectations of where a species can be found and areas not inhabited within range maps represent inappropriate habitat (Ridgely and Tudor 1994) rather than other factors. Thus, we argue that climate based models did a fairly good job determining unsuitable areas within the range map and also achieve a relatively high PPP. Elevation and NDVI based models had significantly lower specificities than any of the other models, again 356

highlighting the fact that these models attain a high sensitivity with large over predictions. When elevation and NDVI datasets were combined, their specificity increased significantly, suggesting datasets were predicting distinct areas of expected absence, and thus when joined, were complementary. Elevation data generally were able to delineate the range of a given species but they did not detect smaller-scale changes within boundaries of the range maps. However, NDVI is more likely to capture small isolated patches where human intervention or any type of disturbance has caused the vegetation to change. Hence, elevation and NDVI might be detecting different areas based on the particular nature of each dataset. Finally, performance measures, such as Kappa, make use of all the information in a confusion matrix and give us a picture of the specific agreement of the model over chance (Fielding and Bell 1997). NDVI based models had the lowest Kappa value. This is likely a result of its particularly low specificity and PPP. In each of the other datasets, the average Kappa values are similar, reflecting in most cases intermediate values of sensitivity and specificity, except in the elevation-only based models, which had both significantly high sensitivity and low specificity (Fig. 3). In general, average Kappa values were relatively low (B/0.4, Landis and Koch 1977) suggesting that model predictions do a slightly better job than chance. Various aspects of the model such as the quality of raw data, the predictor variables and modeling procedure employed, could be contributing to a general low performance. The quality of the environmental data sets can be evaluated through their estimated error. Temperature interpolations had a low error estimate in relation to rainfall, which had a relatively high error because of the complex topography of the Andes. Nonetheless, species ecological niche models have been successfully created with similar error rates in other areas (Faith et al. 2001). Unfortunately, we have no error estimates for NDVI or elevation surfaces; estimation of these errors is out of our reach. One of the potential sources of error in remotely-sensed data such as NDVI is interference with clouds (Asner 2001). We controlled for cloud interference in NDVI data by including only the maximum monthly values from both years. Previous studies using NDVI data have been performed at high elevations in the Andes and suggest that a signal can be captured out of the data (Fuentes et al. 2001). Distribution models may perform poorly if we chose environmental predictor variables that are not biologically meaningful for the study subject. NDVI and climate data are two of the most commonly used predictors of species distributions at a regional scale because of their presumed importance as limiting factors for species distributions (Guisan and Zimmermann 2000, Peterson et al. 2002, Hurlbert and Haskell 2003). ECOGRAPHY 27:3 (2004)

37020 20689 29758 19075 15063 10450 9608 44854 25492 40110 23028 19166 13186 11975 38059 22485 36109 18112 19648 16925 13413 49010 30064 56353 24034 26057 22940 18219 44816 24276 43608 21708 25631 18278 16433 64000 35374 55998 32106 35585 24517 22407 30001 20322 31298 19172 22339 17151 16128 41120 29029 56436 25819 29097 22991 20865 11774 5316 11616 4529 7716 4078 3492 51250 15199 61610 12966 33907 11384 9992 12848 4268 7832 4199 5937 2432 2387 34064 5094 39494 5014 14331 2921 2869 Elevation Climate NDVI Elev /Cli Elev /NDVI Cli /NDVI Elev /Cli /NDVI

Cut Total Total Cut Total

Cut

Total

Cut

Total

Cut

Total

Cut

G. squamigera G. rufula G. ruficapilla G. quitensis G. nuchalis G. hypoleuca

Table 2. Predicted areas of occurrence based on different data sets. Predicted areas within each speciesâ&#x20AC;&#x2122; range map (under column labeled â&#x20AC;&#x2DC;â&#x20AC;&#x2DC;cutâ&#x20AC;&#x2122;â&#x20AC;&#x2122;) and total predicted areas within Ecuador (under column labeled â&#x20AC;&#x2DC;â&#x20AC;&#x2DC;totalâ&#x20AC;&#x2122;â&#x20AC;&#x2122;). Units are squared meters.

ECOGRAPHY 27:3 (2004)

Antpittas are a distinctive group of new world forest dwelling birds that are shy and are not believed to fly long distances (Ridgely and Tudor 1994). These traits combined with historical fragmentation cycles of cloud forests due to pleistocene glaciations can lead to high, localized adaptation (Holt 2003). Grallaria distributions are often restricted to narrow altitudinal ranges along the Andes (Ridgely and Tudor 1994) delineated by narrow environmental variables. These variables should be reflected in the environmental data sets we employed. In the case of NDVI-only based models, we expected performance to improve when we excluded points in currently deforested areas. Nonetheless, there were no apparent changes in terms of Kappa values. The fact that we had a small sample size (40 /140 point localities), makes the exclusion of just a few points out of the model a compromise that could lead to decreased model performance. A final factor, which could contribute to a general low performance is the modeling method employed. Although BIOCLIM is a simple straightforward method / one of its main strengths / it has some flaws (i.e. giving an equal weight to all variables, and sensitivity to outliers, Farber and Kadmon 2003). However, there is no consensus in the literature as to what method is â&#x20AC;&#x2DC;â&#x20AC;&#x2DC;bestâ&#x20AC;&#x2122;â&#x20AC;&#x2122; (Guisan and Zimmermann 2000) and initial testing between BIOCLIM, GARP, and DOMAIN methods indicated that results from each modeling type were similar (results not shown here). Our objective in this research was not to test among modeling methods but to test among environmental datasets. Modeling species geographic distributions is a useful exercise to better understand potential limiting factors or correlates of speciesâ&#x20AC;&#x2122; distributions (Oâ&#x20AC;&#x2122;Connor 2002). Nonetheless, modeling procedures can be misleading if the performance of the models and the types of environmental variables used are not evaluated. The results presented here show that performance has many faces, each representing a cost depending on the purpose of the model (Fielding and Bell 1997, Guisan and Zimmermann 2000). Lack of thorough spatial sampling and the need for effective conservation networks are the usual scenario for many tropical humid forest dwelling birds. Niche modeling can be a useful aid to their conservation (Peterson et al. 2002). As detailed natural history information on these species accumulates, we will be able to relate landscape variables (such as mean patch size or proximity of patches) to species-specific information on tolerance of habitat alteration or minimum patch size requirements to refine and improve our models. Acknowledgements / We would like to thank N. Krabbe and R. Ridgely for his collaboration with the maps. Robert Hijmans and Wayne Heiser provided technical advice. Craig Moritz provided funding for JLP work in the MVZ. Robert Hijmans and Craig Moritz provided useful advice on earlier drafts. JFF would like to thank Programa de Becas de InvestigacioÂń para la ConservacioÂń, FundacioÂń EcoCiencia, and William Belton

357

Donations Program, American Bird Conservancy, for funding research on the distribution and conservation of Grallaria species. The Museum of Vertebrate Zoology (MVZ), NCEAS working group and an NSF grant provided funding for all modeling analyses. All computer software used was provided through the MVZ.

References Anderson, R. P., Gomez-Laverde, M. and Peterson, A. T. 2002. Geographical distributions of spiny pocket mice in South America: insights from predictive models. / Glob. Ecol. Biogeogr. 11: 131 /141. Anderson, R. P., Lew, D. and Peterson, A. T. 2003. Evaluating predictive models of species distributions: criteria for selecting optimal models. / Ecol. Modell. 162: 211 /232. Asrar, G. et al. 1984. Estimating absorbed photosynthetic radiation and leaf area index from spectral reflectance in wheat. / Agr. J. 76: 300 /306. Asner, G. P. 2001. Cloud cover observation of the Brazilian Amazon. / Int. J. Remote Sens. 22: 3855 /3862. Berry, P. M. et al. 2002. Modelling potential impacts of climate change on the bioclimatic envelope of species in Britain and Ireland. / Glob. Ecol. Biogeogr. 11: 453 /462. Busby, J. R. 1991. BIOCLIM / A bioclimate analysis and prediction system. / In: Margules, C. R. and Austin, M. P. (eds), Nature conservation: cost effective biological surveys and data analysis. CSIRO, pp. 64 /68. Carpenter, G., Gillison, A. N. and Winter, J. 1993. DOMAIN: a flexible modeling procedure for mapping potential distributions of plants and animals. / Biodiv. Conserv. 2: 667 /680. Chong, D. L. S., Mougin, E. and Gastellu-Etchegorry, J. P. 1993. Relating the global vegetation index to net primary productivity and actual evapotranspiration over Africa. / Int. J. Remote Sens. 14: 1517 /1546. Faith, D. P. et al. 2001. The Biorap biodiversity assessment and planning study for Papua New Guinea. / Pacific Conserv. Biol. 4: 279 /288. Farber, O. and Kadmon, R. 2003. Assessment of alternative approaches for bioclimatic modeling with special emphasis on the Mahalanobis distance. / Ecol. Modell. 160: 115 / 130. Fielding, A. H. 2002. What are the appropiate characteristics of an accuracy measure? / In: Scott, J. M. et al. (eds), Predicting species occurrences: issues of accuracy and scale. Island Press, pp. 271 /279. Fielding, A. H. and Bell, J. F. 1997. A review for the assessment of prediction errors in conservation presence/absence models. / Environ. Conserv. 24: 38 /49. Fjeldsa˚, J. et al. 1997. Are biodiversity ‘hotspots’ correlated with current ecoclimatic stability? A pilot study using the NOAAAVHRR remote sensing data. / Biodiv. Conserv. 6: 401 / 422. Freile, J. F. 2000. Patrones de distribucioń y sus implicaciones en la conservacioń de los geńeros Grallaria y Grallaricula (Aves: Formicariidae) en Ecuador. PhD thesis, Dept of Biology, Pontificia Univ. Cato´lica del Ecuador, Quito, Ecuador. Fuentes, M. V., Malone, J. B. and Mas-coma, S. 2001. Validation of a mapping and prediction model for human fasciolosis transmission in Andean very high altitude endemic areas using remote sensing data. / Acta Tropica 79: 87 /95. Godown, M. E. and Peterson, A. T. 2000. Preliminary distributional analysis of US endangered bird species. / Biodiv. Conserv. 9: 1313 /1322. Green, R. H. 1979. Sampling design and statistical methods for environmental biologists. / Wiley.

358

Guisan, A. and Zimmermann, N. E. 2000. Predictive habitat distribution models in ecology. / Ecol. Modell. 135: 147 / 186. Henricksen, B. L. and Durkin, J. W. 1986. Growing period and drought early warming in Africa using satellite data. / Int. J. Remote Sens. 7: 1583 /1608. Holben, B. N. 1986. Characteristics of maximum-value composite images from temporal AVHRR data. / Int. J. Remote Sens. 7: 1417 /1434. Holt, R. T. 2003. On the evolutionary ecology of species’ ranges. / Evol. Ecol. Res. 5: 159 /178. Houlder, D. et al. 2000. ANUCLIM user’s guide. Centre for Resource and Environmental Studies (CRES). / Australian National Univ., Canberra, Australia. Huete, A. R. 1989. Soil influences in remotely sensed vegetationcanopy spectra. / In: Asrar, G. (ed.), Theory and applications of optical remote sensing. Wiley, pp. 107 /141. Hugall, A. et al. 2002. Reconciling paleodistribution models and comparative phylogeography in the Wet Tropics rainforest land snail Gnarosophia bellendenkerensis (Brazier 1875). / Proc. Natl. Acad. Sci. USA 99: 6112 /6117. Hurlbert, A. H. and Haskell, J. P. 2003. The effect of energy and seasonality on avian species richness and community composition. / Am. Nat. 161: 83 /97. Hutchinson, M. F. 1991. Climatic analyses in data sparse regions. / In: Muchow, R. C. and Bellamy, J. A. (eds), Climatic risk in crop production. CAB, pp. 55 /71. Hutchinson, M. F. 1995. Interpolating mean rainfall using thin plate smoothing splines. / Int. J. GIS 9: 305 /403. Hutchinson, M. F. 1998. Interpolation of rainfall data with thin plate smoothing splines: II analysis of topographic dependence. / J. Geogr. Inf. Dec. Anal. 2: 168 /185. Hutchinson, M. F. 2000. Anusplin, ver. 4.1. User guide. / Centre for Resource and Environmental Studies, The Australian National Univ., Canberra, Australia. Ichii, K., Kawabata, A. and Yamaguchi, Y. 2002. Global correlation analysis for NDVI and climatic variables and NDVI trends: 1982 /1990. / Int. J. Remote Sens. 23: 3873 / 3878. Joseph, L. and Stockwell, D. 2002. Climatic modeling of the distribution of some Pyrrhura parakeets of Northwestern South America with notes on their systematics and special reference to Pyrrhura caeruleiceps Todd, 1947. / Ornitol. Neotrop. 13: 1 /8. Krabbe, N. et al. 1998. An Atlas of avian diversity in the Ecuadorian Andes. / Centre for Research on Cultural and Biological Diversity of Andean Rainforests (DIVA), DIVA Tech. Rep. no. 4. Quito, Ecuador. Landis, J. R. and Koch, G. C. 1977. The measurement of observer agreement for categorical data. / Biometrics 33: 159 /174. Lillesand, T. M. and Kiefer, R. W. 2000. Remote sensing and image interpretation, 4th edn. / Wiley. Loiselle, B. A. et al. 2003. Avoiding pitfalls of using species distribution models in conservation planning. / Conserv. Biol. 17: 1591 /1600. Manel, S., Williams, H. C. and Ormerod, S. J. 2001. Evaluating presence-absence models in ecology: the need to account for prevalence. / J. Appl. Ecol. 38: 921 /931. New, M., Hulme, M. and Jones, P. 1999. Representing twentieth-century space-time climate variability. Part I: development of a 961-90 mean monthly terrestrial climatology. / J. Climate 12: 829 /856. Nix, H. A. 1986. A biogeographic analysis of Australian elapid snakes. / In: Anon. (ed.), Atlas of Australian elapid snakes. Bureau Flora Fauna, Canberra, Australia, pp. 4 /15. O’Connor, R. J. 2002. The conceptual basis of species distribution modeling: time for a paradigm shift? / In: Scott, J. M. et al. (eds), Predicting species occurences: issues of accuracy and scale. Island Press, pp. 25 /33. Oindo, B. O. 2002. Predicting mammal species richness and abundance using multitemporal NDVI. / Photo. Engin. Rem. Sens. 68: 623 /629. ECOGRAPHY 27:3 (2004)

Peterson, A. T. 2001. Predicting species geographic distributions based on ecological niche modeling. / Condor 103: 599 / 605. Peterson, A. T., Stockwell, D. R. B. and Kluza, D. A. 2002. Distributional prediction based on ecological niche modeling of primary occurrence data. / In: Scott, J. M. et al. (eds), Predicting species occurences: issues of accuracy and scale. Island Press, pp. 617 /623. Peterson, A. T. and Robins, C. R. 2003. Using ecological niche modeling to predict barred owl invasions with implications for spotted owl conservation. / Conserv. Biol. 17: 1161 / 1165. Renjifo, L. M. et al. 1997. Patterns of species composition and endemism in the northern Neotropics: a case for conservation of montane avifaunas. / Ornit. Monographs 48: 577 / 594. Richard, Y. and Poccard, I. 1998. A statistical study of NDVI sensitivity to seasonal and interannual variations in Southern Africa. / Int. J. Remote Sens. 19: 2907 /2920.

ECOGRAPHY 27:3 (2004)

Ridgely, R. S. and Tudor, G. 1994. The birds of South America. vol. I: the oscine passerines. / Univ. of Texas Press. Sańchez-Cordero, V. and Martıńez-Meyer, E. 2000. Museum specimen data predict crop damage by tropical rodents. / Proc. Natl. Acad. USA 97: 7074 /7077. Stockwell, D. R. B. and Noble, I. R. 1991. Induction of sets of rules from animal distribution data: a robust and informative method of data analysis. / Math. Comp. Sim. 32: 249 / 254. Sierra, R. 1999. Propuesta preliminar de un sistema de clasificacioń de vegetacioń para el Ecuador continental. / Proyecto Inefan/Gef-Birf y EcoCiencia. Quito, Ecuador. Sierra, R., Campos, F. and Chamberlin, J. 2002. Assessing biodiversity conservation priorities: ecosystem risk and representativeness in continental Ecuador. / Landsc. Urban Plann. 59: 95 /110. Zar, J. H. 1996. Biostatistical analysis, 3rd ed. / Prentice Hall. Zinner, D., Pelaéz, F. and Tokler, F. 2002. Distribution and habitat of grivet monkeys (Cercopithecus aethiops aethiops ) in eastern and central Eritrea. / Afr. J. Ecol. 40: 151 /158.

359

360

ECOGRAPHY 27:3 (2004)

Predicted

Elev /Cli /NDV (7)

Cli /NDV (6)

Elev /NDV (5)

Elev /Clim (4)

NDVI (3)

Climate (2)

Elevation (1)

Cut

Data sets

Species

/ / / / / / / / / / / / / / / / / / / / / / / / / / / /

9408 751 7250 631 3616 6543 2987 4894 5302 4857 3930 3951 3600 6559 2971 4910 4883 5276 3653 4228 2043 8116 1673 6208 2031 8128 1661 6220

/ 24656 47167 5598 3004 1478 70345 1281 7321 34192 37631 3902 4700 1414 70409 1228 7374 9448 62375 2284 6318 878 70945 759 7843 838 70985 726 7876

G. hypoleuca

11293 318 5783 224 6637 4974 3053 2954 8834 2777 4646 1361 6498 5113 2976 3031 8371 3240 4325 1682 4983 6628 2360 3647 4880 6731 2296 3711

G. nuchalis

39957 30414 5991 4090 8562 61809 2263 7818 52776 17595 6970 3111 6468 63903 1553 8528 25536 44835 3391 6690 6401 63970 1718 8363 5112 65259 1196 8885

Observed

18545 738 14662 620 14553 4730 11676 3606 16125 3158 13147 2135 14257 5026 11397 3885 14722 4561 11950 3332 12430 6853 10243 5039 12151 7132 9978 5304

G. quitensis

22575 40124 15339 8934 14476 48223 8646 15627 40311 22388 18151 6122 11562 51137 7775 16498 14375 48324 10389 13884 10561 52138 6908 17365 8714 53985 6150 18123

/ 24059 124 18875 46 17268 6915 12816 6105 16742 7441 13696 5225 16936 7247 12556 6365 15869 8314 12969 5952 12355 11828 9735 9186 12164 12019 9554 9367

/ 39941 17858 25941 14550 18106 39693 11460 29031 39256 18543 29912 10579 15170 42629 9152 31339 19716 38083 12662 27829 12162 45637 8543 31948 10243 47556 6879 33612

G. ruficapilla

15184 967 12949 656 10102 6049 8270 5335 11560 4591 9657 3948 8938 7213 7412 6193 9170 6981 7728 5877 7369 8782 6054 7551 6559 9592 5434 8171

G. rufula

33826 32005 25110 13901 19962 45869 14215 24796 44793 21038 26452 12559 15096 50735 10700 28311 16887 48944 11920 27091 15571 50260 10871 28140 11660 54171 7979 31032

5015 225 4018 156 3733 1507 3025 1149 3167 2073 2424 1750 3619 1621 2926 1248 2839 2401 2215 1959 2278 2962 1810 2364 2207 3033 1749 2425

39839 36903 33002 23537 21759 54983 17664 38875 36943 39799 27334 29205 19409 57333 16149 40390 16327 60415 12848 43691 10908 65834 8640 47899 9768 66974 7859 48680

G. squamigera

Appendix. Confusion matrices of models based on each dataset used. Numbers correspond to counts of square kilometer cells. Confusion matrices for predictions within each speciesâ&#x20AC;&#x2122; range map are coded under 1 in the column â&#x20AC;&#x2DC;â&#x20AC;&#x2DC;cutâ&#x20AC;&#x2122;â&#x20AC;&#x2122;; matrices for total predictions within Ecuador are coded as 0. Positive â&#x20AC;&#x2DC;â&#x20AC;&#x2DC; /â&#x20AC;&#x2122;â&#x20AC;&#x2122; sign refers to presence and negative â&#x20AC;&#x2DC;â&#x20AC;&#x2DC; /â&#x20AC;&#x2122;â&#x20AC;&#x2122; to absence.

FORUM i!> a lighter channel of communication between readers and contnbutors, it aims to stimulate discussion and debate, particularly by presenting new ideas and by suggesting alternative interpretations to the more formal research papers published in ECOGRAPHY and elsewhere A lighter prose is encouraged and no summary is required Contributions should be concise and to the point, with a relatively short bibliography Formal research papers, however short, will not be considered

The elevational gradient of species richness: a uniform pattern? Carsten Rahbek, Centre for Tropical Biodiversity, Zoological Museum, University of Copenhagen, Universitetsparken 15, DK-2100 Copenhagen 0, Denmark.

The decline m species nchtiess with increasing elevation IS widely accepted as a general pattern (Table 1) In as much as the elevational gradient is often claimed to mirror the latitudinal gradient, spectes richness is assumed to decrease monotonically (l e because of reduced temperature and consequent decrease in producUvity) Perhaps because such a relation is intuitive, biologists have readily generalized the results of a few early studies of tropical birds as supporting a general biogeographic

Table 1 Typical examples of statements atwut the relationship between species nchness and elevation from books and papers in major journals

pattern This has resulted in "citation inbreeding" Here, I outline the supporting evidence for the generalization and discuss the influence of samplmg regime and the often Ignored influence of area I then present a quantitative review of the data already present, although often ignored, in the literature Altogether 97 papers (with 163 examples) have been reviewed. The analysis of these empincal data support the view that species nchness declines with elevation, but not the view that this decline IS necessanly monotonic Some possible reasons for variation in the exact shape of the relationship between species nchness and elevation for different taxa and zoogeographic areas are commented, but our understanding of the relation between elevation and species nchness suU appears to be immature

"For all of these reasons, we expect the number of species to decrease with altitude and, m fact, it does " (MacArthur 1972, p 107) "In terrestrial habitats, vanation in species diversity along gradients of elevauon and available soil moisture are [sic] almost as general and striking as latitudinal vanation " (Brown and Gibson 1983, p 502) "In terrestnal environments, a decrease in species with altitude IS a phenomenon almost as widespread as a decrease with latitude " (Begon et al 1990, p 805) "Just as change of physical conditions with altitude resembles in many respects the vanauon with latitude, so the decreasing diversity of most organisms with increasing elevation mirrors in most respects the latitudinal gradient of species nchness" (Brown 1988, p 62) "biologists have long recogmzed that elevational and latitudinal species-nchness gradients mirror each other " (Stevens 1992, p 899) "In terrestnal ecosystems, diversity generally decreases with increasing altitude there appear to be no substantiating data for [the] 'mid-altitudinal bulge' as a general phenomenon" (World Conservation Momtonng Centre 1992, pp 43,45) "Decrease in the number of species with decreasing temperatures at higher altitudes is as conspicuous as the decrease with latitude (e g Brown and Gibson 1983), although exceptions occur" (Rohde 1992, p 522)

200

The empirical basis for the "general pattern" The generalization (Table 1) grew mamly from two studies dealing with tropical birds one from the Peruvian Andes (Terborgh 1977), and the other from New Guinea (Kikkawa and Williams 1971) The textbook example from New Guinea was published as a short note based on compilation of the published distnbutional data (Kikkawa and Williams 1971) It was conducted at a Ume when the knowledge of the elevational distnbution of New Guinean btrds was still somewhat rudimentary Another more detailed study on birds of New Guinea has also been cited as proof of the general pattern (Diamond 1972, cited in MacArthur 1972) However, Diamond's New Guinea data actually show a small peak tn species nchness at 1100 m with a marked decline m species nchness only above this level The second textbook example was based on a carefully conducted field survey and cntical data analysis (Terborgh 1977). Based on mist-netting and opportunisuc ECXX3RAPHY 18 2 (1995)

A 200

B 70

1000

2000

3000

4000

1000 2000 3000 ELEVATION (M)

ELEVATION (M)

4000

Fig I Species nchness of syntopic birds versus elevation on an Amazonian slope of the Andes in Peru Figure 1A is based on data not standardized for area and sampling effort, whereas Fig IB is based on standardized samples of 300 mist-netted birds (data from Terborgh 1977) I have fitted the lines by distance-weighted least-squares smoothing

field observattons at camps situated along ati elevational gradient on the humid east slope of the Peruvian Andes, Terborgh showed that species nchness declined monotonically with elevation if the number of species is simply plotted against elevation (Fig lA) However, when Terborgh tned to eliminate the effect of sampling time by standardizmg his mist-netting data, a different pattern emerged (Fig IB) Terborgh explained the emerging "hump-shaped" curve, which had a peak in species nchness around 1400 m, as the result of a local "hot-spot" in resources (Terborgh 1977) Although this unexpected pattern is addressed through most of his discussion, only the first non-standardized graph is usually cited in the literature (e.g Brown 1988) Though the Ube of under-

storey mist-netting data limited the scope of this study, it serves to demonstrate the strong effect that sampling effort can exert - especially in species-nch tropical forest where most species occur at low densities

The importance of area The effect of area on the relationship between spwcies nchness and elevation has rarely been considered, although we would expect area to have a significant impact on the form of the elevational pattern, as the relationship between area and species nchness seems as universal as

A 2500 2000

1500

o ffl

1000

500

ELEVATION (M)

Fig 2 Species nchness of South Amencan tropical landbirds versus elevation Figure 2A is based on data not standardized for area, whereas Fig 2B is based on data standardized for area set to 50000 km^ using equations for species-area curves (log S/log A transformation) of each elevauonal zone (based on data from Rahbek unpubl) Area is set to 50000 km' because it is within the range of the ongmal data upon which the equaUons are based, and is a size within the geographical regional scale (e g Wiens 1981, Rosenzweig and Abramsky 1993) ECOGRAPHY 18 2 (1995)

201

Table 2 Number of data sets (n = 163) found in the literature with data on the variation of species nchness with elevation summanzed by biomes (NT = non-tropical biomes, T = tropical biomes) and scale (R = regional, L = local), listed for studies conducted on mainland and island, respectively, and subdivided for whether the researcher(s) have attempted to standardize for the effect of area and sampling regime and/or effort, only one of these factors, or none ("non-standardized") Vertebrates

Invertebrates

NT(R/L) T(R/L)

Plants

Totals

NT(R/L)

T(R/L)

NT (R/L)

T(R/L)

NT (R/L)

T (R/L)

Mainland Area and sampling Area Sampling Non-standardized

0/2 0/0 4/1 9/11

0/0 0/0 3/4 4/6

0/6 0/0 0/6 3/0

0/1 1/0 3/1 17/4

0/10 0/0 0/1 1/0

0/6 4/3 0/0 8/3

0/18 0/0 4/8 13/11

0/7 5/3 6/5 29/13

Island Area and sampling Area Sampling Non-standardized

1/0 0/0 0/0 0/5

3/2 1/0 0/5 2/5

0/0 0/0 0/0 0/0

0/0 0/3 0/1 5/2

0/4 0/0 0/0 0/0

0/0 0/0 0/0 2/0

1/4 0/0 0/0 0/5

3/2 1/3 0/6 9/7

14/19

13/22

3/12

26/12

1/15

14/12

18/46

53/46

Total

the latitudinal gradient To understand the relationship between area and species nchness along elevational gradients, especially on a regional scale, the effect of area must be considered since areas of equal-sized elevational belts may vary with elevation Thus, areas often decrease with elevation because of generally steeper terrain toward the highest peaks When landbird data from tropical South Amenca, compiled at a regional scale using countries as units, are standardized for area, the relationship between species nchness and elevation gives a humpshaped curve (Fig 2) However, area alone is unlikely to explain any global pattern of species nchness, as close couplings can be expected to exist between biological diversification and habitat complexity (see also Rosenzweig 1992)

A quantitative review of the literature The repeated citation of the same few studies provides a false picture of the amount and diversity of data published on the issue I have been able to find the surpnsmgly high number of 97 papers, with 163 examples that give data on the vanation in number of species with elevation. It is highly probable that additional data exist, as many data sets are published in httle known journals, or in the "gray" literature. Table 2 summanzes some charactenstics of each data set (taxonomic group, region, scale and data treatment) The mfluences of sampling regime/effort and the effect of area are among the most influential biasing factors in most field studies of species-nchness patterns, and, unfortunately, equally difficult to eliminate successfully I have thus only judged whether an attempt was made to deal with these two factors, either in the design of the survey or afterward, during the data analyses. Remaricably, many of the pajjers reviewed do not give any details 202

on this subject These data sets are classified as "Nonstandardized" together with those with no attempt to standardize data (Table 2) Unlike the traditional histoncal trend within most fields of biological research, most data sets are from the tropics (99 out of 163) The focus on the tropics is presumably related to the circumstance that tropical elevational gradients compnse a wider range of climatic vanation than temjserate elevational gradients Out of 163 data sets, 68 are on invertebrates, 53 on vertebrates (including 36 on birds), and 42 on plants The majonty of the data sets comes from mainland biota (122), whereas 41 data sets are fi-om islands As shown, standardizing for sampling effort or effect of area can have a significant influence on the emerging shape of the relationship between sjjecies richness and elevation In 87 of the 163 data sets, the data have not been standardized for area or sampling effort (corresponding figures for tropical and non-tropical biomes are 59% and 45%) Only in 35 (21 %) cases has a standardization been attempted for area as well as sampling effort (figures for tropical and non-tropical biomes are 12% and 36%, respectively) Considenng the high mobility of birds compared to other groups, the reliance by most reviewers on primanly bird examples to illustrate a universal relationship seems ill-founded, especially when the bulk of data in the hterature actually denves from invertebrates and plants Table 2 also provides an overview for which combmations of, for example, taxonomic groups, scale, and region we lack studies - especially those that consider the effect of area and samphng regime on data.

Methodological problems and patterns In many studies, a stated decline in species nchness with ECOGRAPHY 18 2 (1995)

Table 3 The relationship between species nchness and elevation summarized by type of pattern Only the 90 data sets (of 163) that provide data points spanning from below 500 m to above 1500 m are included (see text) The classification of each pattern is based on a visual examination of bivanate plots NT= non-tropical biomes and T = tropical biomes Scale

Monotomcally decreasing

Horizontal, then decreasing NT

Hump-shaped

Regional Invertebrates Vertebrates Plants

10 2 1 1

elevation was restncted to only a part of the elevational gradient In other cases, mid-elevational data were lacking Others used correlation tests on data sets that include few stations from low- and mid-elevation, but many from higher elevations, thereby biasing their findings toward a strong negative correlation In such instances, the data are inapplicable to support a monotonic relationship Conclusions based on correlation tests sometimes ignore that stations at mid-elevation actually have more species than stations at low elevation To analyze the general vanation of species richness with elevation, a minimum requirement for any data set is that It includes data spanning the entire gradient, albeit it becomes lncreasmgly difficult to find appropnate gradients with continuous natural habitat along the entire gradient This is es{)ecially a problem with respect to lowland stations, as lowlands and foothills often are the most disturbed elevational zone(s) Of the 163 data sets, 47 do not include data from below 500 m. In the descriptive analysis of the elevational pattern of species nchness, I have only included data sets which are based on a gradient from below 500 m to above 1500 m (see Table 3) This hmits the analysis to 90 data sets of the onginal 163 37 on invertebrates, 26 on vertebrates (lncludmg 19 on birds), and 27 on plants. As for the entire data set, this subsample of data sets is biased toward the tropics with 73 data sets compared to only 17 from non-tropical biomes It also has more data sets from mainland (n = 71) than islands (n = 19) Non-standardized data sets are dominant (n = 49) In only 13 data sets (14%) have attempts been made to take the effect of area and sampling into account. The corresponding figure for tropical data sets, which have been the main source of generalizations on the elevational gradient, is only 7% (5 of 73 tropical data sets), in contrast to 47% of the nontropical data sets ECOGRAPHY 18 2 (1995)

Total regional

Total

Other

7 6 10

Local Invertebrates Vertebrates Plants Total local

Increasing

To conduct proper descnpuve statistical analyses of the vanation of species nchness with elevation, the stations (l e the data points) must be reasonably equally distnbuted over the gradient, and the number of data points sufficient to refiect any marked changes in habitats/ biomes over the analyzed gradient. Unfortunately these two requirements for an optimal data set are rarely fulfilled As It IS difficult to judge especially the latter cntenon for most of the published data no attempt has been made to select or exclude data on this basis. Compansons of elevational patterns between taxa, latitudinal climatic zones, biogeographic regions or mainland versus islands could be misleading without correction for the area effect and differences in sampling regime Compansons of studies are also biased by differences in the species included, and sometimes further by limitations to Sjjecific trophic levels, guilds or habitat These problems and the pronounced heterogeneity of the quality of data sets make it difficult to conduct proper cntical statistical tests for each data set of the relationship between sp>ecies nchness and elevation that are mutually comparable Still, disregarding these biases, a rough companson based on a classification of the pattern in each data set by visual examination of bivanate plots serves to illustrate how ambiguous the pattern is, both withm region, spatial scale and crude taxonomic groupmgs (Table 3). A decline in species nchness with elevation seems to be a general trend Yet, a pattern where the speciesrichness curve IS almost honzontal up to a certain elevaUon before declining, or is hump-shaped, seems more typical than a monotomc declme (Table 3)

203

1000

2000

3000

4000

5000

ELEVATION (M) Fig 3 Patterns of species nchness versus elevation may vary within the same area for different taxa, here illustrated by New Guinean bats (*) and rodents (â&#x20AC;˘) (data compiled <Tom Flannery [1990]) Data are not standardized for area and sampling effort Lines are fitted by distance-weighted least-squares smoothing

Just a reflection of the latitudinal gradient? The decrease in sjjecies nchness from the equator toward the poles is one of the most universal biogeographic patterns This pattern has been shown across an array of taxa in aquatic as well as terrestnal ecosystems Exceptions are few, and typically restncted to taxa with relatively few species (Rohde 1992) At first glance, the elevational gradient appears to share many climatic charactenstics with the latitudinal gradient Thus, Stevens (1992) claimed that ecologically restnctive climatic conditions appear to increase with elevation as they do with latitude Furthermore, the harsh climate and relatively low species nchness at the equatorial treeline seem to resemble the corresponding conditions found at northern and southern temperate latitudinal zones The apparent resemblance of climate at high elevauons and extreme latitudes makes it intuitively tempting to expect the elevational gradient of species nchness to just mirror of that of the latitudinal gradient The impression of a mirrored relationship is also indirectly imposed upon us by terminology traditionally used, as we often divide tropical mountains into tropical, subtropical, temperate and arctic (or alpine) zones (eg Chapman 1917). Yet, detailed analyses and compansons of vanation in biologically significant climatic parameters between the latitudinal and elevational gradient will reveal several important differences For example, an important feature of climatic vanation on the latitudinal gradient is the increased seasonality towards the poles In contrast to this, mean tem-

204

perature remains fairly constant year round within the same bands on a tropical elevational gradient, and variation in temperature regime seems pnmanly to be diurnal Interestingly, both types of vanation in temperature regime are represented on temperate elevational gradients One would expect this dissimilanty to have different impact on fwpulation maintenance processes on the two gradients and speciation Actually, the existence of a "plateau" or a "hump" on a curve companng species nchness with elevation should not be regarded as unexpected considenng that - although temperature declines with elevation, another lifesupport factor, stable water supply, increases (at least to a certain elevation) Most elevational gradients have a more or less stable condensation zone (cloud zone) at a certain level, especially conspicuous in the tropics, causing favorable conditions for certain taxa (e.g ephiphytes at mid-elevation, which in turn create microhabitats and food for other taxa) As local climate can vary prominently over a few kilometers or hundred meters (e g between opfwsite slopes of the same mountain) in the tropics, the exact location of such a "climatic optimum" can vary considerable regionally and locally, causing differences in the shape of the elevational gradient even within the same taxa The latitudinal gradient does not have such a "humidity peak" One could also add that habitat fragmentation necessanly increases with elevation but not necessarily with latitude. Altogether, there seem to be no a pnon reasons to believe that, climatically, the elevational gradient simply mirrors the latitudinal gradient A negative correlation between species richness and elevation fits well with the general acceptance that the lowland tropical rain forest has the nchest biota on Earth (e g. MacArthur 1972, Erwin 1988). Recent research has shown that this may not always be true on a regional scale (South Amencan mammals [Mares 1992], and birds [Rahbek unpubl ]). Histoncally, a monotonic decline in Sjjecies nchness with elevation corresponds well with the many theones suggesting mechanisms by which increased energy availability often results in proliferation of different species rather than larger populations of existing species (e.g Hutchinson 1959, Pteston 1962, MacArthur 1972, Brown and Gibson 1983, Wnght 1983) We now know that such an increase in species nchness with productivity is not universal (e.g Rosenzweig 1971, Carson and Barrett 1988) In fact, it has recently been suggested that the relationship is hump-shaped (Rosenzweig 1992, Rosenzweig and Abramsky 1993)

Final remarks Understanding elevational patterns must be based on well collected qualitative data, and explanations must pnmarlly be based on unravelling pnmary mechanisms, such as ECOGRAPHY 18 2 (1995)

physical causes, including climatic factors, the narrow width of the elevational gradient, and histoncal perturbations that shape the available species pool Biotic mteractions are secondary mechamsms that can be reflected in the emergent patterns The vanation of sjjecies nchness with elevation might be connected to the reduction of temjjerature with elevation and the assumed corresfwndlng reduction in productivity However, other factors such as vanation in steepness, geological perturbations, alterations of precipitation patterns, etc might also be involved, probably with varying impact from case to case For any correlation of species nchness with variables measured over an elevational gradient, we need to examine whether the correlation reflects a direct coupling, or if It could be a result of interactions of several other factors

The complete list of papers which this Forum contnbution is based on is available from the author

References

Begon, M , Harper, J L and Townsend, C R 1990 Ecology Individuals, populations and communities (2nd e d ) Blackwell. Oxford Brown, J H 1988 Species diversity - In Myers, A A and Giller, P S (eds). Analytical biogeography - an integrated approach to the study of animal and plant distnbution Chapman and Hall, New York, pp 57-89 - and Gibson. A C 1983 Biogeography - C V Mosby, St Louis Carson, W P and Barrett, G W 1988 Succession in old-field plant communities effects of contrasting types of nutrient The observation that high elevation suppwrts fewer ennchment - Ecology 69 984-994 species than low elevation, which indirectly acted as a Chapman, F M 1917 The distnbution of bird-life in Colombia a contnbution to a biological survey of South Amenca catalyst for the belief in a general monotonic decline, Bull Amer Mus Nat Hist 36 1-729 seems to be a general pattern (Table 3) Still, the pattern Colwell, R K and Hurtt, G C 1994 Nonbiological gradients in of species nchness at low- and mid-elevations may differ sfiecies richness and a spunous Rapoport effect - Am Nat between taxa as well as within taxa between different 144 570-595 regions, and within the same region, at least on a regional Diamond, J M 1972 Avifaunaof the eastern highlands of New Guinea - Nuttall Omith Club 11 1^38 scale (see example in Fig 3) It is important to discnmiL 1988 The tropical forest canopy the heart of biotic' nate between patterns refiecting recent diversification Erwin.T diversity - In Wilson, E O (ed), Biodiversity National and those reflecting long-term accumulation of species Academy Press, Washington. D C . pp 123-129 (Fjeldsa 1994) The latter could well be an equilibrium, FjeldsS, J 1994 Geographical patterns for relict and young species of birds m Afnca and South Amenca and implicaprovided that we compare areas that not only have similar tions for conservation pnonties - Biodiv Conserv 3 207average conditions but also resemble each other in habitat 226 mosaicism and dynamism This could be a reason why Flanner>', T 1990 Mammals of New Guinea - Robert Brown the position of humps or bend of curves vanes between and Ass . Sydney different sets of data Vanous taxa are also differently Hutchinson. G E 1959 Homage to Santa Rosalia, or "why are there so many kinds of animals''" - A m Nat 93 145-159 affected by abiotic factors, such as, for example, humidKikkawa. J and Williams, E E 1971 Altitude distnbution of ity land birds in New Guinea - Search 2 64-65 Further studies, including analyses of pnmary-level MacArthur. R H 1972 Geographical ecology - Harper and Rowe Publ , New York processes that could influence the pattern, are needed to Mares. M A 1992 Neotropical mammals and the myth of reveal whether general patterns exit within biogeographic Amazonian biodiversity - Science 25 976-979 regions, taxa, spatial scale, mainland versus islands, etc , Preston, F W 1962 The canonical distribution of commonness and rarity - Ecology 43 185-215, 410-^32 or various combinations of these Appropriate null-models should be considered before explaining elevational Rohde, K 1992 Latitudinal gradients in species diversity the search for the pnmary cause - Oikos 65 514-527 gradient patterns as results of climatic, biological and Rosenzweig, M L 1971 Paradox of ennchment destabilization histoncal processes (Colwell and Hurtt 1994) Although a of exploitation ecosystem in ecological time - Science 171 385-387 difficult task, the development of testable hypotheses is - 1992 Species diversity gradients we know more and less important to achieve significant progress within this field than we thought - J Mammal 73 715-730 and to contnbute to our general understanding of di- and Abramsky. Z 1993 How are diversity and productivity versity patterns This would be preferable to founding related"' - In Ricklefs, R and Schluter, D (eds). Species diversity in ecological communities Histoncal and geonew generalizations on simple accumulations of case graphical perspectives Univ of Chicago Press, Chicago, pp studies Much is still to be learned about this topic, for 52-65 now, we have to accept the unsatisfactory realization that Stevens. G C 1992 The elevational gradient in altitudmal we do not know whether a general relationship exists range an extension of Rapoport's latitudinal rule to altitude between species nchness and elevation, or whether an - A m Nat 140 893-911 Terborgh. J 1977 Bird species diversity on an Andean elevauniversal explanation or model can be given tional gradient - Ecology 58 1007-1019 Wiens, J A 1981 Scale problems in avian censusing - S t u d Avian Biol 6 513-521 Acknowledgements - 1 thank G Graves and J Fjeldsa for many discussions and cnUcally reading the manuscnpt Also thanks to World Conservation Monitonng Centre 1992 Global diversity status of the Earth's living resources - Chapman and Hall, J Lovett,J M CardosadaSilva,R Zusi,D ThurberandM E London Petersen for their cntical comments to the manuscnpt The study was conducted dunng a stay at The National Museum of Natural Wnght, D H 1983 Species-energy theory an extension of species-area theory - Oikos 41 496-506 History, Smithsonian Instituuon supported by a Fulbnght grant ECOGRAPHY 18 2(199'>)

205

ECOGRAPHY 29: 375 384, 2006

Altitude and woody cover control recruitment of Helleborus foetidus in a Mediterranean mountain area Jose´ M. Ramı´rez, Pedro J. Rey, Julio M. Alca´ntara and Alfonso M. Sa´nchez-Lafuente

Ramı´rez, J. M., Rey, P. J., Alcańtara, J. M. and Sańchez-Lafuente, A. M. 2006. Altitude and woody cover control recruitment of Helleborus foetidus in a Mediterranean mountain area. Ecography 29: 375 384. This study explores how variation of macro- and micro-climatic conditions associated with changes in altitude affect early recruitment dynamics of the perennial herb Helleborus foetidus (Ranunculaceae). We also analyse the relevance of facilitation by woody vegetation on seedling recruitment along altitudinal gradient. We conducted a sowing experiment testing the effect of altitude (using three populations located at 1100, 1400 and 1650 m a.s.l.) and woody cover (open areas vs cover of woody species) on seedling emergence during two years and survival three years after sowing. Simultaneously, we characterised elevations and cover types in terms of climatic factors (surface air temperature and relative humidity) throughout a whole year, and light conditions (global site factor and red/infrared ligh ratio) using hemispheric photographs. We detected a significant effect of elevation on seedling emergence, with a higher emergence at lowest altitude. Woody cover greatly affected seedling survival and recruitment, both rates being higher under woody species than in open areas. Emergence was negatively correlated with winter stress factors, which increased with elevation. Survival and recruitment were negatively correlated with summer stress factors, which were ameliorated by woody cover and with altitude. Amelioration of climatic factors by woody cover was not influenced by altitude. Implications for species persistence in Mediterranean mountains under climate change scenarios are discussed. J. M. Ramı´rez (jmrpardo@ujaen.es), P. J. Rey and J. M. Alcańtara, Dept de Biologıá Animal, Biologıá Vegetal y Ecologıá, Univ. de Jaeń, ES-23071 Jaeń, Spain. A. M. Sańchez-Lafuente, Dept de Biologıá Vegetal y Ecologıá, Univ. de Sevilla, ES-41012 Sevilla, Spain.

Seed and seedling mortality approaches 100% in many plant species (Harcombe 1987, Houle 1994, Hampe and Arroyo 2002, Navarro and Guitia´n 2003) and may affect plant population dynamics (Fenner and Kitajima 1999). Plant performance during the early life cycle stages is sensitive to environmental factors. Biological (e.g. herbivores, fungi) or micro-environmental mortality factors (e.g. low light) will usually be restricted to small spatial scales and rarely affect full cohorts of seeds or seedlings (Hulme 1997, Figueroa and Castro 2000, Alca´ntara et al. 2000, Manzaneda et al. 2005). In contrast, macroclimatic conditions may affect all individuals in a given

cohort in a large area. In the Mediterranean area climatic conditions are considered a major cause of the high seedling mortality rates of many plant species (Herrera et al. 1994, Rey and Alca´ntara 2000, Garrido 2003). Mediterranean mountain climate imposes two seasons of extreme conditions; droughts in summer and frosts in winter, both of which are well-known to slow down plant growth (Merino and Infante 2004, Montserrat-Martı´ et al. 2004). However, the combined effect of both periods on seedling survival, establishment and regeneration dynamics of plant populations in the Mediterranean area is still largely unknown.

Accepted 7 March 2006 Copyright # ECOGRAPHY 2006 ISSN 0906-7590 ECOGRAPHY 29:3 (2006)

375

In regions with harsh environmental conditions for a given species, small-scale spatial heterogeneity in the environment may allow the ocurrence of microhabitat patches suitable for seedling emergence and establishment (Harper 1977, Fenner and Kitajima 1999, Go´mezAparicio et al. 2005). One of the main sources of environmental heterogeneity may be vegetation itself. Thus, temperature, humidity, or water availability beneath a plant, are different to those in its immediate surroundings (Joffre and Rambal 1993, Moro et al. 1997, Breshears et al. 1998, Maestre et al. 2003). Although seedling establishment is often reduced by competition (Bullock 2000, Kitajima and Fenner 2000), the presence of other plants may sometimes be beneficial, i.e. facilitation (Callaway and Walker 1997, Garcı´a et al. 2000, Maestre et al. 2003). Several authors have proposed a dynamic balance between beneficial and detrimental interactions, with facilitation predominant in stressful environments and competition predominant under productive conditions (Bertness and Callaway 1994, Pugnaire and Luque 2001, Callaway et al. 2002). However, if such a shift occurs is currently highly controversial (Casper 1996, Pennings et al. 2003, Maestre et al. 2005). Most examples of seedling facilitation refer to long-lived woody plants (Niering et al. 1963, Callaway 1998, Maestre et al. 2003, Rey et al. 2004) with little available data for herbs (but see Pugnaire et al. 1996, Callaway et al. 2002, Sans et al. 2002, Espigares et al. 2004). The present study investigated recruitment of the perennial herb Helleborus foetidus (Ranunculaceae) along its altitudinal range in southern Spain. In particular, we addressed three main issues. First, we studied the effects of climatic conditions on seedling recruitment in H. foetidus populations along an altitudinal gradient by assessing the most stressful situations faced by seedlings at different altitudes. Second, we analysed the role of facilitation as a determinant for seedling recruitment by considering if woody plant cover enhances recruitment. Third, we assessed if facilitation becomes more important when environmental conditions are more stressful.

Methods Study species and site Helleborus foetidus (Ranunculaceae) is a perennial herb widely distributed in western Europe (Werner and Ebel 1994). In the Iberian Peninsula, it typically grows in clearings, understory of forests, and scrublands. Plants consist of one or a few ramets, each producing a single terminal inflorescence after several years of vegetative growth. Flowering takes place during January March. Seed mass ranges between 4.57 and 17.46 mg (Garrido et al. 2002). Seeds have an extended dormancy, with the 376

vast majority germinating during the second spring after entering the seed bank. Seed germination and seedling emergence occur from late December to early May, with a peak in March (Garrido 2003). Details of the floral biology, and seed and seedling ecology in the Iberian Peninsula can be found in Herrera et al. (2001, 2002), Garrido et al. (2002, 2005), Guitia´n et al. (2003), Fedriani et al. (2004). The present study was conducted in Sierra Ma´gina Natural Park (Jae´n Province), a mountainous area in SE Spain (37843?N 3830?W), between October 2000 and October 2003. The area is occupied by a mixed oak (Quercus ilex and Q. faginea ) forest that at high altitude changes to spiny-shrub vegetation dominated by Q. ilex , Berberis hispanica, Crataegus monogyna, and Erinacea anthyllis ). At all altitudes, the vegetation is a mosaic of discrete dense scrub patches and open areas without shrubs.

Experimental design We selected three populations along the altitudinal range of the species in the study area: Castillejo (1100 m a.s.l.), Pasaı´lla (1400 m a.s.l.) and Puerto de La Mata (1650 m a.s.l.). Sites were chosen within a single catchment, so that environmental conditions other than altitude were similar between sites: orientation (south-facing), slope ( B/5%), parent material (limestone), and soil (calcareous lithosols). Horizontal distance between consecutive sites was ca 1.5 km. To analyse the effects of elevation and woody cover on seedling emergence and survival, we established 10 randomised experimental blocks in Pasaı´lla and Puerto de La Mata, and 9 in Castillejo. Each block was set around a scrub patch of several meters of diameter and comprised a set of four plots of 15 /15 cm, each covered by wire mesh to avoid predation by rodents. Two plots were placed under scrub cover and the other two were placed 2 3 m apart, outside the cover of scrubs (‘‘open areas’’). Blocks within site were 10 30 m from each other. In each plot we sowed between 34 and 40 seeds in November 2000. Between January and May 2002 and 2003 we registered seedling emergence by monthly counts. By the end of 2003 (i.e. three years after the seeds were sowed) we registered seedling survival.

Abiotic conditions During 2002 we characterised the light environment and microclimate of the study sites. We recorded surface air temperature and relative humidity at hourly intervals using Omega RH-temp OM-44 sensors. At each altitude sensors were set up in two points for each cover type (presence/absence of woody cover), thus climatic data ECOGRAPHY 29:3 (2006)

were not associated to sowing plots. Using a digital camera Nikon Coolpix 995 with a fish-eye lens we also took hemispheric photographs of each experimental plot in order to characterise the light conditions of each sowing plot. Photographs were subsequently analysed using Winphot (ter Steege 1996) to calculate the global site factor (GSF) and red/infrared ratio (R/IR). GSF is the fraction of radiation that reaches the soil surface relative to the amount of radiation above the canopy. It is generally used to assess light availability at small scales and is inversely related to canopy plant cover (Clark et al. 1996, Myers et al. 2000). Hence, we used GSF to estimate vegetation cover. R/IR was used as an additional measure of the light environment that may affect germination (Frankland and Poo 1980, Smith 1982, Larcher 1995).

Data analyses The following dependent variables were considered: seedling emergence (seedlings emerged/seeds sown), seedling survival (seedlings alive/seedlings emerged) and seedling recruitment (seedlings alive/seeds sown). All dependent variables were analysed by linear mixed models (Proc MIXED in SAS; Anon. 2003) after angular transformation of the original data. Altitude and woody cover were considered as fixed factors. Blocks, nested within altitude, were used as random factors. We used the ESTIMATE option in Proc MIXED to analyse effect size of a given factor when any factor was significant; effect size is calculated as the predicted difference of means between the extreme levels of the factor, providing also a t-test for significance of that effect size. When we found a significant interaction effect, we performed tests of simple main effects (SLICE option in Proc MIXED) to compare the levels of the interaction. Data on environmental variables were analysed in relation to altitude and woody cover. Variation in GSF and R/IR (n /174) was analysed by linear mixed models (Procedure MIXED, SAS/STAT) in a similar way as

indicated above. We computed maximum, minimum and mean monthly temperatures (by averaging daily records), the length of time with temperature 5/08C (‘‘frost period’’, FP), the number of hours per month with relative humidity /100% (RH100) or B/25% (RH25). All these variables were analysed with repeated measures ANOVAs (GLM in STATISTICA; Anon. 2004), considering the month as the within-subjects effect. Finally, we tested for direct relationships between the abiotic variables and Helleborus emergence, survival and recruitment. Each dependent variable was included in logistic regression models with GSF and R/IR as independent variables. Logistic models yield a generalised determination coefficient (‘‘R-square’’ in SAS output; Procedure LOGISTIC) whose maximum value depends on the specific model, and is always B/1 for discrete models. For example, for logistic regression, in a dataset with 50% y /1 and 50% y /0 observations, the maximum possible value of this generalised determination coefficient would be 0.75, which would be reached if all observations were correctly predicted by the model (Nagelkerke 1991). Because this is a serious limitation for a R2 coefficient, Nagelkerke (1991) proposed an adjusted R2 (‘‘Max-rescaled R-Square’’ in SAS output), which achieves a maximum value of 1, and will be referred to when discussing R2 in logistic regression models. We also computed the Spearman rank correlations between climate and seedling performance variables using their averages, obtained for each altitude by woody cover combination (n /6).

Results Sowing experiment A total of 458 seedlings emerged (10.0% of the seeds sown), 190 of which were found in open areas, and 268 under woody cover. There was a significant effect of altitude on the probability of emergence (emergence was higher at 1100 m than at 1400 and 1650 m; Table 1A; model estimated means and 95% confidence intervals of

Table 1. Results of the linear mixed model testing the effects of elevation and woody cover on seedling emergence, survival and recruitment. Block (nested in elevation) was treated as a random factor. Effects

Elevation Woody cover E /W

(A) Emergence

ECOGRAPHY 29:3 (2006)

Error DF

2 1 2

26 84 84

8.95 2.39 1.11

0.001 0.126 0.335

30.04 57 57

1.55 27.17 0.46

0.229 B/0.0001 0.624

26 84 84

2.86 30.03 2.30

0.076 B/0.0001 0.106

1.69

0.091

1.65

0.100

1.40

0.160

Random effects Block (Elevation)

(B) Survival

377

(560 h yr 1). The mean surface air temperature was also influenced by the interaction between elevation and woody cover, but this effect was not significant for maximum or minimum temperature. As shown in Fig. 1, mean surface air temperature during the hottest months was buffered by the cover of woody plants, but the buffering effect decreased with decreasing mean temperature. As a consequence the difference between mean temperature under woody cover and in open areas was 4.08C at 1100 m, but only 1.98C at 1650 m. RH100 varied with elevation and woody cover (Table 2). RH100 was longer at higher elevations, although differences disappeared during the summer months (Fig. 1c). For RH25 the effect of elevation was only marginally significant (Table 2), albeit differences between elevations were more pronounced in summer, when low humidity conditions were more prolonged at lower elevations. RH100 was longer in open areas than under woody cover, but differences were low during spring and most of the summer (Fig. 1g; Table 2). Nevertheless, the most relevant effect of woody cover on relative humidity was a reduction in RH25, particularly during spring and summer (Fig. 1h). The interaction between elevation and woody cover was significant for GSF (Table 2, Fig. 2a). Thus, GSF varied among elevations for each cover type (tests of simple main effects: F2,84 ]/4.50, p5/0.01, for both cover types), and among cover types for a given altitude (F1,84 ]/ 67.69, pB/0.0001, for all altitudes). We found higher GSF in open areas than under woody cover at all elevations (Fig. 2a). GSF was highest at 1650 m in open areas and lowest at 1400 m under woody cover. Similarly, impact of woody cover on R/IR was related to altitude (Table 2, see also Fig. 2). R/IR increased with elevation in open areas (test of simple main effects: F2,84 /58.1, p B/0.0001; Fig. 2b). Although R/IR also differed among elevations under woody cover (test of simple main effects: F2,84 /30.28, pB/0.0001), there were no differences between mid and low altitude.

emergence proportion were 0.14 [0.09, 0.20] at 1100 m, 0.05 [0.03, 0.10] at 1400 m, and 0.03 [0.01, 0.06] at 1650 m). Increase in emergence probability between the extremes of the gradient was 0.12 (t26 /4.18, p /0.0003, effect size analysis). However, we found no significant effect of woody cover; a result that was consistent at all elevations, as suggested by the absence of a significant interaction effect. In contrast, seedling survival was unrelated to altitude, but significantly affected by woody cover (Table 1B; model estimated means and 95% CI were 0.003 [ /0.007, 0.036] in open areas and 0.22 [0.12, 0.34] under woody cover; t60.6 /5.85, pB/0.0001, effect size analysis). This result was consistent at all elevations (non-significant interaction on survival, Table 1B). Finally, most variation found in seedling recruitment was related to presence/absence of woody cover. Recruitment was significantly higher under woody cover than in open areas (Table 1C; model estimated means and 95% CI were 0.0001 [ /0.0004, 0.0017] in open areas and 0.013 [0.007, 0.021] under woody cover), although the magnitude of the difference was small (0.01; t84 /5.48, pB/0.0001, effect size analysis).

Abiotic conditions Table 2 shows the results of analyses of variation in abiotic environmental variables registered in the experimental sites. Surface air temperature decreased with altitude (Fig. 1a), thus the site at highest altitude experienced harder conditions in winter, with longer frost periods (996 h yr 1 at 1650 m and 302 h yr 1 at 1100 m), but milder conditions in summer (mean daily temperature was 58C lower at 1650 m than at 1100 m between June and August). Mean and maximum temperature were higher in open areas than under woody plants, particularly in summer (Fig. 1e). However, minimum temperature was also lower in open areas than under woody cover, with a longer frost period in open areas (706 h yr 1) than under woody cover

Table 2. Effects of elevation and woody cover on the variation in abiotic conditions. Effects on climatic variables were obtained by repeated measures ANOVAs, using month as the within-subjects effect (pB/0.0001 in all analyses). Results of the variables related to light environment (GSF and R/IR) were obtained from linear mixed models, using block as a random effect (p 5/0.01 in both analyses). Variables

Elevation DF

58.43 0.0001 3.98 0.0794 55.39 0.0001 84.96 B/0.0001 36.71 0.0004 3.60 0.0937

DF 1, 1, 1, 1, 1, 1,

6 6 6 6 6 6

78.34 0.0001 50.01 0.0004 25.27 0.0024 11.34 0.0151 23.89 0.0027 112.26 B/0.0001

DF 2, 2, 2, 2, 2, 2,

6 6 6 6 6 6

Mean monthly temperature Maximum monthly temperature Minimum monthly temperature Duration of frost period (FP) Time periods with relative humidity of 100% (RH100) Time periods with relative humidity B/25% (RH25)

2, 2, 2, 2, 2, 2,

Global site factor (GSF) Red/infrared ratio (R/IR)

2, 26 7.77 0.0023 1, 84 338.43 B/0.0001 2, 84 6.19 2, 26 53.60 B/0.0001 1, 84 209.39 B/0.0001 2, 84 17.50

378

6 6 6 6 6 6

E /W

Woody cover

9.56 4.78 0.81 0.775 2.44 1.94

p 0.0136 0.0573 0.4881 0.5019 0.1677 0.2227 0.0031 B/0.0001

ECOGRAPHY 29:3 (2006)

Fig. 1. Climatic characterization of the experimental sites. Temperatures expressed in 8C and time in hours. Data are predicted means (with 95% CI) obtained by repeated measures ANOVAs. RH100: number of hours with 100% relative humidity; RH25: number of hours with relative humidity values below 25%. n /12.

Plant responses to the abiotic conditions Seedling emergence, survival, and recruitment were all negatively related to GSF and R/IR (Table 3). However, the variance explained by the fitted models was always low (0.015/R2 5/0.21 in all cases, see Table 3), which does not suggest a primary role of these abiotic variables for the spatial recruitment dynamics. Regarding climatic variables (Fig. 3), seedling emergence was negatively related to the duration of the frost periods (rs /0.94, p /0.005) and positively to minimum temperature (rs / /0.94, p /0.005). Seedling survival was negatively related to summer RH25 (averaged through June to August; rs / /0.94, p /0.005) and maximum surface air temperature during summer months (rs / /0.77, p /0.07). Similarly, seedling recruitment was negatively related to the duration of summer RH25 (rs / /0.83, p /0.04). The high correlation coefficients suggests a primary effect of the climatic variables on the recruitment dynamics, in contrast to the results for light.

Discussion Our investigation describes how climatic and abiotic micro-environmental factors undergo altitudinal and cover type variation, thereby causing different stress ECOGRAPHY 29:3 (2006)

situations during a year. Open areas became more extreme microhabitats in summer, with highest maximum and minimum temperatures and lowest relative humidity. The effect of altitude, however, seems to be responsible for two different extreme situations throughout the year. Thus, populations located at higher altitude undergo milder environmental conditions in summer (lower maximum and minimum temperatures, and shorter periods with low relative humidity). However, conditions at high altitude are harder in winter (lowest temperatures and longer frost periods). A contrasting situation occurs in populations located at lower altitude. Furthermore, the climatic effects of altitude and woody cover were independent of each other, thereby causing a mosaic of micro-climatic conditions when combined. As in H. foetidus, seedling emergence in many Mediterranean plants occurs in winter, e.g. Phillyrea latifolia (Herrera et al. 1994), Muscari comosum , M. neglectum , M. commutatum , M. weissii (Doussi and Thanos 2002), Olea europaea (AlcaÂ´ntara and Rey 2003), Rhamnus alaternus (Gulias et al. 2004), Quercus ilex (GoÂ´mez 2004), Cistus ladanifer, Erica umbellata , and Rosmarinus officinalis (Quintana et al. 2004). The results of this study suggest that low winter temperatures may become an important limiting factor for seedling emergence. Germination success was inversely related to altitude, with more seeds germinating at lower altitudes, and was also negatively related to the duration and 379

Fig. 2. GSF (global site factor) and R/IR (red/infrared ratio) variation related to elevation and woody cover. Within each plot, different letters indicate significant differences in mean values (obtained in post-hoc tests). Data are model estimated means with 95% CI. n /116.

intensity of the frost periods. The hardest winter conditions at higher elevation coincide with the time of germination and seedling emergence (November through March; Garrido 2003; Fig. 1a, b). The increasingly harder environmental conditions along the elevation gradient shown in this study seem to determine the gradient found for emergence; which in turn may be related to the probability of recruitment (see below). Although mean emergence was slightly higher under

woody plants than in open areas, we did not find any significant effect of woody cover on the probability of emergence, which indicates that vegetation cover did not successfully ameliorate the stress on emergence imposed by the low winter temperatures in the present study. Other studies have shown a significantly higher emergence probability under scrub cover than in open interspaces (Rey and Alca´ntara 2000, Go´mez 2004). It is possible that low emergence values obtained in our study masked any possible effect of woody cover. As expected, summer conditions in our study area impose severe restrictions on seedling survival. Climatic variation along the elevation gradient, however, did not affect seedling survival; thus, even when temperature and humidity variation were attenuated at higher elevations, stressful conditions still occurred. Nevertheless, both seedling survival and recruitment (estimated 3 yr after seed sowing) were higher under woody plant cover than in open areas, which suggest that woody cover offers improved environmental conditions to Helleborus seedlings, as it has been found for different Mediterranean woody species (Garcı´a et al. 2000, Maestre et al. 2003, Castro et al. 2004, Rey et al. 2004). Under Mediterranean climate, high temperatures and very low rainfall occur in summer, which result in high evapotranspiration and photoinhibition (Demming-Adams 1992, Angelopoulos et al. 1996). Seedlings may be particularly affected by these conditions, as they have not yet developed efficient mechanisms to defend themselves against water stress (Kitajima and Fenner 2000). Thus, one of the advantages of woody cover seems to be an attenuation of the extreme summer conditions (Moro et al. 1997, Go´mez-Aparicio 2004, Rey et al. 2004). Our results indicate that the higher the temperature and the lower the humidity, the harder the environmental conditions experienced by seedlings (Fig. 3c, d). Thus, the more effective buffering of temperature and humidity under woody cover was found in summer (Fig. 1e, h). This is also suggested by the negative relationship found between GSF and seedling survival. That is, scrub cover increased seedling survival during summer, as a more dense plant cover restricted light and ameliorated the temperatures, probably reducing the evapotranspiration of seedlings. One of the aims of this study was to explore if a presumed facilitation to seedling establishment, offered

Table 3. Results of the logistic regressions of variables related to light conditions (global site factor and red/infrared ratio) on seedling emergence, survival and recruitment (expressed as proportions). GSF

Emergence Survival Recruitment

380

R/IR

x21

/1.98 /11.34 /10.42

18.47 63.18 73.99

B/0.0001 B/0.0001 B/0.0001

0.008 0.211 0.097

x21

/7.69 /14.18 /17.70

95.33 36.13 80.85

B/0.0001 B/0.0001 B/0.0001

0.043 0.124 0.106

ECOGRAPHY 29:3 (2006)

Fig. 3. Relationship between climatic variables and plant responses (seedling emergence, seedling survival and seedling recruitment, expressed as proportions). Figures by each point indicate elevation of that point. Temperatures expressed in 8C and time in hours. In all cases, a significant Spearman rank correlation coefficient was obtained (pB/0.05). All climatic data are means per month. Minimum temperature: mean of minimum monthly temperatures through the year; frost periods: mean of monthly frost periods through the year; summer maximum temperature: mean of maximum monthly temperatures through the summer; summer RH25: mean of monthly number of hours with relative humidity B/25% through the summer.

by woody plant cover, would be modified by variation in the conditions found at different elevations during the Mediterranean summer. The positive effect of woody cover on survival and recruitment was observed regardless of elevation. Thus, it can be concluded that, although environmental conditions softened at higher elevations, the facilitation effect offered by woody species was still present. Similar conclusions have been obtained in other facilitation studies carried out with woody plants in mountain areas of the Mediterranean region (GoÂ´mez-Aparicio et al. 2004). Further, the effect ECOGRAPHY 29:3 (2006)

of elevation, if present (only marginally on recruitment likelihood), did not affect facilitation negatively; rather, the effects of elevation and woody cover were additive. A number of studies in the Mediterranean region have found that facilitative interactions determined seedling establishment under highly variable conditions of elevation and humidity (Pugnaire et al. 1996, Sans et al. 2002, Rey et al. 2004, GoÂ´mez-Aparicio et al. 2005, Lloret et al. 2005). Consequently, it can be concluded that positive interactions among plants may be ubiquitous in the Mediterranean area. 381

Our results do not confirm the shift in the balance between positive and negative interactions with the change in abiotic stress (Bertness and Callaway 1994, Pugnaire and Luque 2001; but see Maestre and Cortina 2004, Maestre et al. 2005). Relaxing the summer stress (decreasing temperatures and increasing air humidity) at high altitudes did not affect the woody facilitation on seedling survival. It is possible that the softening of summer conditions in our elevation gradient were not sufficient to produce such shift. The distribution range of a plant species may be highly correlated with the geographic variation of the climatic conditions (Beerling 1993, Huntley 1994, Sykes et al. 1996, Gaston 2003). This may be explained by the limiting climatic conditions generally found in the boundaries of a species range, while conditions in the core of the range may be particularly suitable for the maintenance and growth of its populations (Brown 1984, Lawton 1993, Sagarin and Gaines 2002). This paper has considered most of the elevation range occupied by the study species in the Mediterranean mountains, which allows to explore the recruitment dynamics of this species along practically all possible climatic conditions under which it occurs. Our results suggest that the altitudinal distribution of H. foetidus in the Mediterranean mountains may be determined by seedling tolerance to climatic conditions. Both extremes of the year-round climatic variation in Mediterranean mountains, winter and summer, have a clear effect on seedling recruitment. While seedling emergence may be considered a key factor affecting recruitment in winter, seedling survival becomes most important in summer. It is the combination of these two climatic extremes which determines the limits of the distribution range of this plant species. This result is particularly interesting in our study area, because most studies carried out so far in Mediterranean areas have found that the main limiting factor for recruitment is caused by the stressing summer conditions (Garcı´a-Fayos and Verdu´ 1998, Ogaya and Pen˜uelas 2004, Lloret et al. 2004a, b, 2005). This emphasis on the role of the summer conditions disregards the fact that many Mediterranean plant species have an upper or northern distribution limit that is unlikely to be determined by summer climate. We suggest that future studies should pay attention to the role of the climatic conditions during winter as an additional climatic factor influencing population dynamics of Mediterranean plant species, in particular those with seeds germinating during autumn and winter.

Implications for plant species persistence under climate change Climate change models predict an increase of mean annual temperature in Europe during the coming decades, and this warming would be more intense in 382

southern Europe (Anon. 2001). In response species are expected to shift their ranges not only in latitude (Pitelka et al. 1997, Honnay et al. 2002), but also in altitude (Peters and Darling 1985, Peters 1992, Franklin et al. 1992), especially in the Mediterranean basin, where changes in distribution area are expected to be dramatic (Skov and Svenning 2004, Thuiller et al. 2005). Low dispersal capacity of many forest herbs may compromise their ability to track climatic changes (Skov and Svenning 2004), but this limitation will be stronger for latitudinal than for altitudinal migration because of the very different migration distances needed. Recruitment and patch occupancy dynamics in H. foetidus are limited by dispersal (Rey et al. 2005). Skov and Svenning (2004) have predicted a dramatic reduction of H. foetidus ’ distribution area in Mediterranean basin. Thus, we expect altitudinal migration to become critical for maintaining H. foetidus ’ Mediterranean populations. Our results suggest that facilitation by woody cover will become more pronounced with rising summer temperatures. As Fig. 3c e suggests, an increase in summer temperatures will easily lead to zero seedling survival probability and recruitment in open areas, in which case persistence would only be possible under woody cover, regardless of altitude. Thereby, our results highlight the need to account for the effects of community interactions when predicting biotic responses to climate change (Davis et al. 1998, Lortie et al. 2004). Acknowledgements We are grateful to Alejandro Casas for help with the field work and the Junta Rectora of Sierra Ma´gina Natural Park for working permission. This study was supported by grants BOS2000-1122-C03 and BOS2003-03979-C02-02 from Direccioń General de Investigacioń, Ministerio de Ciencia y Tecnologıá. JMR and AMSL were supported by grants FP20006455 and BOS2003-0292, respectively, from Direccioń General de Investigacioń, Ministerio de Ciencia y Tecnologıá.

References Alca´ntara, J. M. and Rey, P. J. 2003. Conflicting selection pressures on seed size: evolutionary ecoloy of fruit size in a bird-dispersed tree, Olea europaea . J. Evol. Biol. 16: 1168 1176. Alca´ntara, J. M. et al. 2000. Early effects of rodent postdispersal seed predation on the outcome of the plant-seed disperser interaction. Oikos 88: 362 370. Angelopoulos, K. et al. 1996. Inhibition of photosynthesis in olive trees (Olea europaea L.) during water stress and rewatering. J. Exp. Bot. 47: 1093 1100. Anon. 2001. Climate change 2001: the scientific basis. Contribution of Working Group I to the Third Assessment Report of the Intergovernmental Panel on Climate Change, Cambridge Univ. Press. Anon. 2003. SAS/STAT software. Ver. 9.1. SAS Inst. Anon. 2004. STATISTICA (data analysis software system), ver. 7. StatSoft. Beerling, D. J. 1993. The impact of the temperature on the northern distribution limits of the introduced species Fallopia japonica and Impatiens glandulifera in north-west Europe. J. Biogeogr. 20: 45 53. ECOGRAPHY 29:3 (2006)

Bertness, M. D. and Callaway, R. 1994. Positive interactions in communities. Trends Ecol. Evol. 9: 191 193. Breshears, D. D. et al. 1998. Effects of woody plants on microclimate in a semiarid woodland: soil temperature and evaporation in canopy and intercanopy patches. Int. J. Plant Sci. 159: 1010 1017. Brown, J. H. 1984. On the relationship between abundance and distribution of species. Am. Nat. 124: 255 279. Bullock, J. M. 2000. Gaps and seedling colonization. In: Fenner, M. (ed.), Seeds: the ecology of regeneration in plant communities. CABI International, pp. 375 395. Callaway, R. M. 1998. Competition and facilitation on elevation gradients in subalpine forests of the northern Rocky Mountains, USA. Oikos 82: 561 573. Callaway, R. M. and Walker, L. R. 1997. Competition and facilitation: a synthetic approach to interactions in plant communities. Ecology 78: 1958 1965. Callaway, R. M. et al. 2002. Positive interactions among alpine plants increase with stress. Nature 417: 844 848. Casper, B. 1996. Demographic consequences of drought in the herbaceous perennial Cryptantha flava : effects of density, associations with shrubs, and plant size. Oecologia 106: 144 152. Castro, J. et al. 2004. Seedling establishment of a boreal tree species (Pinus sylvestris ) at its southernmost distribution limit: consequences of being in a marginal Mediterranean habitat. J. Ecol. 92: 266 277. Clark, D. B. et al. 1996. Landscape scale evaluation of understory light and canopy structure: methods and application in a neotropical lowland rain forest. Can. J. For. Res. 26: 747 757. Davis, A. J. et al. 1998. Making mistakes when predicting shifts in species range in response to global warming. Nature 391: 783 786. Demming-Adams, B. 1992. Photoprotection and other responses of plants to high lights stress. Annu. Rev. Plant Physiol. Plant Mol. Biol. 43: 599 626. Doussi, M. A. and Thanos, C. A. 2002. Ecophysiology of seed germination in Mediterranean geophytes. 1. Muscari spp. Seed Sci. Res. 12: 193 201. Espigares, T. et al. 2004. Is the interaction between Retama sphaerocarpa and its understorey herbaceous vegetation always reciprocally positive? Competition-facilitation shift during Retama establishment. Acta Oecol. 26: 121 128. Fedriani, J. M. et al. 2004. Geographical variation in the potential of mice to constrain an ant-seed dispersal mutualism. Oikos 105: 181 191. Fenner, M. and Kitajima, K. 1999. Seed and seedling ecology. In: Pugnaire, F. I. and Valladares, F. (eds), Handbook of functional plant ecology. Marcel Dekker, pp. 589 621. Figueroa, J. A. and Castro, S. A. 2000. Effect of herbivores and pathogens on the survival and growth of seedlings in a fragment of the Chiloe rainforest, Chile. Rev. Chilena Hist. Nat. 73: 163 173. Frankland, B. and Poo, W. K. 1980. Phytochrome control of seed germination in relation to natural shading. In: De Greef, J. (ed.), Photoreceptors and plant development. Antwerpen Univ. Press, pp. 357 366. Franklin, J. F. et al. 1992. Effects of global climatic change on forests in northwestern North America. In: Peters, R. L. and Lovejoy, T. E. (eds), Global warming and biological diversity. Yale Univ. Press, pp. 244 257. Garcıá, D. et al. 2000. Yew (Taxus baccata L.) regeneration is facilitated by fleshy-fruited shrubs in Mediterranean environments. Biol. Conserv. 95: 31 38. Garcıá-Fayos, P. and Verdu´, M. 1998. Soil seed bank, factors controlling germination and establishment of a Mediterranean shrub: Pistacia lentiscus L. Acta Oecol. 19: 357 366. Garrido, J. L. 2003. Semillas y plańtulas de Helleborus foetidus L. (Ranunculaceae): variacioń geogra´fica, ecologıá y evolucioń. Ph.D thesis, Univ. de Jaeń, Jaeń, Spain. Garrido, J. L. et al. 2002. Geographical variation in diaspore traits of an ant-dispersed plant (Helleborus foetidus ): ECOGRAPHY 29:3 (2006)

are ant community composition and diaspore traits correlated? J. Ecol. 90: 446 455. Garrido, J. L. et al. 2005. Pre- and post-germination determinants of spatial variation in recruitment in the perennial herb Helleborus foetidus L. (Ranunculaceae). J. Ecol. 93: 60 66. Gaston, K. J. 2003. The structure and dynamics of geographic ranges. Oxford Univ. Press. Go´mez, J. M. 2004. Importance of microhabitat and acorn burial on Quercus ilex early recruitment: non-additive effects on multiple demographic processes. Plant Ecol. 172: 287 297. Go´mez-Aparicio, L. 2004. Papel de la heterogeneidad en la regeneracioń del Acer opalus subsp. granatense en la montanã mediterrańea: implicaciones para la conservacioń y restauracioń de sus poblaciones. Ph.D thesis, Univ. de Granada, Granada, Spain. Go´mez-Aparicio, L. et al. 2004. Applying plant facilitation to forest restoration: a meta-analysis of the use of shrubs as nurse plants. Ecol. Appl. 14: 1128 1138. Go´mez-Aparicio, L. et al. 2005. Response of tree seedlings to the abiotic heterogeneity generated by nurse shrubs: an experimental approach at different scales. Ecography 28: 757 768. Guitiań, J. et al. 2003. Variation in structural gender in the hermaphrodite Helleborus foetidus (Ranunculaceae): withinand among-population patterns. Plant Syst. Evol. 241: 139 151. Gulias, J. et al. 2004. Critical stages in the recruitment process of Rhamnus alaternus L. Ann. Bot. 93: 723 731. Hampe, A. and Arroyo, J. 2002. Recruitment and regeneration in populations of an endangered South Iberian Tertiary relict tree. Biol. Conserv. 107: 263 271. Harcombe, P. A. 1987. Tree life tables: simple birth, growth, and dead data encapsulate life histories and ecological roles. BioScience 37: 557 568. Harper, J. L. 1977. Population biology of plants. Academic Press. Herrera, C. M. et al. 1994. Recruitment of a mast-fruiting, birddispersed tree: bridging frugivore activity and seedling establishment. Ecol. Monogr. 64: 315 344. Herrera, C. M. et al. 2001. Geographical variation in autonomous self-pollination levels unrelated to pollinator service in Helleborus foetidus (Ranunculaceae). Am. J. Bot. 88: 1025 1032. Herrera, C. M. et al. 2002. Interaction of pollinators and herbivores on plant fitness suggests a pathway for correlated evolution of mutualism- and antagonism-related traits. Proc. Nat. Acad. Sci. USA 99: 16823 16828. Honnay, O. et al. 2002. Possible effects of habitat fragmentation and climate change on the range of forest plant species. Ecol. Lett. 5: 525 530. Houle, G. 1994. Spatiotemporal patterns in the components of regeneration of four sympatric tree species, Acer rubrum , A. saccharum , Betula alleghaniensis and Fagus grandifolia . J. Ecol. 82: 39 53. Hulme, P. E. 1997. Post-dispersal seed predation and the establishment of vertebrate dispersed plants in Mediterranean scrublands. Oecologia 111: 91 98. Huntley, B. 1994. Plant species’ response to climate change: implications for the conservation of European birds. Ibis 137: S127 S138. Joffre, R. and Rambal, S. 1993. How tree cover influences the water balance of Mediterranean rangelands. Ecology 74: 570 582. Kitajima, K. and Fenner, M. 2000. Ecology of seedling regeneration. In: Fenner, M. (ed.), Seeds: the ecology of regeneration in plant communities. CABI International, pp. 331 359. Larcher, W. 1995. Physiological plant ecology. Springer. Lawton, J. H. 1993. Range, population abundance and conservation. Trends Ecol. Evol. 8: 409 413.

383

Lloret, F. et al. 2004a. Experimental evidence of reduced diversity of seedlings due to climate modification in a Mediterranean-type community. Global Change Biol. 10: 248 258. Lloret, F. et al. 2004b. Establishment of co-existing Mediterranean tree species under a varying soil moisture regime. J. Veg. Sci. 15: 237 244. Lloret, F. et al. 2005. Effects of vegetation canopy and climate on seedling establishment in Mediterranean shrubland. J. Veg. Sci. 16: 67 76. Lortie, C. J. et al. 2004. Rethinking plant community theory. Oikos 107: 433 438. Maestre, F. T. and Cortina, J. 2004. Do positive interactions increase with abiotic stress? A test from a semi-arid steppe. Proc. R. Soc. B (Suppl.) 271: S331 S333. Maestre, F. T. et al. 2003. Positive, negative, and net effects in grass-shrub interactions in Mediterranean semiarid grasslands. Ecology 84: 3186 3197. Maestre, F. T. et al. 2005. Is the change of plant-plant interactions with abiotic stress predictable? A meta-analysis of field results in arid environments. J. Ecol. 93: 748 757. Manzaneda, A. J. et al. 2005. Effects of microsite disturbances and herbivory on seedling performance in the perennial herb Helleborus foetidus (Ranunculaceae). Plant Ecol. 179: 73 82. Merino, J. and Infante, J. M. 2004. Ecofisiologıá. In: Herrera, C. M. (ed.), El monte mediterrańeo en Andalucıá. Consejerıá de Medio Ambiente, Junta de Andalucıá, pp. 47 57. Montserrat-Martı´, G. et al. 2004. Fenologıá y caracterı´sticas funcionales de las plantas lenõsas mediterrańeas. In: Valladares, F. (ed.), Ecologıá del bosque mediterrańeo en un mundo cambiante. Ministerio de Medio Ambiente, EGRAF, S. A., pp. 129 162. Moro, M. J. et al. 1997. Mechanisms of interaction between Retama sphaerocarpa and its understory layer in a semi-arid environment. Ecography 20: 175 184. Myers, G. P. et al. 2000. The influence of canopy gap size on natural regeneration of Brazil nut (Bertholletia excelsa ) in Bolivia. For. Ecol. Manage. 127: 119 128. Nagelkerke, N. J. D. 1991. A note on a general definition of the coefficient of determination. Biometrika 78: 691 692. Navarro, L. and Guitiań, J. 2003. Seed germination and seedling survival of two threatened endemic species of the northwest Iberian peninsula. Biol. Conserv. 109: 313 320. Niering, W. A. et al. 1963. The saguaro: a population in relation to environment. Science 142: 15 23. Ogaya, R. and Pen˜uelas, J. 2004. Phenological patterns of Quercus ilex , Phillyrea latifolia , and Arbutus unedo growing under a field experimental drought. Ecoscience 11: 263 270. Pennings, S. C. et al. 2003. Geographic variation in positive and negative interactions among salt marsh plants. Ecology 84: 1527 1538.

384

Peters, R. L. 1992. Conservation of biological diversity in the face of climate change. In: Peters, R. L. and Lovejoy, T. E. (eds), Global warming and biological diversity. Yale Univ. Press, pp. 15 30. Peters, R. L. and Darling, J. D. S. 1985. The greenhouse effect and nature reserves. BioScience 35: 707 717. Pitelka, L. F. et al. 1997. Plant migration and climate change. Am. Sci. 85: 464 473. Pugnaire, F. I. and Luque, M. T. 2001. Changes in plant interactions along a gradient of environmental stress. Oikos 93: 42 49. Pugnaire, F. I. et al. 1996. Facilitation between higher plant species in a semiarid environment. Ecology 77: 1420 1426. Quintana, J. R. et al. 2004. Time of germination and establishment success after fire of three obligate seeders in a Mediterranean shrubland of central Spain. J. Biogeogr. 31: 241 249. Rey, P. J. and Alca´ntara, J. M. 2000. Recruitment dynamics of a fleshy-fruited plant (Olea europaea ): connecting patterns of seed dispersal to seedling establishment. J. Ecol. 88: 622 633. Rey, P. J. et al. 2004. Seedling establishment in Olea europaea : seed size and microhabitat affect growth and survival. Ecoscience 11: 310 320. Rey, P. J. et al. 2005. Seed- versus microsite-limited recruitment in a myrmecochorous herb. Plant Ecol. DOI: 10.1007/ s11258-005-9066-3. Sagarin, R. D. and Gaines, S. D. 2002. Geographical abundance distributions of coastal invertebrates: using one-dimensional ranges to test biogeographic hypotheses. J. Biogeogr. 29: 985 997. Sans, F. X. et al. 2002. Positive vs. negative interactions in Picris hieracioides L., a mid-successional species of Mediterranean secondary succession. Plant Ecol. 162: 109 122. Skov, F. and Svenning, J. C. 2004. Potential impact of climatic change on the distribution of forest herbs in Europe. Ecography 27: 366 380. Smith, H. 1982. Light quality, photoperception, and plant strategy. Annu. Rev. Plant Physiol. 33: 481 518. Sykes, M. T. et al. 1996. A bioclimatic model for the potential distributions of north European tree species under present and future climates. J. Biogeogr. 23: 203 233. ter Steege, H. 1996. Winphot 5: a programme to analyze vegetation indices, light and light quality from hemispherical photographs. Tropenbos Guyana Rep. 95-2, Tropenbos Guyana Programme. Thuiller, W. et al. 2005. Climate change threats to plant diversity in Europe. Proc. Nat. Acad. Sci. USA 102: 8245 8250. Werner, K. and Ebel, F. 1994. Life-history of the genus Helleborus L (Ranunculaceae). Flora 189: 97 130. Subject Editor: Jens-Christian Svenning .

ECOGRAPHY 29:3 (2006)

ECOGRAPHV 21: 327-,135. Copenhagen 1998

Community structure of crustacean zooplankton in subarctic ponds - effects of altitude and physical heterogeneity Milla Rautio

Rautio. M. 1988. Community structure of crustacean zooplankton in subarctic ponds - elTects of altitude and physical heterogeneity. Bcography 21: 327 335. Crustacean zooplankton (Cladocera. Copcpoda) distribution patterns, commutiity composition and response to altitude, temperature, pH and surface area were studied in 17 hshless subarcUc ponds in the Kilpisjarvi area, NW Finnish Lapland. Despite their harshness, the ponds harboured diverse groups of zoopknkton Altogether 50 spectcs were found from ice-out in June-August 1994. There was both a marked deeline in the speeies number and a change in the composition of pond communities with increasing altitude and decreasing temperature as well as decreasing pH. Pond surface area was lea.st signiticant in determining the species composilion Ponds at low elevations harboured up to 21 species while the fell top ponds usually had < 10 species. Chydoridaen eladocerans were the most dominant group even though their number greatly diminished in ponds above the timber line.

M. Raulio (nnUa.rautio@helsinki.fi). Depi oJ Ecology and Svswmatics. Div of Hvdrohiology, FIN-00014 Univ, of Helsinki. Finland

Ponds in subarctic and arctic regions are subjected to great fluctuations in their physical and chemical conditions because they are shallow and small in size, and located in a harsh climate (Hcbert and Hann 1986, Btetchko 1995). During long winters, the ponds are regularly covered by ice. some freezing to the bottom, whereas in summer the water volume often quickly diminishes due to evaporation, and shallow ponds inay totally dry out. Ponds at different elevations also experience regional microclimalic variations even within a small spatial scale resulting from the amount of vegetation and the pond's location to the timber line as well as the regional topography (e.g. Picnitz and Smol 1993). The occurrence t)f aquatic invertebrates in temporary and semilemporary ponds is largely determined by species' tolerance to the changing environmental conditions or their dispersal and colonization capability in adverse periods (e.g. Carter et al. 1980, Hebert and Hann 1986, Girdner and Larson 1995). Among invertebrates, zooplankton have the widest tolerance in tnany

respects, and therefore they often predominate ponds. As they can reproduce fast (e.g. Allan 1976) and arc drought and freezing resistant in certain developmental stages {e.g. Wiggins et al. 1980, Marcus 1990). zooplankton can tolerate a broad range of differen! waters (e.g. Schmitz 1959. Reed 1962. Tash and Armitage 1967, Anderson 1974). Moreover, the "twolevel"' structure of the food chain, that is the absence of vertebrate predators, which is typical for small water bodies, favours zooplankton (Hansson el ai. 1993) and often results in higher species number in ponds than in nearby lakes with fish (Anderson 1971, Arnott and Vanni 199.1). Although the total number of zooplankton in an area may be high, a single pond seldom contains more than a few species that occur in the region (e.g. Patalas 1964. Anderson 1971. Hebert and Hann 1986. Girdner and Larson 1995). Even small changes in altitude may alter the zooplankton species number. According to Patalas (1964) 5-8 mesozooplankton species are usually found at ponds at 1400-1700 m a.s.l. whereas at 2500 3200

327

m a.s.l. 2-8 species are found and >320O m only 1- 4. The community composition of a single pond is also affected by the capacity of the pond to hold a certain community, that is the habitat size and diversity (Brown 1981, Fryer 1985). productivity level (Connell and Orias 1964), and niche number which is detenniiied by competition and predation (Hammer and Sawchyn 1968. Sprules 1972. Ranta 1979, Hebert and Loaring 1980). Subarctic and arctic waters arc among the ecosystems that are thought will first react to the global warming (Boer et al. 1990). Arctic and alpine lakes have therefore recently been studied with respect to global change (e.g. Psenner and Schmidt 1992, Douglas et al. 1994, Smol et al. 1996. Weckstrom et al. 1997). However, there liave been few studies because of their remote location, and apparent ecological and economical insignificance as they are fishless. However, these marginal ecosystems, especially in Fennoscandia. are among the most undisturbed ecosystems as far as the atmospheric deposition is concerned (Ruhling 1992). Acidification and eutrofication have had no significant effect on these waters yet, making them a valuable reference habitat for polluted waters. Ponds also provide a fascinating field for experimental and theoretical ecology because of their high number and spatial patchiness. The objective of this study is firstly to record the crustacean zooplankton community composition of several subarctic ponds in NW Finland, and secondly to test the efiect of altitude, pH. temperature and pond surface area on the species eomposition. The study will provide ecological and biogeographical information for research projects studying the functioning of arctic and alpine water ecosystems.

Seventeen ponds were chosen to be studied at altitudes of between 490 and 940 m a.s.l. (Fig. 1). AH but one of the ponds are crystal clear, their water originating mainly from melting snow and rain. One pond, however, is located in a Sphagnum growing mire and is therefore very stained and humic. Ponds below the timbeiiine are rounded by an abundant macrophyte vegetation which constitutes mainly Care.x and Eriophorum species. In addition, the bottom of these low elevation ponds is eovered by a thick algae carpet which partly floats on the surface, especially among the macrophyte vegetation. The proportion of the macrophytes and algae declines above the timberline so that the ponds on tundra are relatively similar to ponds below the timberline but the ponds situated on iell tops and sides are completely without macrophytes and visible epilithic algae. All the study ponds are fishless to the best of our present knowledge. Samples from the ponds were collected once a week for three months between June and August in 1994. A 50 \xm plankton net was pulled from the suriace to the bottom and back to the surface in the same way in each pond in Older to take semi-quantative and comparable samples. In order to avoid "unclean" samples with pieees of vegetation and fioating algae turfs, the sampling was performed just outside the possible littoral macrophyte zone. Although this technique, especially in low elevation ponds, may not necessarily cateh all the individuals that hide near the sediment or among the macrophyte belt, the results show that typical semiplanktonic species were also caught. During the fording through macrophytes towards the centre of the pond, the water eolumn and sediment were also mixed, which results in the samples most likely to represent the

Material and methods Site description and sampling The Kilpisjarvi region (69°02'N. 20°50'E) in the northwest tip of Finland constitutes the most eastern part of the Scandinavian mountain chain (Fig. 1). As a part of the subarctic region, the maximum summer temperature seldom rises above 15°C, The mean summer (June-August) temperature is 9.0°C whereas the mean annual temperature is only -2,6°C (Jarvinen 1987). The timber line {Betula puheseens tortuosa) follows the 600 m contour. Lakes and ponds are eovered with ice for approximately nine months of the year. The short growing period is, however, eompensated by the sun staying continuously above the horizon for 62 days (during winter it does not rise for 54 days). The area encompasses several high altitude ponds, most of glaeial origin. 328

Fig. I. Locauon of the study sitei by Lake Kilpisjiirvi. Ponds are black and numbered. The timber line (Bciula puheseens tortuosa) follows the 600 m contour line. ECOGRAPHY

Table I, Locations, environmental eharacteiislics and number of species in the study ponds. Values for depth are estimates except for those marked with a star. Values for pH and temperature outside the parenlhoses arc Ihe averages for the season, in tiic parentheses the minimum and ma.\imum values. Minimum values for the lemix'rature are measured duriny the fnvst sampling in .lune except for those marked with a star. Zone

i'ond

A I'ea (ha)

Deplh |m|

Altitude Oil)

1 2 3 4 5 6 7

2.6 2.7 0.2 0.2 1.4 ^).^ (1.1

1.0 2.0 1.0 1.0 2.0 1..^ 0..^

490 510 510 510 5.10 520 570

7.5 7.4 6.5 7.2 7.2 7.2 5.3

8 9 10 11 12

11.1 0.2 0.8 1.3

0.5 1..S 2.0 i.O I.(I

680 680 680 710 710

0,3 O.I 0.6 1,4 0,7

1.0 0.5 7.5* 3,2* 1.0

770 860 850 930 940

(I.X

1.7 1,0 0.5 7,5

674 680 490 940

iiumber

Birch forest

Tundra

Fell top or side

1.1 14 \5 16

17 Mean Median Minimum Maximum

O.I 2.7

species community in the pond, at least in the srnalles! ponds. Temperature and pH measurements were taken from the surface water by using a normal thermometer and a pH pen. Altitude and surface area of each pond were calculated by using a geographical map of the area. Depth was cither cstitnatcd or in the mosl difficult cases measured by snorkelling. Samples were preserved in a 4'^ formaldehyde solution and later identified and counted in the laboratory under a light microscope. All together 52 crustacean species were found of which all but two were used in data analyses. Folyartemia forcipata and Branehinecla paludo.sa (Anostraca) were not quantified because of their large size and net escaping abilities. Five cyclopoid species were not identified but they all clearly represented separate species.

Temperature (C)

Species number

[6.5-7,7] [6,5 -7.81 [6.3 6.6] [6.4 7.3] [7.0 7.4] [7.0 7,.t] [4,8 6.0]

10,4 [1.0 16,0] 12,8 [6,5 16,2] 13,3 [8,2 16 8] 12.2 [4.2 16.0] 12,7 [7,5 17,2] 13,4 [7,5 -18,0] 13,5 [6,0 17,5]

18 18

7.2 6,8 7,2 7.4 7.5

[6.2 7.1] [6.2-7.1] [6.7-7.4] [6.5 7.8J [6.7 7.8]

11,0 [4,2 14,1] 12,0 [4,2 15,2] 10.7 [5,5*- 13.3] 9,3 [2.5-17.2] 9.0 [2.5-16.2]

6.4 6.2 b.^ 6.4 5.9

[5.8 7,1] [5.5 6,5] [6,0 6,7] [6.1 6.9] [5.1-6.0]

8.2 [1,5 10,2 [0,3 9,0 [0.5 8.5 [2,5 10,0 [2,2*

6.8 7.2 5.3 7.5

II.O 10,7 8.2 13,5

15.8] 16,4] 14.6] 14,7] 16,5]

21 14 21 16 10 13 18 16

14 17 9 8 9 II 9

14 14 9 21

pond communities: therefore, species relative abundances were not used but only each species" presence or absence during the samphng season. In the study of the relationship between individual species and environmental factors another multiregressional method was used. CCA (Canonical Correspondence Analyses) assumes the species to have bell-shaped response with respect to environmental variables, this is the case if the ranges of the environmental factors are relatively large (Ter Braak 1986). Data for CCA consisted of absolute numbers of each species at each sampling occasion as well as the absoltite temperature measurements. Due to the inaccuracy of the pH-pen, only the average pH measurement of each pond throughout the season was used. This has not altered the results significantly since the pH variation was low (Table 1).

Results Statistical analyses The relatioitship between the number of species in a pond and environmental parameters was tested with a linear regression analysis. For more accurate results of the relative similarities of zooplankton communities between different ponds. OCA (Detrended Correspondence Analyses) was used. DCA computes coordinates for a set of points so that the shorter the distance between points, the more similar are the communities in ponds that the points represent. DCA was run only to illustrate the differences in species occurrence between tit'OCiRAI'HV 21:.^

The typical characteristics of each pond based on seasonal measurements are shown in Table I. The lowest pH valties were recorded during the thawing period in spring in all but the iiattuall) very acid pond no. 7 as well as in fell top and side ponds vvliich also had relatively low pHs. In these ponds thawing increased pH. After the ice and snow melting period pH varied little in individual ponds. Differences in minimum temperatures are due to the time differences in the ice cover melting process during the first sampling. Some of the ponds at the lowest elevations were already uncovered 329

while most mountain top ponds were still totally covered with ice at the first sampling oecasion. Species number was found to be dependent on the ahitude (p = 0,001). the pH (p-0.017) and the average water temperature during the sampling season (p =

A Poods on tundra (680-710 m) O Ponds on M tops (770-940 mj A Ponds bekwtimbertina (490470 m)

250-,

p=0.001

O13

700

Altitude (m)

250 DCA axes 1

Fig. 3, DCA-plot based on species composition in different ponds during the sampling season,

0,029) (Fig 2). Ponds at lowest elevations had the highest number of species. 21 at most, while mountain top ponds usually had <1() species. Acidic waters contained less species than more alkaline waters as did wariT) ponds in comparison with cold ponds. Surface area had no significant effect on the number of species (p = 0.174). DCA divided ponds into two major groups with only two exceptions (Fig. 3). Mountain top ponds formed one group while another group consisted of ponds at lower elevations. Pond no. 7 differed significantly from other pond communities. It was the only humie SphagÂŤw/n-rounded pond in the study and also the only one that dried out during the season. The reason that pond no. II was different from other ponds in the low elevation group is probably due to its large size. Sampling, which was performed in one place, may not have reached all the species in the pond. 10 11 12 The structural dissimilarity between two major pond Water temperature groups, indicated by DCA. is mainly due to the distribution patterns of certain species. Most cladocerans only occurred in low elevation ponds as did also the calant)id Euiliaptonms graciloides while copepods Mixodiapiomus laeinialm and Cyclop.s seulifer were only found in mountain top ponds. Individual species preterenees for environmental conditions are shown in CCA ordination in Fig. 4. p=0.131 In the two-dimensional CCA solution the alignment of the pH axes with the first canonical axes indicates that pH was the most important of the four environ3.0 2,5 1.0 1.5 ZO 0.S 0.0 mental factors considered in determining the strueture Area (ha) of the zooplankton community. In the diagram, the Fig, 2, Linear regression analyses. The relationship between points of each species can be perpendicularly projected the speeies number in the study pond and a) alutude, b) pH, c) onto each environmental variable axis. The projection water temperature and <J) surface area.

ECOGRAI'HY 2\-3

330

A)c2

Fig. 4. Ordinalion diagram of the canonical correspondence analysis (CCA) on the relationship between environmental variables and zooplankton species. Filled circles :ire redrawn in the lower left box for clarity. Full taxon names are given in Appendix I. A = Large area.

_ Acan cur O + Oaph Ion

OAIC

Stre ser O Diap braO AlonexiO

O Acan ver

MicrgraQ

High temperature

Ax. 1

Ocycl scu

Holo gib HighpH

High altitude

BythlonO ycl A Akjn cos

point of a species corresponds approxitnalely lo the ponds at intermediate elevations and neutral pH. Large weighted average, i.e. the oplimum of the species with surface area determined the occurrence of the species respect to the environtnental variable (Ter Braak 1987Holopedium gihherum. Ahma quadrangular is, A. guitata. 1992). According to the CCA. high elevatioti ponds A. affiriis,, Lathonura rectiroslris. Bosmina obtusirostris. with a cold water were occupied by copepods Cyclops scutifer, Mixodiaplomus lacinialus and Acanthocydops Fjiryccrcus lamcllatus. F.ucyclops scrrtdalus, Acanthocvcnissicaudis whereas low elevation ponds with relative clops capil/alus, Mcgacychps gigas. M. viridis and the warm water were occtipied by cladocerans Di- unidentified cyclopod E. The cladocerans Akma costata, Rhyncholahma aphano.soma hrachyuruni. Ahnella exigua. Sida crystalUna, Chydorus splmcricus, Cerlodaphnia quadrangula, falcala, Polyphemus pediculus, Drepanolhrix dentScaphoteheris mucromifa, Simocephalus velu/us and ata. Alorwlta nana. Acroperus elongatus and A. harpae Ahma rectangula as well as by the copepods Eiuliap- were situated close to the origin. These species tomus graciloides, Microcycbps gracilis and tinidetitilied occurred in all type of ponds, thus none of the environmental factors studied controlled their distribucyclopods C and D. tion. Low pH determined the occttrrence oi' Acanlholehcris currirostris. Daplniia louf^i.spina. Alonclla cxcisa. Acanthocydops vcmatis and Dunlwvedia crassa. Of these species A. verncdis and D. cms.sa also oeeurred mainly Discussion in ponds with a small surface area. The point for cladoceran Strephcerus scrricaudalus indicated that it Species number in different ponds occurred iti ponds with a relative high temperature, low Decreasing maximum species number per pond with pH and a small area. Bythotrcphes longimanus and the nicreasing altitude is a typical pattern (e.g. Patalas unidentified eyclopods A and B occupied eold water 1964, Hebert and Hann 1986, Raina and Vass 1993). In fcCOCiRAPHY 2l:.l (1998)

331

aluminium. The organic matter of humus is also a the Kilpisjiirvi region the ponds in birch forests harboured 10-20 species whereas the high iiltitude ponds significant source of food tor zooplankton (Salonen and Hammar 1986). hiirboured 8-12 speeies (Table I). The larger the surface area and greater the depth, The CCA analysis indieates that altitude, either directly (dispersal and eolonization abilities) or indirectly the more species are usually found (e.g. Pennak 1958. (low temperature leading to short growing period, Fryer 1985). This pattern is not, however, always true amount of vegetation), was the main factor in deter- for individual species and the size may not directly mining zooplankton eommunities. According to the determine the number of microhabitats. Accordmg to theory of island biogeography (MeArthur and Wilson Whiteside's (1974) experiment with ehydorids (Clado1967). the more isolated the habitat (here pond), the cera). the number and variability of habitats arc more smaller the probability for species to colonize it. espe- significant than the size of the water body for the cially for species like zooplankton that disperse by abundance of zooplankton. By adding turfs covered passive means (Maguire 1963). Therefore, the small with bacteria and algae, the number of ehydorids innumber of species in mountain top ponds can be partly creased from 3-4 to 13. Anderson (1974) obtained explained by the isolation of the water body. More- similar results. He did not find a rclalionship between over, the abundant littoral vegetation and a thick algae surface area and the mesozooplankton abundance. He carpet on the bottom of the low elevation ponds indi- speculated that the reason for this is that ponds are cate the niche diversity to be greater in these ponds more sensitive to the ehanges in weather than lakes: than in the stony and coarse high elevation waters water temperature follows the air temperature, during without vegetation. An abundant vegetation is usually dry seasons ponds relatively easily dry out etc. Contina sign of relatively high productivity (Wetzel 198:^). uously changing conditions inhibit only some species and therefore of the amount of available food re- from dominating the entire community as the duration sourees for zooplankton. Vegetation also serves as a of environmental optima for an individual species is hiding place for zooplankton from planktivorous short. Therefore, more species are usually found in predators (Schwartz et al. 1983) such as insect larvae pond ecosystems than in nearby lakes with more eonand predatorous zooplankton (Cyclopoids. Polyphemi- stant conditions (Anderson 1971). In the Kilpisjarvi dacsl. Although inseet predators were not quantified in ponds too. the surface area was least significant in the study, they were present. Most commonly seen determining the speeies distribution. It is therefore were larvae of the phantom midge (Cbaohorus sp.). probable that some other environmental factor than Corixids, Notonectids and Odonatas. Therefore, surface area would better explain the species distribuzooplankton in low elevation ponds with vegetation tion. In fact, the amount of macrophytes and algae, must have been able to avoid predators more sueccss- including productivity and refuges from predators, fuUy than in high elevation ponds without vegetation, would again serve as good factor. which in turn may have affected the zooplakton community composition in the altiludinal gradient. Low pH caused by human activity is a stress factor Individual species distribution in relation to for most organisms but also naturally acidic waters are environmental factors avoided by many species (e.g. Schindler et al. 1985. Based on the information obtained in the CCAArvola et al. 1986). In the study ponds, the number of ordination. the reasons for certain species distribuspecies deelined with declining pH value (Fig. 2), With tion among study ponds can be estimated. Temperaone exeeption (pond no. 7), lowest pi I values were ture and altitude gradients divide speeies into two separecorded in ponds at high altitude. Therefore, altitude rate groups (Fig. 4). The copepods Cyclops scutifer and may indirectly determine the number of species also m Mixodiaptomiis laeiiiiatus occupied high elevathis case. Also, as the pH values are not particularly tion and cold water ponds. It is strengthened by low except tor pond no. 7. differences in pH may here earlier studies (Reed 1962, Carter 1971) that C- scutifer indicate differences in productivity more than pH per prefers habitats with relatively deep and cold se. Acid lakes are often nutrient-poor eompared to waters, in Kilpisjarvi it was also most abundant in more neutral lakes which leads to low primary producthe two deepest ponds: 3.5 and 6.5 m at 930 m and 850 tivity (e.g. Jansson et al. 1986). m a.s.l.. respectively. In northern Europe C. seutifer The high content of humic substances indicated by stained colour makes pond no, 7 (pH = 5.3) excep- has been found up to 1500 m a.s.l. and in Russia it tional in the study area. According to Sarvala and only occurs in taiga and tundra regions (Rylov 1948). Halsinaho (199O| the influence of acidity is substan- Mixodiaptomiis laciniatus is an obligate high altitude tially modified by humic matter. Humus may amelio- speeies (Dussart 1967) which is found in clear mounrate the toxic effects of heavy metals sueh as labile tain lakes and ponds > 50U m a.s.l. ECOGRAl'HV 21:3

332

Most cladoceran species ure coordinated in the CCA on the left side of the ordination which indicates that they occur in low elevation and relative warm ponds. Temperature has an effect on the lengtli of the lile-cycle of /ooplankton. According to Allan (1976). eUidoceraiis complete their life-cycle in 7-8 days at 2U°C' whereas at 10°C the development from an egg to an adult requires 20-24 days. Therefore, the average water temperature in fell top and side ponds may not have been high enough for cladoccrans to successfully complete their life-cycle, resulting in them being scarce in high elevation ponds. This, however, is not indicated by the hnear regression analyses (Fig. 2). Low temperature also shortens the growing season which in addition with the tolal freezing of the pond from surface to the bottom inhibits the occmrence of higher vegetation in high elevation ponds {Nedler and Pennak 1955. Federley 1972). Maerophytes especially affect the distribution of ehydoridaes and daphniidaes. All chydorid species live mainly in vegetation (Fryer 1985). Of the daphniid species found in Kilpisjarvi ponds Daphnia longispiua was the only species that typically occurs in all kind of waters, whereas all other species arc associated with littoral vegetation (Ward and Whipple 1959). Acroperus liarpae was the only cladoceran that was coordinated at the right side of the CCA ordination, indicating that it was abundant in relative high elevation ponds. Acroperus liarpae is an exception in the family of Chydoridae since it generally occurs in the pelagic zone and in ponds without higher vegetation {Ward and Whipple 1959). Only a few species were found at low pH. Cladoceran Aeantlwleheris curvirostris was only found in pond no. 7 which had the lowest pH-value (5.3). According to Ward and Whipple (1959) it prefers Sphag;»»)/-roundcd humid waters. Also. Daphnia loiijii.spina was very abundant in pond no. 7 as well as the eyclopod Acaiitliocyclops reriialis which occurred in acidic waters in general among the study ponds (Appendix 1). According to Rylov (1948). A. vermdis occurs mainly in shallow bogs rounded by Sphaiinum. It can, however, be found in waters with pH"s > 8 . Although D. lon^iispina seems to be more acidity-toleranl than other Daphnia species (Arvola et al. 1986. IJimonen-Simola and Tolvanen 1987) the high abundance in pond no. 7 may not be directly determined by low pH. Absence of competition IVom macrophyte demanding zooplankton may have resulted the distribution of D. longispiiui m the CX^A ordination. The location of the cladocerans Strehlocenis serrivaudatus and Alonella C-Wisa in several ponds (Appendix I) as well as their location in the CCA ordination (I'ig. 4| indicate that they can tolerate a wide range of pli's. Despite the four environmental factors (temperature, surface area. pH and altitude) used in CCA do

not explain all the reasons for the species distribution. CCA is still a powerful lool. Although the altitude of the pond and water temperature do not always directly determine the species distribution, their impaet on zooplankton species richness is still stronger than other environmental variables', at least via indirect effects. Altitude has an elTecl on temperature and the temperature has an effect on the amount of vegetation, and through vegetation also on the level of prodtiction and niche number. Pond pH sets limits to the species number due to physiological constraints. Surface area, although not signiticant in this study in determining the species distribution, can in some cases be an indicator of niche diversity. The information obtained from CCA is a good start for advanced studies. Therefore, other more relevant environmental factors may be chosen according to the results in this study. Especially the elTect of vegetation requires more studies as well as the productivity level and number and type of predators in each pond. Moreover, as the clear northern waters have been predicted to be the first to respond intensively to climate change (e.g. Pearce 1996. Schmdier et al. 1996) it would be reasonable to experimentally compare the effects of rising temperatures and more intense UVradiation in clear water ecosystems. Studies by Hessen and Sorensen (1990) are already suggesting that zooplankton produce more pigments when exposed to intense IJV-radiation. .•ickiiowlcdiU'iiieiils 1 ihank Hcikki Salcma;i for imrodLiciiig me to llic nortlicrn walcr'^ ;ind his valuable comments on llic work and the rnanuscript- llppo Viiorinen for commenting on tho iimmiscrlp!. and l-lcnii Gorokliova, University of Stockliolm, for teaching me zooplankton Uixonomy. Kilpisjiirvi biological station provided facilities for the study during summer r

References Allan, .1. D- 197(1. Life history pattcrus in zooplaukton. Am. Nat. IK); 165 180. Anderson. R. S. 1971. Crustacean plankton of 146 alpine and subalpine lakes and ponds in Western Canada. J. Fish- Res. Bd. Can. 2K: 311-321, 1974- Crustacean plankton communities of 340 lakes and ponds in and near the national park of the Canadian Rocky Mountains. - J. Kish. Res- Bd. Can. 31: 855 869. Arnott. S. E. and Vanui. M. J. i993. Zooplankton assemblages in fisliless boa lakes: influence of biotic and abiotic factors. - Hcology 74: 2361 2380. Arvola. L. et al. I9K6. Elfects of experimental aeidification on phyto-, bacteiio- and /ooplankton in enclosures of a highly humic lake. Int. Rev. (ies, Hydiobiol- 71: 737 758. ' Boer. M- M., Koster, li. A- and Lundberg. H- 1990. Greenhouse impact in Fennoscandia preliminary findings of a European workshop on the effects of climatic change. Am bio 19: 2 10. Brctscko, G. 199,^. Opportimities for high alpine research, the lake "Vorderer Finstertalcr See" as an example (Kiihtai. Tirol, 2237 m a.s.l.)Limnologica 25: lOS 108.

333

Brown, J. H. 1981. Two decades of homage to Santa Rosalia: toward a general theory of diversity. Am. Zool. 21: 877-8X8. Carter, J. H. C. 1971. Distribution and abundance of planktonic Crustacea in ponds near Georgian Bay (Ontario, Canada) in relation to hydrography and water chemistry. - Arch. Hydrobiol. 68: 204 231. - et al. 1980. Distribution and zoogeography oi planklonic crustaceans and dipterans in glaciated eastern North America. - Can. ,1. Zool. 58: 1355-1387. Connell J. II. and Orias, H. 1964. The eeologieal regulation of species diversity. - Am. Nat. 98; 399 414. Douglas. M. S. V., Smol, J. P. and Welston, B. Jr. 1994. Marked post-l8th century environmental change in high arctic ecosystems. - Seience 226; 416 419. nussart B 1967. Les Copcpodes des eaux continentales, d'EuVope occidentale. Part 1. Ed. N. - Boubee and Cie. Pederley B 1972. Introduction: the area, its investigation and'Ihe plant cover. - In; Krogerus (ed.). The invertebrate fauna of the Kilpisjarvi area. Pinmsh Lapland. Aeta Soe. Fauna Elora Fenn, 80: 5 36. Fryer, G. 1985. Crustacean diversity in relation to the size ol water bodies: some facts and problems. Freshw. Biol. 15; 347-.^61. ^ , Girdner, S. F. and Larson, G. L. 1995. Effects ot hydrology on zooplankton communities in high-mountain ponds. Mount Rainier National Park, USA. J. Plank. Res. 17: 1731 1755. Hammer, U. T. and Sawchyn, W. W. 1968. Seasonal succession and congeneric association of Diaptomus spp. (Copepoda) in some Saskatchewan ponds. Limnol. Oceanogr. 13; 476-484. Hansson, L.-A., Lindell. M. and Tranvik, L. J. IW.l Biomass distribution among trophic levels in lakes lacking vertebrate predators. Oikos 66; 101 106. Hebert, P. D. N. and Hann, B. J. 1986. Patterns m ihe composition of arctic tundra pond microcrustacean communities. - Can. J, Fish. Aq. Sci. 43; 1416-1425. - and Loaring, J. M. 1980. Selective predation and the species composition of arctic ponds. Can. J. Zool. 58: 422-426. Ilessen, O. and Sorensen. K. 1990. Photoprotective pigmentation in alpine zooplankton populations. - Aq. Fenn. 20; 165 170. Jansson M,, Persson, G. and Broberg, O. 1986. Phosphorus in acidified lakes; The example of Lake C5ardsjon. Sweden. - Hydrobiologia 139: 81 96. . larvinen A 1987. Basic climatological data on the Kilpisjarvi area, NW Finnish Lapland. Kilpisjarvi Notes 10; MacArthur, R. II. and Wilson, E. O. 1967. The theory of island biogeography. Princeton Univ. Press. Maguire, B. 1963. The passive dispersal ot small aquatic organisms and their eolonisation of isolated bodies ol water. - Ecol. Monogr. 33: 161- 185. Marcus, N. 1990. Calanoid copepod. cladoceran and rotiler eggs in sea-bottom sediments of northern Calitomian coastal waters: identitieation, occurrence and hatching. Mar. Biol. 105: 413-418. Nedler K. H. and Pennak, R. W. 1955. Seasonal launal variations in a Colorado alpine pond. - Am. Midi. Nat. 5^; 419 430. .. . Patalas, K. 1964. The crustacean plankton communities in 52 lakes of different altitudinal zones of Northern Colorado. - Verh. Int. Ver. Limnol. 15; 719 726. Pearce, F. 1996. Canadian lakes suffer triple blow. New Sci! 2018; 16. Pennak R W 1958. Regional lake typology in northern Colorado. U.S.A. Verh. Int. Ver. Limnol. 13; 264-283 Pienilz, R. and Smol. J. P. 1993. Dialom assemblages and their relationship to environmental variables in lakes

334

from the boreal forest-tundra eeotone near Yellowknife Territories, Canada. - Hydrobiologia 269/270; 391-404. Pscnner, R. and Schmidt, R. 1992. Climate-driven p\\ con^ trol of remote alpine lakes and effeets of acid deposition. â&#x20AC;˘ Nature 358: 781-783. Raina, H. S. and Vass, K. K. 1993. i:)istribution and speeies composition of zooplankton in Himalayan ecosystems. Int. Revue. Ges. Hydrobiol. 78; 295 307. Ranta. L. 1979. Niche of Daphnia in rock pools. - Arch. Hydrobiol. 87; 205 223. Reed, h. B. 1962. Freshwater plankton crustaeea ol the Conville river area, northern Alaska. Arctic 15; 27 50. RLihliiig A (ed.) 1992. Atmospheric heavy metal deposition in northern Europe in 1990. - Nord 12. Nordic Council of Ministers, Copenhagen. Rylov, V. M. (ed.) 1948. Fauna of U.S.S.R.; Ireshwaier Cyclopoida. Israel I'rogram Sci, Transl. Salonen, K. and Hammar, T. 1986. On the importance ot dissolved organic matter in the nutriotion of zooplankton in some lake waters. - Oecologia 68; 246-253. Sarvala J, and Halsinaho, S. 1990. Crustacean zooplankton of Finnish forest lakes in relation to acidity and other environmental factors. In; Kauppi. P., Anttila, P. and Kenttamies, K. (eds). Acidification in Finland. Springer, pp. 1009-1027. Schindler, D. W. et al. 1985. Long-term ecosystem stress: the effects of years of experimental acidilication on a small lake. - Science 228; 1395 1401. et al 1996. Consequences of climate warming and lake acidification for UV-B penetration in North Amencan boreal lakes. Nature 379; 705-708. Schmitz F H 1959. Seasonal biotic events m two Colorado alpine tundra ponds. Am. Midi. Nat. 61: 424-446. Schwartz, S. S., Hann, B. J. and Hebert, P. D. N. 1983. The feeding ecology o\' Hydra and possible implications in the structure of pond zooplankton communities. Biol. Bull. 164: 136-142. Smol, J. P. et al. 1996. Inierring past climatic changes in Canada using paleolimnologieal techniques. Geosci. Can. 21: 113 117. ^ . Sprules W. G, 1972. Effects of size-selective predation and food competition on the high altitude zooplankton communities. - Ecology 53; 375 386. Tash, J. C. and Armitage. K. B. 1967. Ecology of zooplankion of the cape Thompson area, Alaska. - Ecology 48; 129 139.

"ler Braak, C. J. F. 1986. Canonical correspondenee analysis; a new eigenvector lecniqiie for multivariate direct gradiant analyses. Ecology 67: 1167 1179. 1987-1992. Canoco - a Fortran program lor Canonical C\mimunity Ordination. - Microcomputer Power. New York. Uimonen-Simola, P. and Tolonen, K. 1987. Ellects ot recent acidification on Cladoeera in small clear-water lakes studied by means of sedimentary remains. - Hydrobiologia 145; 343 351. Ward, H. B. and Whipple, G. C. (eds) 1959. Freshwater biology. John Wiley. Weekstrom. J.. Korhola, A. and Blom. T. 1997. The relationship between Diatoms and water temperature in thirty subarctic I ennoscandiau lakes. Arct. Alp. Res. 29; 7^ 92. Wetzel R 1983. Limnology. Saunders. Whiieside M. C. 1974. Chvdorid (Cladoeera) ecology; seasonal palterns and abundance of populations in Elk Lake, Minnesota. - Ecology 55; 538 550. Wiggins G. B., Mackay. R. J. and Smith, I. M. 1980. Evolutionary and ecological strategies of animals in annual temporary pools. - Arch Hydrobiol. Suppl. 58; 97 206.

htOGR.'M'HY

Appendi.x 1. List of species found during the study and their presence-absence distribution in the study ponds. Code Sida cry Diap bra Holo gib Daph Ion Simo vet Ceri quad Scap miic Oplir gra Drep den Acan cur Stre ser Lath rcc Eury lam Aero elon Aero har Dunh era Chvd sph Rh'yn fal Alon alT Alon qua Alon gut AloLi cos Alon rcc Alon nan Alon exe Alon exi Bosm obi I'oly ped Uyth Ion Eudi gra Mixo lac

Taxon name Cladocera Sida vrysialliiKi Diaphaiiosonia brticbyiiniiii Holopedium gibbcriini Dcipbi 1 in IOI /S; isp in a Siiiwcepbahis veiulus Ceriodaphnia quudraiigiila Sctipbdk'beris nuieroiiaUi Opliryo.\us gracilis Drcpanodiri.x tieiiiaiii Acaiitlwk'beris curvirostris Slrehloecrus scrricauduhis Lalhoimra reetirosiris Emycercus laincUatiLs Act operas clongalus A. liarpae Diiiilicrediii cras.sa Cbydoriis spbacricus Rbyiicbotalonii fakani .tIfiiKi a/iiii.s

A. A. .1. A.

i/utnirangNlaris y.uiui!(i cosiata recHingula

Aloiielki lUiiHi

A. e\eisa A. e.yiguii Bosmina oblusiroslris Polyphemus pedicuhis BydioHepbes longinuimis Caiunoidi)

Eucy ser Cycl scu Mega vir Mega gig Acan ver Acan cap Acim era Micr gra Cycl A Cycl B Cycl C Cyel D Cyel E

Eiidiaplomiis '^raciioidcs Mixodiaptoiiius laciniatus Cyclopoida Eiiryclops serrulatus Cyclops scutifer Mc'gcieyelops viridis M. gigas Acaiitlioeyclops veriuilis A. eapillalus A. crassicaudis Mieroeyclops gracilis Unidentified Cyclopoida A Unidentified Cyelopoidia B Unidentified Cyclopoidia C Unidentified Cyclopoidia D Unidentified Cyclopoida E

Poly for Bran p;il

.\iiostraca Polyancmia 1 tin iputa Brambiiu'cta paludosa*

1 0 0 I) 0 1 1 1

II II

1 1 0 1 0 1

0 0

u 11

1 II I) 1 0 (1 ! 1 II 1 I)

1 0 II 1 () 1 1 1 0 1 II 1 1 0 0

II 1 1 1) 1)

II II 0 1

0 0 0 1 1 1 1 II 1 11 1 1 1 0 1 II II II II 0 0 1 1

0 11

1 0 (1 (1 1 1 II II 11 I) 1 0 1 11 1

0 0 1

1 1 0 (1 1 II 1 1 1 1 1 0 0 11 1

0 0 1 (1 tl 11 II II 1 1 0 0 (1 1 0 0

(1 0 11 11

0 0 0 (1

0 1 1 II

11 1 1 II

(1 II II i

1 1

0 0 0 1) 1 0 0 0 II II

1 1 1 1 1 11

1 1 0 1 1 1 1 1 1 1 1 1 1 1 1 11

II 11 0 0 0 0

1 0 1 (1 1 II 1 1)

tl 11 11 1 II II 0 1) 1 0 (1

II 11 1

1 0

1 11

0 11

0 (1

1 0

0 0 1 0

11 11 1 0 0 0 0 1 0 (1 1 1 II

11 11 0 11 1 II 1 0 0 0 (1

0 0 0 0 0 1 1

0 1)

(1 0 1 0 0 0 (1 1 II

1) I) 0 1 (1 0 II II

a u 1 0

0 1 1) 1 0

0 0 0 0 0 0 II 1 II II

0 0

1 0 1 1 0

1 II

II II II

1 1 1 1 1)

t 0 1 1

0 0 0 (1 11 1 0 1 1 0

0 11 1 1

1 1 1 1 0

0 0

1 1 1 0

1 (1 11

I 1 1)

II 1 0 II

0 0 1 (1 (1

(1 0 0

0 (1 1

(1 0 0 11 0 1 1 1

(1 0

(1 0 0 II II

(1 0 0 (1 0 i 0

0 (1 0 0 0

1 1 0

1) 0 1 1 1

1 1 0 1 1 0

11 0 1 0 0 0 (1 0 II II 1 II 1 1 0 0 0 II II 11 1 0 0 1 0 11

1 0

1 II

! 11

II 11

II 1

1 0 0 0 1 1

0 0 1 0 0

(1 1 0

0 (1 II II

1 1 11 0 0 (1 1 1 1 0 1 0 0 0 1 II 1

(1 1 1 1 0 1 0 0 1 II II 11

0 0 11 11 1 1 1 0 0 0 1 1 1 II

0 0

0 0 0 1 1 II 11 0 0

0 0 0

11 11

II II 1)

0 1 0 0 11 11 11 1 1 11 0 0

1 11

0 II

1 11

0 0

1 0

0 0

1 11 11 I) 0 (1 II II

II 1 II 0 0 0 0 II II

0 0 0 11 1 II 11 (1 0 0 0 0 II 1 0 1 0 0 0 II

I) 0 1) 0

(1 0 1 1 1 0

0 0 0 0 II 0 1 (I 0 1 1 0

II 1

II l]

II 1

0 0 0

l] 11 0 0 0 (1 11 II 11 11

1 1 II

0 1

II 0 1 1 (1

II 11 1 0 0 1 1

0 0 I)

I 0 0 11 0 0 0 0 ! 1 II 1 0 0 0 0 II

II 1 0 0 1 1 0 II 1 II 1 0 0 0 0 0 II II 11

0 0 II II II II II 0

0 1 0 II II II II

11 0 0

1 0

0 0

1 I)

0 II 11 (1 0 0

* BranchiiiecHi lycdtuhsa was only found in a fell top pond that was not a study site.

ECOdRM'HV :i

335

ECOGRAPHY 27: 129 /136, 2004

The region effect on mesoscale plant species richness between eastern Asia and eastern North America Robert E. Ricklefs, Hong Qian and Peter S. White

Ricklefs, R. E., Qian, H. and White, P. S. 2004. The region effect on mesoscale plant species richness between eastern Asia and eastern North America. / Ecography 27: 129 /136. The greater number of plant species in temperate eastern Asia compared to eastern North America has been ascribed to both local environment and regional characteristics, but the relative contributions of each have not been resolved. In this analysis, we related species richness of flowering plants in mesoscale floras ( B/104 km2) dominated by temperate forest vegetation to area, elevation, latitude, and several climate variables. When analyses were conducted separately within each region, area and, in eastern Asia, elevation, were the primary determinants of species richness. It appears that the number of species in mesic temperate floras within these regions is largely unrelated to the relatively narrow range of local climate factors associated with these floras. Analysis of covariance of the logarithm of species richness with the logarithm of area (b /0.148) and climate measurements as independent variables revealed a region effect, with species richness in eastern Asia exceeding that in eastern North America by 0.294 log10 units, or a factor of 2.0. Similar regional differences in species richness were apparent in floras compiled from larger areas. Understanding differences in plant species richness between regions requires consideration of regional influences, whose effects should be tested in comparative analyses based on floristic surveys of ecologically characterized small areas. R. E. Ricklefs, (ricklefs@umsl.edu), Dept. Biology, Univ. Missouri-St. Louis, St. Louis, MO 63121-4499,USA. / H. Qian, Res. Coll. Center, Ill. State Museum, 1011 East Ash Street, Springfield, IL 62703, USA. / P. S. White, Dept of Biology, Univ. North Carolina-Chapel Hill, Chapel Hill, NC 27599-3280, USA.

Broad-scale patterns in species richness have been related to variation in the capacity of environments to support coexisting species (local processes) as well as to variation between regions in the balance between proliferation and extinction of species (regional processes). The hypothesis that local processes, including competition and consumer-resource interactions, constrain local species richness has been supported by correlations between richness and climate or other environmental conditions (Wright 1983, Currie and Paquin 1987, Adams and Woodward 1989, Currie 1991, Huston 1993, O’Brien 1993, 1998, Francis and Currie 1998, 2003, O’Brien et al. 2000, Kleidon and Mooney 2000, Badgley and Fox 2000, Whittaker and Field 2000). Local processes are

responsible for assembly of local communities from the larger regional pool (Pa¨rtel et al. 1996, Zobel 1997, Weiher et al. 1998). Differences in local species richness between regions with similar environments have been cited in support of the additional influence of regional factors, including aspects of geography, climate, and history that influence rates of species production and extinction (Orians and Paine 1983, Latham and Ricklefs 1993, Ricklefs and Latham 1993, Ricklefs et al. 1999, Qian and Ricklefs 2000). These hypotheses can be evaluated in a framework of analysis of covariance in which the statistical effect of region on local species richness is tested with environmental conditions included as covariates to account for differences within and

Accepted 19 November 2003 Copyright # ECOGRAPHY 2004 ISSN 0906-7590 ECOGRAPHY 27:2 (2004)

129

between regions in local conditions. Specifically, region effects could be rejected if differences in species richness between regions paralleled differences in the local physical environment independently of regional differences in physiography and history. The temperate floras of eastern Asia (EAS) and eastern North America (ENA) have figured prominently in an ongoing discussion concerning the relative contributions of local and regional factors (Latham and Ricklefs 1993, Francis and Currie 1998, 2003, Ricklefs et al. 1999, Qian and Ricklefs 2000, Whittaker and Field 2000). Much of this discussion has focused on analyses of the species richness of trees. In an early study, Currie and Paquin (1987) found that the number of tree species within 2.58 grid squares (2.58 /58 longitude north of 508N; 0.5 /1 /105 km2) in North America was most strongly related to actual evapotranspiration (AET), a measure of habitat productivity (Rosenzweig 1968). Several other variables, including latitude, longitude, temperature, precipitation, insolation, elevation, and variation in these measures within latitude-longitude grid cells, made significant but less important contributions to variation in species richness. The relationship in North America also estimated tree species richness reasonably well in Europe (Currie and Pacquin 1987) and eastern Asia (Adams and Woodward 1989), lending support to the hypothesis that species richness is determined by local ecological conditions. These results have been supported by further analyses of species richness in trees (Currie 1991) and family richness in flowering plants (Francis and Currie 2003, but see Qian and Ricklefs 2004), as well as similar analyses of the diversity-climate correlation elsewhere (e.g., O’Brien 1998, O’Brien et al. 2000). Latham and Ricklefs (1993) analyzed tree species richness in smaller sampling areas (‘‘mesoscale,’’ 101 / 104 km2) based on local studies or floras for small regions. They found an additional region effect, even when AET was accounted for statistically, indicating that tree species richness in eastern Asia exceeded that of North America. In the present study, we consider the species richness of all angiosperm (flowering) plants in a sample of complete floras from mesoscale areas of 10 to 104 km2 in eastern Asia and eastern North America. These floras each comprise B/0.1% of the total area of each region. Our analysis extends previous work in that complete floras are considered (contrasted with tree floras [Latham and Ricklefs 1993] or EAS-ENA disjunct genera [Qian and Ricklefs 2000]), and both survey area and physical environment are included in the analysis to control for the effects of these factors on species richness.

northeast to southwest, and Korea. Arid regions of western China (Xizang [Tibet], Qinghai, Xinjiang, Gansu, Ningxia, and inner Mongolia provinces or autonomous regions) were excluded. We also excluded the major islands of eastern Asia, such as Japan, Taiwan, and Hainan, to remove the effect of island endemics on plant species richness. Eastern North America includes the eastern half of the United States (east of the western boundaries of Minnesota, Iowa, Missouri, Arkansas, and Louisiana). Most of the areas within the two regions are located in warm temperate climate zones (Mu¨ller 1982, Domro¨s and Peng 1988, Fig. 1) and are forested (Wu 1980, Barbour and Billings 1999). To document floristic data, we searched literature sources including journal articles, floras, checklists, monographs, and atlases pertinent to the floras of eastern Asia and eastern North America. In total, we assembled 124 regional floras (39 in eastern Asia and 85 in eastern North America) having areas B/104 km2. The areas of the selected floras ranged from 10 km2 to 3981 km2 in eastern Asia and from 10 km2 to 9333 km2 in eastern North America. Because we limited the size of floras to B/104 km2, the characteristic dimension of each flora was B/100 km, or about 1 degree of latitude. Thus, these floras were smaller, by one to four orders of

Materials and methods We define eastern Asia to include the eastern temperate and warm temperate portion of China extending from 130

Fig. 1. Map showing the location of floras in eastern Asia and eastern North America. ECOGRAPHY 27:2 (2004)

magnitude, than the latitude-longitude blocks used by Currie (1991) and Francis and Currie (2003). For each flora, we recorded the latitude (8N) of the geographic midpoint, area (km2), highest elevation (m), and the number of species of indigenous angiosperms. We also obtained several climate variables for the latitude-longitude half-degree grid point closest to the geographical midpoint of each flora using data in the International Institute of Applied System Analysis (IIASA) climatic database (Leemans and Cramer 1991). This database provides values for each 0.5 degrees of latitude and longitude interpolated from numerous climate stations worldwide. It has been widely used in ecological and biogeographic studies (e.g., Monserud et al. 1993, Tchebakova et al. 1993, Peng et al. 1995, Shao and Halpin 1995). The climate variables were January temperature (8C), July temperature (8C), May through August (‘‘summer’’) precipitation (mm), and September through April (‘‘winter’’) precipitation (mm). In addition, we included two derived climate indices: actual evapotranspiration (AET, mm) and potential evapotranspiration (PET, mm), which is proportional to the drying power of the environment, primarily a function of temperature. AET and PET were calculated following the approach developed by Cramer and Prentice (Cramer and Prentice 1988, Prentice et al. 1992, Prentice et al. 1993). Climate seasonality, which is correlated with species richness in studies covering a broader range of ecological conditions than ours (e.g., O’Brien 1993, 1998), is incorporated in this study by contrasting measures of precipitation and temperature at different seasons. Derived variables, such as the difference in temperature between January and July and the ratio of summer to winter precipitation, had no statistical effect beyond that of the original variables in our analyses and were not considered further. Several floras were excluded because they were statistical outliers ( /3 SD units [p B/0.001] from the mean of the remaining floras) for one or more climate variables. Specifically, we excluded seven floras in eastern Asia with summer precipitation exceeding 1500 mm or B/200 mm, winter precipitation B/200 mm, or January temperature B/ /108C. We also excluded three floras in eastern North America with summer precipitation exceeding 600 mm. As a result of excluding these floras, the data conformed more closely to the normality assumptions of parametric statistics and the range of environmental conditions matched more closely between the regions. Number of species and area were log10transformed to normalize the distributions of the data and to achieve a linear relationship between the two variables. Summer and winter precipitation were also log10-transformed to normalize their distributions. After trimming the dataset and transforming variables, none of the variables had skewness exceeding an absolute value of 1.3; all absolute values for kurtosis were B/2.0 ECOGRAPHY 27:2 (2004)

and most were B/1.0. The final data set included 32 floras in eastern Asia and 82 floras in eastern North America, for a total of 114 floras (Fig. 1). Analyses of species richness of floras within regions included (i) simple correlations of species richness with each independent variable, (ii) stepwise regressions to identify independent variables making the strongest statistical contributions to variation in species richness within each region, and (iii) analysis of covariance to explore differences in species richness between eastern Asia and eastern North America, with continent as a main effect and the independent variables as covariates. All statistical analyses were carried out using procedures in the Statistical Analysis System (SAS) version 6.12 software (Anon. 1990): correlations (Proc CORR); stepwise regressions (Proc STEPWISE with forward selection); analysis of covariance (ANCOVA, Proc GLM) testing differences between regions when species richness was related to the independent covariates area, latitude, elevation, and climate variables (Legendre and Legendre 1998). Details of these analyses are presented in the Results section.

Results Descriptive statistics Variables are compared between regions in Table 1. Average angiosperm species richness in floras in eastern Asia exceeded that in eastern North America by 0.20 log10 units, or a factor of 1.58. On average, floras were obtained from larger areas in eastern North America than in eastern Asia (3.09 vs 2.26, a difference of 0.83 log10 units, or a factor of 6.8) (Fig. 2). In addition, maximum elevations were higher in eastern Asia (17429/ 636 m) than in eastern North America (7839/459 m). The Asian sites were distributed over a wider and, on average, more southerly latitudinal range, owing to the extension of temperate vegetation far to the south at higher elevations in China. Thus, the most southerly Asian sites were 8.9 degrees further south (23.78N) than the most southerly North American sites (32.68N). Differences in climate between the two regions derive in part from the monsoon weather system of eastern Asia, which results in higher summer (May /Aug) precipitation (714 vs 383 mm) but similar winter (Sep /Apr) precipitation (635 vs 660 mm) in eastern Asia. The ranges of PET and AET were similar in the two regions. Our samples did not include semiarid, typically non-forested areas.

Relationship of species richness to environmental variables within regions When variables were tested individually, only area was strongly correlated with species richness in eastern North 131

0.320 0.719 0.417 0.239 0.073 0.012 0.735 0.027 0.025 0.001 B/0.0001 B/0.0001 B/0.0001 B/0.0001 0.0036 0.25 B/0.0001 0.083 0.094 0.79 3.21 41.60 2025 3.97 8.0 28.6 499 929 1044 1153 0.13 2.44 459 0.67 3.06 1.68 33.1 93.2 71.3 93.4 0.13 2.94 636 0.65 4.13 3.42 182.6 168.7 105.3 97.4 3.11 29.72 1742 2.26 2.75 24.18 714.3 634.7 908.9 925.2 LOGSPP LAT ELEV LOGAREA TEM1 TEM7 PRES PREW AET PET

2.74 23.68 964 1 /7.5 17.0 330 312 642 723

3.41 37.33 3306 3.60 12.3 31.6 1087 926 1094 1107

2.91 38.84 783 3.09 0.64 23.62 383.3 660.1 880.0 930.4

2.61 32.58 27 1 /4.4 19.9 327 567 773 784

52.6 286.0 80.2 35.1 8.9 1.34 310.7 3.06 2.86 0.07

R2 p F1,112 Max Min SD SD Mean Code

Min

Max

Mean

ANOVA Eastern North America (n /82) Eastern Asia (n /32) Variable

America (log-transformed variables, r /0.75, pB/ 0.0001). Species richness was positively related to elevation (r /0.30, pB/0.05) and negatively related (pB/0.05) to summer temperature (r / /0.18), summer precipitation ( /0.32), and annual evapotranspiration ( /0.25). In eastern Asia, area (r /0.49) and elevation (r /0.56) were significantly correlated with species richness. No climate variables had significant simple correlations with species richness in Asia. Stepwise multiple regressions incorporate correlations among the independent variables in their influence on species richness. Stepwise regressions accounted for 56% of the variation in species richness in eastern Asia and 65% in eastern North America. The analyses provided support for the association of species richness with elevation and January temperature in eastern Asia, and with area, January temperature, and summer precipitation (weak and negative) in eastern North America (Table 2).

Angiosperm species richness (log10) Latitude (8N) Elevation (m) Area (log10 km2) January temperature (8C) July temperature (8C) Precipitation in May /August (mm) Precipitation in September /April (mm) Annual actual evapotranspiration (mm) Annual potential evapotranspiration (mm)

The region effect on species richness

Variable

Table 1. Descriptive statistics for angiosperm species richness and site characteristics for floras in temperate eastern Asia and temperate eastern North America.

132

Fig. 2. The relationship between the logarithm of angiosperm species richness (vertical axis) in floras in eastern Asia (solid symbols) and eastern North America (open symbols) and the logarithm of flora area. Regression lines are from the analysis of covariance in Table 3.

To assess the effect of region on species richness, we included region as a main effect in an analysis of covariance. Because of its strong correlation with climate variables, latitude was excluded from the analysis. Flora area and January temperature were the only variables making significant unique contributions to species richness (Table 3). When the analysis was restricted to these covariates, the logarithmic slope of species richness with respect to area was 0.1489/0.013, and species richness increased by 2.4% for each 18C increase in January temperature. Region had a strong effect on species richness, with values in eastern Asia exceeding those in eastern North America by a factor of almost 2.0 (Table 3 and Fig. 2). ECOGRAPHY 27:2 (2004)

Table 2. Analysis of variance of the log-transformed species richness of flowering plants with respect to area, latitude, elevation, and climate variables. Models were reduced by stepwise regression to exclude non-significant variables. Model

Eastern Asia (n /32)

Eastern North America (n /82) 2

Log10 area Elevation (km) January temperature (8) Summer precipitation (m)

F2,29 /18.6, pB/0.0001, R /0.562

F3,78 /47.8, p B/0.0001, R2 /0.648

Estimate

120.5

0.0001

0.164

0.015

31.1 16.2

0.0001 0.0004

0.142 0.0157

0.025 0.0039

18.9 4.5

0.0001 0.0378

0.014 /0.575

0.003 0.272

Estimate

Note: Estimated slope for elevation in meters was multiplied by 1000.

To examine the possibility that region effects may differ among floras having different area, we also analyzed data for seven floras in eastern Asia and 34 floras in eastern North America having areas between 104 and 105 km2, that is, similar to the areas considered by Currie and his colleagues (0.5 /1.0 /105 km2). Species richness in these areas in eastern Asia exceeded that in eastern North America by 0.30 log10 units, or a factor of 2.0. When these floras were subjected to an analysis of covariance, the effect of region was significant (F1,37 /23.9, pB/0.0001), with species richness in eastern Asia exceeding that in eastern North America by 0.2459/0.050, a factor of 1.76. The logarithmic regression of species richness on area had a slope of 0.1749/ 0.056. Among floras having areas between 105 and 106 km2 (EAS n /50, ENA n /70), the slope of the species-area relationship increased to 0.2999/0.028. The more than two-fold excess of species richness in these large floras in eastern Asia was reduced when area, elevation, and January temperature were included as covariates, to a difference of 0.1199/0.019 log10 units, or a factor of 1.32 (F1,115 /40.2, pB/0.0001).

Discussion Our analyses indicated that the roughly six-fold range of variation in angiosperm species richness in forested areas

of both eastern Asia and eastern North America is, in part, related to climate variables, with winter (January) temperature having the greatest effect, and that species richness in eastern Asia exceeds that in eastern North America by a factor of almost two when the influences of area and climate are taken into account. These results for mesoscale ( B/104 km2) floras extend the conclusions of Currie and Paquin (1987), Currie (1991), Whittaker and Field (2000), and Francis and Currie (2003) that climate exerts a dominating influence on species richness on a global scale. Our analyses do not reject the existence of a region effect on species richness that is independent of differences between regions in local climate conditions. The analyses of Currie and Paquin (1987), Currie (1991), and Francis and Currie (2003) applied to larger sampling areas and extended over the whole of North America, including areas with scant precipitation in the western part of the continent. The lower limit of AET in those studies was about 100 mm / too little to support forested habitats / compared to 642 mm in eastern Asia and over 773 mm in eastern North America in this study. The extended range of climate variation including areas with low plant species richness undoubtedly increased the statistical correlation between species richness and climate, reducing the relative importance of the region effect. It also appears that the difference in species richness between regions diminishes at larger

Table 3. Analyses of covariance of the common logarithm of species richness with region as an effect. Effect

Type III SSa

SEb

0.131c 0.017d 0.274e

0.015 0.006 0.062

0.148c 0.010d 0.294e

0.013 0.003 0.021

All covariates included: F9,104 /27.3, pB/0.0001, R /0.702, SS /2.729 75.9 B/0.0001 0.592 Log10 area January temperature 8.1 0.0055 0.063 0.154 Region 19.8 B/0.0001 Remaining variables B/2.2 /0.05 1.108 Error 0.812 2 Only significant covariates included: F3,110 /77.6, pB/0.0001, R /0.679, SS /2.729 Log10 area 122.9 B/0.0001 0.979 January temperature 15.4 0.0002 0.122 Region 189.1 B/0.0001 1.505 Error 0.876 a

SS /sums of squares. bSlope of the regression equation, followed by the standard error of the slope. cLogarithmic regression slope. Units are log10 units per 8C. eLog10 units by which eastern Asia exceeds eastern North America.

ECOGRAPHY 27:2 (2004)

133

( /105 km2) sampling areas. Without a better understanding of how species richness is related to environmental heterogeneity, and how the scale of environmental heterogeneity differs between the regions, it is difficult to interpret differences between the results for larger and smaller areas.

The effect of region on species richness The angiosperm flora of the southern part of North America north of Mexico (an area including almost the entire contiguous USA with 7.4 million km2) has about 14 240 native species of angiosperms while an area of the same size and latitude range in eastern Asia has about 20 000 native species of angiosperms, a factorial difference of about 1.4 (Qian 2002). The continental-scale difference in species richness between eastern Asia and North America is greater in older groups of plants with strong tropical affinities, and it is reduced or even reversed in more recent groups with temperate centers of species richness (Qian and Ricklefs 1999). Moreover, the difference in species richness is more pronounced in taxa restricted to primarily forested habitats in eastern North America, while some taxa distributed in the physiographically more heterogeneous west are more diverse than their Asian relatives (Qian and Ricklefs 2000). In this study, we found that the average species richness of angiosperms in floras ranging from 101 to nearly 104 km2 in area is greater in temperate eastern Asia than in temperate eastern North America. This region effect appears to be maintained, albeit less strongly, in larger floras closer to the size of the latitude-longitude blocks in the analyses by Currie and his collaborators (0.5 /1.0 /105 km2), although climate covariates for a single point might represent climate data less well within the larger areas. Data for smaller areas in eastern Asia are largely lacking. Latham and Ricklefs (1993) found that tree species richness on ca 1-ha plots (10 2 km2) was higher in Japan than in eastern North America, but such comparisons are not yet available more broadly. Lacking information on environmental heterogeneity within flora areas, we cannot comment on the contribution of heterogeneity to the difference in species richness between eastern Asia and eastern North America. Maximum elevations are greater in eastern Asia, which suggests greater topographic heterogeneity. However, when maximum elevation was included in an analysis of covariance among mesoscale floras, the difference in species richness between the two regions persisted. If the regions did not differ in environmental heterogeneity, the greater species richness of small floras in eastern Asia would result from greater plant species richness within habitats at a local scale. This is supported by tree species 134

richness in small plots in Japan, as mentioned above, but data for the richer floras of temperate China are lacking. This study has addressed the relative contributions of local conditions and region effects to local species richness of angiosperm plants. Differences in environment between the regions, including a greater range of elevations within floras and a monsoon pattern of precipitation in eastern Asia, and the lack of floristic data from small, ecologically defined areas, has made this task difficult. We have no doubts that our analysis will be modified as sampling improves and more data become available, but we believe that the difference in species richness between eastern Asia and eastern North America found in this study will remain. Whether the region effect reflects unmeasured environmental differences between eastern Asia and eastern North America cannot be resolved at this point. Ecological conditions within mesoscale areas can be characterized only crudely at present. For example, data are lacking on soils, whose variation can have a strong effect on plant diversity (e.g. Huston 1993). The environmental variables used in most studies, including ours, are average or representative conditions within the area of each flora and do not reflect the heterogeneity of these variables. Nor do most studies include information about the distribution of species within flora areas that would show how mesoscale species richness is constituted from local species richness within habitats and turnover of species between habitats within floras. A complete resolution of patterns of species richness between eastern Asia and eastern North America will require planned comparisons among carefully matched, environmentally characterized plots of 1 /10 ha, nested within larger sample areas, to assess both local species richness and turnover of species along the same ecological gradients of conditions. Finally, the role of history and physiographic heterogeneity within regions in generating and maintaining the regional species pool must be taken into account (Qian and Ricklefs 2000). The difference in species richness in local floras between eastern Asia and eastern North America might have resulted from some combination of regional factors, including a higher rate of invasion of plant lineages in Asia from more tropical regions to the south, a higher rate of production of new species within temperate forest regions of eastern Asia (Qian and Ricklefs 2000), and less extinction in eastern Asia due to late Tertiary climate cooling and glaciation. Addressing these issues will require new data. Many elements of the temperate flora of eastern Asia have close relatives in the tropics (Latham and Ricklefs 1993), where angiosperm plants are thought by many authors (e.g., Takhtajan 1969, Wu 1980, Lidgard and Crane 1990) to have originated and diversified. The invasion of comparatively harsh temperate environments, particularly the requirements of frost tolerance and a long dormancy period, may depend upon physioECOGRAPHY 27:2 (2004)

logical adaptation that slows the invasion of temperate regions by tropical lineages (Farrell et al. 1992, Latham and Ricklefs 1993). If this were the case, transitions between tropical and temperate latitudes might have occurred more frequently in eastern Asia than in eastern North America, where there is no direct connection between tropical and temperate floras and where temperate floras harbor fewer lineages with close affinities to the tropics (Latham and Ricklefs 1993). The latitudinal range of temperate habitats in eastern Asia also exceeds that in eastern North America, primarily because mesic temperate regions in North America are bounded by the Gulf of Mexico to the south. North America also lacks high mountains in southern latitudes comparable to the Himalayas and other ranges of southern China. The mountains of southern China are an area of extremely high plant species richness (Barthlott et al. 1996). Furthermore, cooling climates of the late Tertiary and glacial climates of the Pleistocene may have had less influence on the floras of eastern Asia, in spite of the fact that tundra environments moved southward to 43 /448 N (northeastern China) during glacial maxima. Until comparative studies include hypotheses incorporating regional as well as local influences on species richness, it is unlikely that such studies will be able to unravel the relationships between species richness, environment, and history. Acknowledgements / We thank David Currie, Curtis Flather, Michael Huston and Jeremy Lichstein for comments on the manuscript. We thank James Beck, Jie Chang, Lisa DeCesare, Jingyun Fang, W. John Hayden, Jinsheng He, Lynn Heilman, Jianhui Huang, Walter S. Judd, Mingchun Luo, Ann F. Rhoads, Roger Sanders, Xunlin Yu, and Jintun Zhang for their help in data collection. We thank Wolfgang Cramer, Changhui Peng, and Guofan Shao for making the climate data available.

References Anon. 1990. SAS/STAT user’s guide. Ver. 6, 4th ed. SAS Institute, Cary, N.C., USA. Adams, J. M. and Woodward, F. I. 1989. Patterns in tree species richness as a test of the glacial extinction hypothesis. / Nature 339: 699 /701. Badgley, C. and Fox, D. L. 2000. Ecological biogeography of North American mammals: species density and ecological structure in relation to environmental gradients. / J. Biogeogr. 27: 1437 /1467. Barbour, M. G. and Billings, W. D. 1999. North American terrestrial vegetation. 2nd ed. / Cambridge Univ. Press. Barthlott, W., Lauer, W. and Placke, A. 1996. Global distribution of species diversity in vascular plants: towards a world map of phytodiversity. / Erdkunde 50: 317 /326. Cramer, W. P. and Prentice, I. C. 1988. Simulation of regional soil moisture on a European scale. / Norsk Geografisk Tidsskrift 42: 149 /151. Currie, D. J. 1991. Energy and large-scale patterns of animal species and plant species richness. / Am. Nat. 137: 27 /49. Currie, D. J. and Paquin, V. 1987. Large-scale biogeographical patterns of species richness of trees. / Nature 329: 326 /327. Domro¨s, M. and Peng, G.-B. 1988. The climate of China. / Spinger. ECOGRAPHY 27:2 (2004)

Farrell, B., Mitter, C. and Futuyma, D. J. 1992. Diversification at the insect-plant interface. / BioScience 42: 34 / 42. Francis, A. P. and Currie, D. J. 1998. Global patterns of tree species richness in moist forests: another look. / Oikos 81: 598 /602. Francis, A. P. and Currie, D. J. 2003. A globally consistent richness-climate relationship for angiosperms. / Am. Nat. 161: 523 /536. Huston, M. 1993. Biological diversity, soils, and economics. / Science 262: 1676 /1680. Kleidon, A. and Mooney, H. A. 2000. A global distribution of biodiversity inferred from climatic constraints: results from a process-based modelling study. / Global Change Biol. 6: 507 /523. Latham, R. E. and Ricklefs, R. E. 1993. Global patterns of tree species richness in moist forests: energy-diversity theory does not account for variation in species richness. / Oikos 67: 325 /333. Leemans, R. and Cramer, W. P. 1991. The IIASA database for mean monthly values of temperature, precipitation and cloudiness on a global terrestrial grid. / Res. Rep. RR-9118. Intl. Inst. Appl. Syst. Anal., Laxenburg, Austria. Legendre, P. and Legendre, L. 1998. Numerical ecology. 2nd ed. / Elsevier. Lidgard, S. and Crane, P. R. 1990. Angiosperm diversification and Cretaceous floristic trends: a comparison of palynofloras and leaf macrofloras. / Paleobiology 16: 77 /93. Monserud, R. A., Denissenko, O. V. and Tchebakova, N. M. 1993. Comparison of Siberian paleovegetation to current and future vegetation under climate change. / Climate Res. 3: 43 /159. Mu¨ller, M. J. 1982. Selected climatic data for a global set of standard stations for vegetation science. / Dr. W. Junk Publ. O’Brien, E. M. 1993. Climatic gradients in woody plant species richness / towards an explanation based on an analysis of southern Africa’s woody flora. / J. Biogeogr. 20: 181 /198. O’Brien, E. M. 1998. Water-energy dynamics, climate, and prediction of woody plant species richness / an interim general model. / J. Biogeogr. 25: 379 /398. O’Brien, E. M., Field, R. and Whittaker, R. J. 2000. Climatic gradients in woody plant (tree and shrub) diversity: waterenergy dynamics, residual variation, and topography. / Oikos 89: 588 /600. Orians, G. H. and Paine., R. T. 1983. Convergent evolution at the community level. / In: Futuyma, D. J. and Slatkin, M. (eds), Coevolution. Sinauer, pp. 431 /458. Pa¨rtel, M. et al. 1996. The species pool and its relation to species richness: evidence from Estonian plant communities. / Oikos 75: 111 /117. Peng, C.-H., Guiot, J. and van Campo, E. 1995. Reconstruction of past terrestrial carbon storage in the northern hemisphere from the Osnabru¨ck biosphere model and palaeodata. / Climate Res. 5: 107 /118. Prentice, I. C. et al. 1992. A global biome model based on plant physiology and dominance, soil properties and climate. / J. Biogeogr. 19: 117 /134. Prentice, I. C., Sykes, M. T. and Cramer, W. 1993. A simulation model for the transient effects of climate change on forest landscapes. / Ecol. Modell. 65: 51 /70. Qian, H. 2002. A comparison of the taxonomic richness of temperate plants in East Asia and North America. / Am. J. Bot. 89: 1818 /1825. Qian, H. and Ricklefs, R. E. 1999. A comparison of vascular plant taxonomic richness in China and the United States. / Am. Nat. 154: 160 /181. Qian, H. and Ricklefs, R. E. 2000. Large-scale processes and the Asian bias in temperate plant species diversity. / Nature 407: 180 /182. Qian, H. and Ricklefs, R. E. 2004. Taxon richness and climate in angiosperms: is there a globally consistent relationship that precludes region effects? / Am. Nat. 163.

135

Ricklefs, R. E. and Latham, R. E. 1993. Global patterns in diversity in mangrove floras. / In: Ricklefs, R. E. and Schluter, D. (eds), Species diversity in ecological communities. Univ. Chicago Press, pp. 215 /229. Ricklefs, R. E., Latham, R. E. and Qian, H. 1999. Global patterns of tree species richness in moist forests: distinguishing ecological influences and historical contingency. / Oikos 86: 369 /373. Rosenzweig, M. L. 1968. Net primary productivity of terrestrial communities: prediction from climatological data. / Am. Nat. 102: 67 /74. Shao, G. and Halpin, P. N. 1995. Climatic controls of eastern North American coastal tree and shrub distributions. / J. Biogeogr. 22: 1083 /1089. Takhtajan, A. 1969. Flowering plants: origin and dispersal. / Smithsonian Inst. Press.

136

Tchebakova, N. M. et al. 1993. A global vegetation model based on the climatological approach of Budyko. / J. Biogeogr. 20: 129 /144. Weiher, E., Clarke, G. D. P. and Keddy, P. A. 1998. Community assembly rules, morphological dispersion, and the coexistence of plant species. / Oikos. 81: 309 /322. Whittaker, R. J. and Field, R. 2000. Tree species richness modeling: an approach of global applicability? / Oikos 89: 399 /402. Wright, D. H. 1983. Species-energy theory: an extension of species-area theory. / Oikos 41: 496 /506. Wu, Z.-Y. 1980. The vegetation of China. / Science Press, Beijing China. Zobel, M. 1997. The relative role of species pools in determining plant species richness: an alternative explanation of species coexistence. / Trends Ecol. Evol. 12: 266 /269.

ECOGRAPHY 27:2 (2004)

Ecography 30: 440 448, 2007 doi: 10.1111/J.2007.0906-7590.04954.x Copyright # Ecography 2007, ISSN 0906-7590 Subject Editor: Nathan Sanders. Accepted 11 April 2007

An indirect area effect on elevational species richness patterns Tom S. Romdal and John-Arvid Grytnes T. S. Romdal (tsromdal@bi.ku.dk), Center for Macroecology, Biological Inst., Univ. of Copenhagen, Universitetsparken 15, DK-2100 Copenhagen, Denmark. J.-A. Grytnes, Dept of Biology, Univ. of Bergen, Allegaten 41, NO-5007 Bergen, Norway.

The study of elevational diversity patterns, and the attempt to disentangle the factors that create them, has proved a challenging and controversial research venue over the last few decades (Terborgh 1977, 1985, Stevens 1992, Rahbek 1995, 1997, 2005, KoÂ¨rner 2000, Brown 2001, Lomolino 2001, McCain 2005). Studies of elevational gradients fall into two general categories, regional studies that summarize known distributions from complete elevational bands (zones) within a defined geographical or political region, and local-scale transect studies that are based on a single survey scheme (Rahbek 1995, 2005). Elevational band area (the compiled area of each elevational zone) of a mountainous region is generally expected to decline with altitude (MacArthur 1972, KoÂ¨rner 2000, Lomolino 2001), although the smallest area per elevational zone can also be found at mid-elevation, where slope is sometimes steeper (Rahbek 1997). That area of regional elevational bands can influence the number of species found in each band was first clearly demonstrated by Rahbek (1997), and several studies since then have explicitly investigated this regional area effect on elevational diversity gradients (Sanders 2002, Sanders et al. 2003, Bachman et al. 2004). In recent years there has been a renewed interest in the analysis of standardized species richness data sets from transects in order to evaluate simultaneously the potential drivers of elevational diversity gradients (Md. Nor 2001, Sanchez-Cordero 2001, Grytnes 2003, Kattan and Franco 2004, Herzog et al. 2005). A number of factors such as climate-derived productivity, source-sink dynamics and the mid-domain effect, have been evaluated, and the diversity of results confirm the belief that no single factor is responsible for all observed richness patterns (Rahbek 1995, 2005, Brown 2001, Lomolino 2001, McCain 2006, Dunn et al. 2007).

440

Most transect studies attempt to control the influence of area by sampling equal-area plots, although unstandardized plots also frequently occur (Lieberman et al. 1996, Hoffmann et al. 2001, Herzog et al. 2005). However, in addition to the direct influence of the area of the sampled plot, local communities can be perceived as dynamic samples drawn from a regional species pool (Terborgh 1973, Graves and Gotelli 1983, Cornell 1985, Ricklefs 1987). The size of the surrounding region can be thought of as a surrogate variable for the size of the regional species pool (Terborgh and Faaborg 1980, Cornell and Lawton 1992, Caley and Schluter 1997, Cornell 1999). This additional effect of area on elevational transects was, to our knowledge, first discussed by Beehler (1981), who found that a linear decrease in New Guinean forest birds on an elevational transect was paralleled by a decrease in regional area. Several other authors have suggested that regional habitat area influenced observed elevational richness patterns (Holloway 1987, Stotz 1998, Patterson et al. 1998, Brehm et al. 2003), while Heaney (2001) and Sanders et al. (2003) presented results contrasting with the hypothesis. Some recent studies have attempted to analyse this area effect quantitatively. Grytnes (2003) found that elevational band area was only correlated with site richness in two out of seven elevational gradients of plant richness in Norway. In contrast, a considerable influence from measures of surrounding habitat area has recently been demonstrated on a tropical transect measuring avian richness (Herzog et al. 2005). We will use the term indirect area effect to describe the effect of regional area on the pattern of local species richness. In an attempt to quantitatively evaluate the importance of the indirect area effect we have performed a meta-analysis on data from 71 elevational

richness patterns using all available studies of single transect gradients found in the literature.

The hypothesis of an indirect area effect We use the term local to designate the scale at which sampling is done at one site along an elevational transect, and regional to apply to the entire elevational band or zone (Fig. 1). We will explore the complex relationships linking regional area (AR), regional species richness (SR), local area (AL), and local species richness (SL) for elevational gradients by building on two premises. First, the number of species in an elevational band (SR) is influenced by band area (AR) through the Arrhenius equation (Arrhenius 1921, Connor and McCoy 1979). Second, the number of species in a plot (SL) is correlated with the regional richness (SR). The latter relation can be linear or nonlinear (Graves and Rahbek 2005, He et al. 2005), and has been described mathematically in several versions (Rosenzweig and Ziv 1999, He et al. 2005). Combined, these two well known effects predict a linear relation between log-area of regional bands (AR) and logrichness of plots (SL) (Rosenzweig and Ziv 1999). We therefore use the following working hypothesis to test for what we will term the indirect area effect for local-scale elevational richness patterns: the number of species in equal-area plots SL on elevational transects is correlated with the area of the surrounding elevational zone AR, such that logSL z logA R c

(1)

where z is the slope on log-log scale, and c is a constant. This equation is identical to the Arrhenius equation for a â&#x20AC;&#x2DC;â&#x20AC;&#x2DC;directâ&#x20AC;&#x2122;â&#x20AC;&#x2122; area effect because we merely combine the Arrhenius equation with a linear relation.

Meta-analysis on forest transect data Our aim is to quantify the strength of the relationship between local species richness and regional habitat area, and we are as such not interested in exact parameterizations of z and c in eq. 1. To measure the strength of the AR-SL-relation we used the Pearson correlation coefficients relating observed species richness in samples to the forest area of elevational zones. Data are from a complete survey of literature studies (Rahbek 2005), performed on the ISI Web of Science 18 April 2005 using the following search terms (including variants): (elevation OR altitude) AND (richness OR diversity). This initial search produced 1650 papers, of which only a small minority contained relevant data. For inclusion in the database we considered only studies containing point diversity data from single forested elevational transects. The full list can be seen in the Appendix. A study was included in our analysis only if plot area and sample effort (e.g. number of trap nights or transect walks) varied at most by a factor 1.5, the transect contained a minimum of four sample sites and constituted an elevational gradient of at least 800 m. Furthermore, richness data should be readily available in the paper, and the locality described with enough precision for us to place it correctly within our forest area dataset. We restricted the analysis to forested gradients to avoid several biases. Forest extent is relatively easily defined, and forest species are often specialised to that habitat. Combining studies from different habitats is problematic, because the degree of habitat structuring greatly influences species-area relations (Drakare et al. 2006). In some cases we discarded single sites that were from other habitats, as detailed in the Appendix. We excluded studies that only analysed a subset of the taxon group in question (e.g. restrictedrange species), as well as forest disturbance studies. Our

Fig. 1. The concept of the indirect area effect, as shown for a hypothetical mountain range. At each of two elevations, a standardised plot is sampled for species richness (white rectangles). Two elevational bands (horisontal lines) are outlined. The forest habitat (shaded) of each band holds the potential species pool for the sampled plot. The contiguous area of habitat differs between bands resulting in different number of species in the species pool for each plot. This in turn influences the number of species in the plots.

441

criteria allowed us to include 71 out of 95 local-scale forest studies. The data on forest area were extracted from the MODIS 44b Global 500 m ISIN Grid data set, available from Bhttp://edcdaac.usgs.gov/modis/mod44b.asp . This global dataset was compiled from satellite data from the years 2001 and 2002. The information used was percentage of forest canopy cover in 500 500 m surface area cells based on a integerized sinusoidal projection. Two criteria had to be applied in order to estimate the forest areas (Appendix 1). First, forest was as our default defined as habitat having 50% canopy cover. Second, to delimit individual forest units, we used a criterion of including all forest cells that were connected by direct contact with neighbouring forest gridcells. We made a range of sensitivity tests for other values than the defaults (Appendix 2). The overall results remained very robust to changing these criteria. Forest area in each elevational band was derived by combining the data on forest distribution with elevational data from the Shuttle Radar Topography Mission, available at Bhttp://glcf.umiacs.umd.edu/data/ srtm . By overlaying these two data sets we were able to identify the elevation of all forested gridcells. From this we estimated contiguous forest area of 200 m wide elevational bands centered around each empirical species richness sample point. We performed a standardized and weighted metaanalysis of the relation between regional forest area and local species richness (Gurevitch and Hedges 1999, Gurevitch et al. 2001). The effect size variable z is a transformation of the correlation coefficient r, where 1 1 r z ln 2 1 r Values of z are expected to approximate a normal distribution. The variance of z is measured as vz

1 N 3

We then calculate a weighted average effect size ZÂŻ as n X 1

zi i 1 v i ÂŻ Z n X 1 i 1

with the variance of ZÂŻ being s2

1 n X

i 1

and the confidence intervals ÂŻ a=2(n 1) (s) CI Z9t (Gurevitch et al. 2001). The z-values are backtransformed to r-values before presentation in tables.

A measurable effect Of the 71 transects included, 32 were on plants, 27 were on vertebrates and 12 on invertebrates. A general pattern of species richness is not evident. For instance, there are two studies from Mt. Isarog on Luzon, one concerning ants (Samson et al. 1997) and the other mammals (Heaney 2001), which show almost opposite patterns in elevational diversity. Furthermore, the individual studies showed a wide variety of regional area vs local species richness relationships from highly positive correlations to negative correlations. In most studies there was a considerable positive correlation between forest area and local species richness, and averaging values from both positive and negative correlations yields a positive average correlation coefficient of 0.63 (Table 1). Figure 2 shows some examples of both hump-shaped and monotonically decreasing patterns of area and species richness. Patterns of decreasing species richness are mostly very well

Table 1. Correlation coefficients of the relation between surrounding habitat area and species richness, overall and for taxonomic subgroups. Group Overall Invertebrates Birds Non-volant mammals Bats Higher plants Cryptogames

Number of studies

Lower 95% limit

Upper 95% limit

71 12 13 11 2 27 5

0.631 0.836 0.903 0.294 0.847 0.588 0.233

0.565 0.673 0.826 0.092 0.592 0.485 0.094

0.689 0.921 0.947 0.603 0.996 0.674 0.515

Herptiles (1 data set) not shown. All tests performed with the default forest definition. Averages and confidence limits were retransformed from z to r-values following analysis.

442

30 1000 20 500

Area (#cellsx1000)

40 1500

Species richness

2000

Area (#cells)

c 250

500

1000

1500

200

25 150

100

15 10

0 2000

0 0

r = 0.76

d 500

r = -0.59

160

140

20 15

Species richness

Area (#cellsx1000)

Area (#cells)

Elevation (sites)

Elevation (m)

b 80

r = 0.94

400

120 100

300

80 200

60 40

100

0 0

500

1000

Species richness

r = 0.99

1500

0 2000

Elevation (m)

Species richness

a 2500

0 0

Elevation (sites)

Fig. 2. Examples of the relation between elevational band area and local species richness in individual studies used for the metaanalysis. In these figures actual number of species and actual area (gridcells) is shown, while in correlation analysis logtransformed values were used. Open diamonds forest area. Closed squares species richness. r Pearson correlation coefficient. (a) Bird species in Andohahela, Madagascar (Hawkins 1999). (b) Mites in Nebrodi Mts, Sicily (Migliorini and Bernini 1999). (c) Birds on Whiteface Mt., New York (Able and Noon 1976). (d) Plants in Santa Catalina Mts, Arizona (Whittaker and Niering 1975).

predicted by the area effect, whereas for hump-shaped patterns our results are more variable. Many meta-analyses are haunted by a compilation of test values that are not obtained through identical test procedures among studies (Gurevitch et al. 2001). This is probably not a problem in our case because we obtained the values for the predictor variable, area, from an independent and standardized source, and because we limited our survey to standardized point richness data sets. One problem we do have is that some of the transects used in our analysis come from the same publications and could be said to be pseudo-replicates because they represent adjacent transects or related taxa within a single study. We made a series of sensitivity analyses to investigate these potential non-independence problems, by excluding the concerned studies. The main bias encountered was that the results for higher plants are disproportionately driven by two studies (Kessler 2000b, 2001), but our overall conclusions are not influenced.

Additional analyses We posit a number of testable secondary hypotheses that offer further insight into the indirect area effect and its relation to other biogeographical patterns. 1) The indirect area effect should be stronger for larger spatial grain sizes than for smaller spatial grains In a tiny plot the number of species is more constrained by the low number of individuals than by the number of species in the species pool (Connor and McCoy 1979, Rosenzweig 1995). Studies with small grain size will therefore tend to exhibit less variation in number of species among sites. This bias would apply to all hypothetical determinants, including indirect area effects. Accordingly, we found a positive linear relationship between z (strength of the indirect area effect) and

443

a 3

z value

-1

-2 5

log (grain size)

b 3

z value

-1

-2

10 11 12 log(forest area)

Fig. 3. The relationship between z (the transformed correlation coefficients) and log-transformed grain size (km2) (a) and logtransformed total size area of forest (500 500 m pixels) (b). A linear relationship using weighted least square was tested. For grain size a significant linear relationship with z is found (p 0.0031; z 1.14 0.076 logGrain size). No linear relationship between z and total area is found (p 0.77).

Table 2. Correlation coefficients for the climate zone and dispersal subgroups. Categorya Tropical Subtropical Temperate High dispersal Low dispersal

Number of studies

Lower 95% limit

Upper 95% limit

44 9 18 28 43

0.630 0.565 0.680 0.724 0.567

0.555 0.241 0.475 0.630 0.473

0.695 0.775 0.815 0.800 0.649

a Subtropical category includes studies from Mexico and Taiwan. High dispersal ability group includes only volant animals and cryptogames (spores). All tests performed with default forest definition. Averages and confidence limits were retransformed from z to r-values following analysis.

444

the grain size used in the study, i.e. less influence for the smallest grains (Fig. 3a).

hypothesis, but the result was statistically insignificant (Table 2).

2) The indirect area effect should be stronger for smaller forests than for larger forests

5) The effect will differ between different taxonomic groups

We sum the area for contiguous forests under the assumption that the entire forest represents the immediate species pool for the sampled sites, even for forest areas as large as the Amazon region. Such immense forest areas will have relatively large regional area measured for all elevational bands, meaning that all bands will have large potential species pools, and the variation in area among bands should be less important. The data did not show any support for this hypothesis when testing for a linear relationship between forest size and z (Fig. 3b). However, a visual inspection of the graph indicates that both for the smallest and largest forest sizes there is less effect, and an ad hoc second order polynomial test confirms this (unimodal pattern, p B0.01). Thus for forests 2000 km2 the pattern is qualitatively similar to what is predicted, but the smallest forests do not show a high indirect area effect. We tentatively suggest that some of the small forests from which we have data (Mt. Kenya, Mt. Elgon, Mt. Isarog on Luzon, Mt. Vermion in Greece, Haleakala on Hawaii, Mexican forests) have been reduced in elevational extent, or even isolated as forest fragments, by human activities in the twentieth century or earlier. The elevational patterns of species richness therefore do not reflect the current habitat area.

Certain taxon groups are under particular habitat restrictions which should predictably be more influential than other environmental factors. Since it is against the concept of meta-analysis to exclude particular taxa (Gurevitch et al. 2001, Mittelbach et al. 2003), we instead evaluated the variation within our results by analysing taxon groups separately. Invertebrates, birds and bats (containing between them all the volant animals) are the taxon groups that have the highest r values, all above 0.8 (Table 1). Non-volant mammals and cryptogam organisms show relatively small effects of indirect area, in both cases and for all individual studies the peak of diversity lies at higher elevations than predicted from the area factor. For cryptogams, basic physiological restrictions normally necessitate species to occupy humid conditions, and the most humid elevations are typically above 1000 1500 m (Kessler 2000a, Kuper et al. 2004, Cardelus et al. 2006). Non-volant mammals have been speculated to share this climatic predisposition (McCain 2005), and small mammals have been speculated to be less successful at low elevations where ants are abundant competitors, based on patterns observed on tropical Asian transects (Samson et al. 1997, Heaney 2001, Md. Nor 2001).

3) High latitude data sets should show a stronger effect than low latitude data sets

Conclusions and implications

Species-area relations have shown to be weaker in areas with high available energy, as high number of individuals results in high species richness at all scales (Storch et al. 2005). Accordingly, we expect less influence from variation of area among bands in tropical than in temperate elevational data sets. We did not find such a trend, as studies from tropics, subtropics and temperate regions showed equal effect (Table 2). 4) The magnitude of the indirect area effect should be correlated with dispersal ability Taxa that have potential for long distance dispersal should be less influenced by the area of isolated forests, since they will be more able to cross gaps. With a crude distinction between high-dispersal organisms (volant animals and cryptogams) and low-dispersal organisms (all other), we found a trend in accordance with this

The indirect area effect has considerable potential as a basic influence on elevational diversity gradients. Although there is considerable variation among studies, the overall trend is a high correlation between surrounding forest area and the number of species found in sites. Elevational diversity patterns that are monotonically decreasing or nearly so (as in Fig. 2a, c) are usually well predicted by the hypothesis, as are some gradients with mid-domain peaks, especially for forest areas of small size (as in Fig. 2b). The cases where we see the least explanatory power are for some of the largest forest units we have defined, which are the entire Neotropical and Nearctic (eastern and western) forest masses. For those forest units the lowlands are by far the elevation with largest area, but elevational richness gradients in associated mountain ranges can often show pronounced midelevational peaks (as in Fig. 2d). In such cases our approach of defining the relevant regional area may not be the optimal approach.

445

In this study we have correlated area with diversity in a variety of taxa, while other studies have explored mathematically how local species richness levels can be predicted from the size of regional richness pools (Rosenzweig and Ziv 1999) or from the regional area (He et al. 2005). Constructing a regional species pool is problematic and requires a precise knowledge of distributions for all species, both regarding elevational and geographical limits. Because of variation in isolation and speciation events, high elevations have the largest turnover of allopatric species within a region, but such species are not all potential colonizers (Graves 1985, 1988, Brehm et al. 2003). Predicting diversity levels from species pools mathematically requires determination of z values of species-area relations (Rosenzweig and Ziv 1999, He et al. 2005). Such exact parametrisation should rest on extensive knowledge of the structural complexity and heterogeneity of the habitat type, since, even among different forest types differences occur (Drakare et al. 2006). Nevertheless, attempts to explore the exact relation of regional and local richness levels with predictive models for specific case studies could further advance this field. Together with the fact that the highest correlation was found under the most strict forest definition criterion (Appendix) we suggest that the indirect area effect has the largest influence on elevational diversity gradients at small regional scales rather than at continental scales. The fact that patterns for some gradients, and some organisms, could not be predicted by the hypothesis, also points to the importance of other factors. Among these are the Mid-domain effect (MDE, Colwell and Hurtt 1994, Colwell et al. 2004) and species-energy hypotheses (Evans et al. 2005). MDE has been promoted as a potentially strong explanation for mid-elevational species richness peaks along some gradients (Rahbek 1997, Grytnes and Vetaas 2002, Grytnes 2003, Dunn et al. 2007). However, a recent meta-analysis showed only a weak correlation between MDE and emperical patterns for elevational gradients, and it was suggested that MDE is more important on larger scales than on smaller spatial scales (Dunn et al. 2007). Energy and climate correlates are generally considered to predict monotonically decreasing patterns (Rahbek 1995), but climate can predict other elevational diversity patterns as well, depending on the interaction between energy availability and water availability (McCain 2006). The fact that many elevational patterns are neither monotonically decreasing nor mid-domain peaked, but skewed or hump-shaped (Rahbek 1995, 2005), indicates that several factors may work in concert to determine patterns in species richness. The results from our meta-analyses suggest that area within an elevational band may have a strong effect on species richness patterns along elevational gradients even when a

446

constant plot size and standardized sampling is used among sampling points. Future study designs should concentrate on determining the optimal scale for measuring the indirect area effect and to compare its effects on elevational diversity gradients relative to other potential drivers.

Acknowledgements TSR was supported by the Danish National Research Council, grant no. 104.Dan.8.f. JAG was supported financially by the Norwegian Research Council. Mads. O. Rasmussen extracted and handled the GIS data. Carsten Rahbek is thanked for discussions and comments on the entire project. Jon FjeldsaË&#x161;, JĂ¸rn L. Larsen and Niels Krabbe shared field experience in aid of forest delimitation. Nathan Sanders, Rob Dunn and Robert Colwell commented on previous versions of the manuscript.

References Able, K. P. and Noon, B. R. 1976. Avian community structure along elevational gradients in northeastern United-States. Oecologia 26: 275 294. Arrhenius, O. 1921. Species and area. J. Ecol. 9: 95 99. Bachman, S. et al. 2004. Elevational gradients, area and tropical island diversity: an example from the palms of New Guinea. Ecography 27: 299 310. Beehler, B. 1981. Ecological structuring of forest bird communities in New Guinea. Monogr. Biol. 42: 837 861. Brehm, G. et al. 2003. Unique elevational diversity patterns of geometrid moths in an Andean montane rainforest. Ecography 26: 456 466. Brown, J. H. 2001. Mammals on mountainsides: elevational patterns of diversity. Global Ecol. Biogeogr. 10: 101 109. Caley, M. J. and Schluter, D. 1997. The relationship between local and regional diversity. Ecology 78: 70 80. Cardelus, C. L. et al. 2006. Vascular epiphyte distribution patterns: explaining the mid-elevation richness peak. J. Ecol. 94: 144 156. Colwell, R. K. and Hurtt, G. C. 1994. Nonbiological gradients in species richness and a spurious Rapoport effect. Am. Nat. 144: 570 595. Colwell, R. K. et al. 2004. The mid-domain effect and species richness patterns: what have we learned so far? Am. Nat 163: E1 E23. Connor, E. F. and McCoy, E. D. 1979. Statistics and biology of the species-area relationship. Am. Nat. 113: 791 833. Cornell, H. V. 1985. Local and regional richness of Cynipine gall wasps on California oaks. Ecology 66: 1247 1260. Cornell, H. V. 1999. Unsaturation and regional influences on species richness in ecological communities: a review of the evidence. Ecoscience 6: 303 315. Cornell, H. V. and Lawton, J. H. 1992. Species interactions, local and regional processes, and limits to the richness of ecological communities a theoretical perspective. J. Anim. Ecol. 61: 1 12.

Drakare, S. et al. 2006. The imprint of geographical, evolutionary and ecological context on species-area relationships. Ecol. Lett. 9: 215 227. Dunn, R. R. et al. 2007. When does diversity fit null model predictions? Scale and range size mediate the mid-domain effect. Global Ecol. Biogeogr. DOI: 10.1111/j.14668238.2006.00284.x. Evans, K. L. et al. 2005. Species-energy relationships at the macroecological scale: a review of the mechanisms. Biol. Rev. 80: 1 25. Graves, G. R. 1985. Elevational correlates of speciation and intraspecific geographic-variation in plumage in Andean forest birds. Auk 102: 556 579. Graves, G. R. 1988. Linearity of geographic range and its possible effect on the population-structure of Andean birds. Auk 105: 47 52. Graves, G. R. and Gotelli, N. J. 1983. Neotropical landbridge avifaunas-new approaches to null hypotheses in biogeography. Oikos 41: 322 333. Graves, G. R. and Rahbek, C. 2005. Source pool geometry and the assembly of continental avifaunas. Proc. Nat. Acad. Sci. USA 102: 7871 7876. Grytnes, J. A. 2003. Species-richness patterns of vascular plants along seven altitudinal transects in Norway. Ecography 26: 291 300. Grytnes, J. A. and Vetaas, O. R. 2002. Species richness and altitude: a comparison between null models and interpolated plant species richness along the Himalayan altitudinal gradient, Nepal. Am. Nat. 159: 294 304. Gurevitch, J. and Hedges, L. V. 1999. Statistical issues in ecological meta-analyses. Ecology 80: 1142 1149. Gurevitch, J. et al. 2001. Meta-analysis in ecology. Adv. Ecol. Res. 32: 199 247. Hawkins, A. F. A. 1999. Altitudinal and latitudinal distribution of east Malagasy forest bird communities. J. Biogeogr. 26: 447 458. He, F. L. et al. 2005. The local-regional relationship: Immigration, extinction, and scale. Ecology 86: 360 365. Heaney, L. R. 2001. Small mammal diversity along elevational gradients in the Philippines: an assessment of patterns and hypotheses. Global Ecol. Biogeogr. 10: 15 39. Herzog, S. K. et al. 2005. The elevational gradient in Andean bird species richness at the local scale: a foothill peak and a high-elevation plateau. Ecography 28: 209 222. Hoffmann, M. H. et al. 2001. Phytogeographical analysis of plant communities along an altitudinal transect through the Kuraiskaya basin (Altai, Russia). Phytocoenologia 31: 401 426. Holloway, J. D. 1987. Macrolepidoptera diversity in the Indo-Australian tropics geographic, biotopic and taxonomic variations. Biol. J. Linn. Soc. 30: 325 341. Kattan, G. H. and Franco, P. 2004. Bird diversity along elevational gradients in the Andes of Colombia: area and mass effects. Global Ecol. Biogeogr. 13: 451 458. Kessler, M. 2000a. Altitudinal zonation of Andean cryptogam communities. J. Biogeogr. 27: 275 282. Kessler, M. 2000b. Elevational gradients in species richness and endemism of selected plant groups in the central Bolivian Andes. Plant Ecol. 149: 181 193.

Kessler, M. 2001. Patterns of diversity and range size of selected plant groups along an elevational transect in the Bolivian Andes. Biodiv. Conserv. 10: 1897 1921. KoÂ¨rner, C. 2000. Why are there global gradients in species richness? Mountains might hold the answer. Trends Ecol. Evol. 15: 513 514. Kuper, W. et al. 2004. Large-scale diversity patterns of vascular epiphytes in Neotropical montane rain forests. J. Biogeogr. 31: 1477 1487. Lieberman, D. et al. 1996. Tropical forest structure and composition on a large-scale altitudinal gradient in Costa Rica. J. Ecol. 84: 137 152. Lomolino, M. V. 2001. Elevation gradients of species-density: historical and prospective views. Global Ecol. Biogeogr. 10: 3 13. MacArthur, R. H. 1972. Geographical ecology. Harper and Row. McCain, C. M. 2005. Elevational gradients in diversity of small mammals. Ecology 86: 366 372. McCain, C. M. 2006. Could temperature and water availability drive elevational species richness patterns? A global case study for bats. Global Ecol. Biogeogr. DOI: 10.1111/j.1466-8238.2006.00263.x. Md. Nor, S. 2001. Elevational diversity patterns of small mammals on Mount Kinabalu, Sabah, Malaysia. Global Ecol. Biogeogr. 10: 41 62. Migliorini, M. and Bernini, F. 1999. Oribatid mite coenoses in the Nebrodi Mountains (northern Sicily). Pedobiologia 43: 372 383. Mittelbach, G. G. et al. 2003. What is the observed relationship between species richness and productivity? Reply. Ecology 84: 3390 3395. Patterson, B. D. et al. 1998. Contrasting patterns of elevational zonation for birds and mammals in the Andes of southeastern Peru. J. Biogeogr. 25: 593 607. Rahbek, C. 1995. The elevational gradient of species richness a uniform pattern. Ecography 18: 200 205. Rahbek, C. 1997. The relationship among area, elevation, and regional species richness in neotropical birds. Am. Nat. 149: 875 902. Rahbek, C. 2005. The role of spatial scale and the perception of large-scale species-richness patterns. Ecol. Lett. 8: 224 239. Ricklefs, R. E. 1987. Community diversity relative roles of local and regional processes. Science 235: 167 171. Rosenzweig, M. L. 1995. Species diversity in space and time. Cambridge Univ. Press. Rosenzweig, M. L. and Ziv, Y. 1999. The echo pattern of species diversity: pattern and processes. Ecography 22: 614 628. Samson, D. A. et al. 1997. Ant diversity and abundance along an elevational gradient in the Philippines. Biotropica 29: 349 363. Sanchez-Cordero, V. 2001. Elevation gradients of diversity for rodents and bats in Oaxaca, Mexico. Global Ecol. Biogeogr. 10: 63 76. Sanders, N. J. 2002. Elevational gradients in ant species richness: area, geometry, and Rapoportâ&#x20AC;&#x2122;s rule. Ecography 25: 25 32.

447

Sanders, N. J. et al. 2003. Patterns of ant species richness along elevational gradients in an arid ecosystem. Global Ecol. Biogeogr. 12: 93 102 Stevens, G. C. 1992. The elevational gradient in altitudinal range an extension of Rapoport latitudinal rule to altitude. Am. Nat. 140: 893 911. Storch, D. et al. 2005. The species-area-energy relationship. Ecol. Lett. 8: 487 492. Stotz, D. F. 1998. Endemism and species turnover with elevation in montane avifaunas in the neotropics: implications for conservation. In: Mace, G. M. et al. (eds), Conservation in a changing world. Cambridge Univ. Press, pp. 161 180.

Download the appendix as file E4954 from Bwww.oikos.ekol.lu.se/appendix .

448

Terborgh, J. 1973. Notion of favorableness in plant ecology. Am. Nat. 107: 481 501. Terborgh, J. 1977. Bird species-diversity on an Andean elevational gradient. Ecology 58: 1007 1019. Terborgh, J. 1985. The role of ecotones in the distribution of Andean birds. Ecology 66: 1237 1246. Terborgh, J. W. and Faaborg, J. 1980. Saturation of bird communities in the West-Indies. Am. Nat. 116: 178 195. Whittaker, R. H. and Niering, W. A. 1975. Vegetation of Santa Catalina Mountains, Arizona. 5. Biomass, production, and diversity along elevation gradient. Ecology 56: 771 790.

Ecography 32: 411 422, 2009 doi: 10.1111/j.1600-0587.2008.05538.x # 2009 The Author. Journal compilation # 2009 Ecography Subject Editor: Nathan Sanders. Accepted 8 October 2008

Environmental and geometric drivers of small mammal diversity along elevational gradients in Utah Rebecca J. Rowe R. J. Rowe (rrowe@umnh.utah.edu), Committee on Evolutionary Biology, Univ. of Chicago, 1025 E. 57th Street, Chicago, IL 60637, USA, and Div. of Mammals, The Field Museum, 1400 S. Lake Shore Drive, Chicago, IL 60605, USA, (present address: Utah Museum of Natural History, Univ. of Utah, 1390 East Presidents Circle, Salt Lake City, UT 84112, USA).

The mechanisms shaping patterns of biodiversity along spatial gradients remain poorly known and controversial. Hypotheses have emphasized the importance of both environmental and spatial factors. Much of the uncertainty about the relative role of these processes can be attributed to the limited number of comparative studies that evaluate multiple potential mechanisms. This study examines the relative importance of six variables: temperature, precipitation, productivity, habitat heterogeneity, area, and the mid-domain effect on patterns of species richness for non-volant small mammals along four neighboring mountain ranges in central Utah. Along each of these elevational gradients, a humpshaped relationship of richness with elevation is evident. This study evaluates whether the processes shaping this common pattern are also common to all gradients. Model selection indicates that no one factor or set of factors best explains patterns of species richness across all gradients, and drivers of diversity may vary seasonally. These findings suggest that commonality in the pattern of species richness, even among elevational gradients with a similar biogeographic history and fauna, cannot be attributed to a simple universal explanation.

Investigating patterns of biodiversity within and among taxa, regions, and spatial scales may provide a universal set of principles which can guide both theoretical and applied research (Lawton 1999, Whittaker et al. 2001, Hillebrand 2004). In recent years the evaluation of diversity along elevational gradients has become a popular approach, as elevation is correlated with several environmental variables while providing greater relative constancy in ecological conditions, history, and fauna than can be obtained along continental or global spatial gradients. In about half of the recently reviewed 204 datasets, a mid-elevation peak in richness is the most common pattern documented across taxa and regions (Rahbek 2005). Non-flying small mammal assemblages are no exception, with this hump-shaped pattern predominant worldwide (McCain 2005). Despite increased attention to the richness-elevation relationship and apparent commonality in pattern among many studies, our understanding of the determinants of such patterns remains poorly developed and controversial (Heaney 2001, Lomolino 2001). To tease apart the relative importance of factors shaping small mammal diversity patterns along elevational gradients, I conduct comparative, quantitative analyses across multiple gradients, including multiple factors and the consideration of multi-variable models. Specifically, I evaluate the following six variables: temperature,

precipitation, productivity, habitat heterogeneity, area, and geometric constraints on speciesâ&#x20AC;&#x2122; ranges (i.e. the middomain effect), as determinants of species richness for nonflying small mammals along four mountain ranges within the same biogeographic region in Utah, USA. By using multiple regression models and the Akaike information criterion, I determine whether a given factor or set of factors has a consistent explanatory value across gradients. Along each gradient, a hump-shaped relationship of richness with elevation is evident (Rowe and Lidgard 2009). Therefore this study evaluates whether the processes shaping this common pattern are also common. Climate is considered the most widely supported predictor of biodiversity worldwide (Hawkins et al. 2003, Currie et al. 2004, Storch et al. 2006). Its influence in shaping diversity can be both direct and indirect. Directly, climate can set limits on species distributions by exceeding species physiological tolerances (Connell 1961). Indirectly, gradients of temperature and precipitation are thought to establish trends in energy availability (often measured as primary productivity) which can affect photosynthetic activity and rates of biological processes, and is thus considered a primary driver of biodiversity patterns (Hawkins et al. 2003, Currie et al. 2004, Storch et al. 2006). However, the form of the productivity-diversity relationship appears to be contextual, with different

411

responses contingent upon which taxa are incorporated and on the spatial and ecological scale of the analysis (Mittelbach et al. 2001, Chase and Leibold 2002). Numerous studies along elevational gradients have found strong support for the role of climate in structuring species richness patterns (Bailey et al. 2004, Bhattarai et al. 2004, Fu et al. 2006, Sanders et al. 2007). Habitat heterogeneity is another form of environmental variability that influences the production and maintenance of diversity. An increase in the number of habitat types or greater structural complexity in vegetation can provide more resources than a more uniform environment and may therefore support a greater number of species (Hutchinson 1957, Chase and Leibold 2003). Although traditionally interpreted as local-scale heterogeneity, habitat heterogeneity also can shape diversity at larger spatial scales (Kreft and Jetz 2007, Davies et al. 2007). In elevational gradient analyses habitat heterogeneity appears to be less frequently evaluated than climatic factors, possibly due to difficulties associated with measuring this variable (Heaney 2001). However, when considered, a role for differences in habitat conditions in structuring species richness patterns is often evident (Terborgh 1977, Shmida and Wilson 1985, Rickart et al. 1991). Spatial factors also have been shown to influence diversity patterns. Most predominant are the mid-domain effect (MDE) and species-area effect. MDE postulates that geometric constraints imposed on species ranges within a bounded domain will yield a mid-domain peak in richness, without requiring any recourse to ecological or historical explanatory factors (Colwell and Hurtt 1994, Colwell and Lees 2000). Some elevational gradient studies show good agreement between observed patterns and those simulated under this null model, (Kluge et al. 2006, Watkins et al. 2006), while others find hard boundaries to be poor determinants of richness patterns (Herzog et al. 2005, McCain 2007). Goodness of fit may depend on the region and spatial scale of analysis (Oommen and Shanker 2005, Dunn et al. 2007) or degree of endemicity in the fauna (Fu et al. 2006). Under the area hypothesis larger regions are expected to be more diverse (MacArthur and Wilson 1967, Connor and McCoy 1979, Rosenzweig 1995). This species-area relationship can be accounted for under two principle hypotheses: 1) greater area provides greater habitat diversity which can harbor more species (Williams 1964), and 2) increases in area are accompanied by decreased rates of extinction and increased rates of speciation or colonization due to a greater number of barriers and the maintenance of larger population sizes (Preston 1962, MacArthur and Wilson 1967). Typically the hypothesis asserted varies with spatial extent, where habitat diversity is often considered the primary driver at local to landscape scales and the processes of colonization and extinction predominate at larger regional to global scales (Rosenzweig 1995). Because mountain ranges can fall into either category, elevational gradient analyses should consider both hypotheses (Heaney 2001, McCain 2007). In this study, I evaluate the impact of habitat diversity directly using satellite derived vegetation data. The inclusion of area as a variable therefore reflects the role of evolutionary processes or other elements of habitat heterogeneity (e.g. structural complexity) that 412

may not be subsumed under a diversity metric. Elevational gradient analyses have found variable effects of area on diversity, even when the same richness-elevation pattern is documented (Rahbek 1997, Sanders 2002, Vetaas and Grytnes 2002, Oommen and Shanker 2005, McCain 2007). Many potential mechanisms for elevational patterns of species richness have been proposed and examples can be found that either support or refute different hypotheses, even among studies restricted to non-flying small mammals. Unfortunately, many explanations for small mammal richness patterns are qualitative and where formal tests have been conducted a limited number of variables were considered and typically tested independently. Rigorous comparative analyses are essential to improve our understanding of the processes structuring diversity. Recent metaanalyses (McCain 2005, 2007) suggest that the interaction of climate, area, and MDE are driving small mammal richness patterns along elevational gradients worldwide. Area and MDE, both alone or in combination, can influence these patterns although high variability among studies and low explanatory power suggests these factors are unlikely to be primary drivers (McCain 2007). Although climatic factors have not been directly tested, using the mountain mass effect as a proxy provides a strong climate signal across small mammal studies (McCain 2005). Direct evidence for the role of climate in determining non-volant small mammal richness patterns has been found in largerscale studies. A recent meta-analysis (Hortal et al. 2008) of species richness for small mammals at meso-scale sites (100 10 000 km2) worldwide found that the majority of variation in richness was attributed to a combination of water and energy availability and habitat type. In comparison, measures of area and landscape (including habitat) heterogeneity had limited impact on richness. Comparative analyses are essential to place our understanding of pattern and process in a more comprehensive framework. By directly evaluating environmental and spatial hypotheses, independently and in combination, this study provides a comprehensive assessment of the relative contribution of factors shaping diversity patterns for non-flying small mammals along elevational gradients. Furthermore, by examining the same suite of factors along a series of neighboring gradients this study provides insight into whether the factor(s) structuring diversity are uniform under conditions of relative constancy in regional history and fauna.

Material and methods Region and fauna The Uinta Mountains, Wasatch Plateau, Aquarius Plateau, and Markagunt Plateau (Fig. 1) all are located within the Rocky Mountain region of the western USA, and thus share a common geological history, flora, and fauna, which are distinct relative to the neighboring biogeographic regions (Harper and Reveal 1978). Previously connected by forest vegetation during the last glacial maximum ( 20 000 18 000 ybp), today, these mountain ranges are isolated from one another by narrow passes of dry habitat at low

Figure 1. Map showing the location of the four mountain ranges in Utah. UI: Uinta Mountains, WA: Wasatch Plateau, AQ: Aquarius Plateau, and MA: Markagunt Plateau. The black line indicates lowland boundary (veil line) of the mountain range, gray polygons represent area ]2300 m elevation and open circles represent localities.

elevation or connected by narrow ridges at higher elevation. Although these mountain ranges share a common regional history, they vary in species richness, area, and peak height (Table 1, Fig. 1) as well as local habitat and microclimate conditions. Sampling occurred across at least 80% of each gradient, as measured from the lowland boundary (veil line) to the peak height (Table 1). To establish the veil line I combined a low elevation contour line (1600 m) with the major stream channels surrounding each mountain range, ultimately using topography and natural drainage patterns to circumscribe each local geographic gradient (Fig. 1). Data from recent field surveys and museum specimen records were used to determine the richness-elevation relationship for the non-flying small mammal faunas on each mountain range. A total of thirty-seven species are represented, thirty-one are rodents from five families: Muridae (13 species), Heteromyidae (3), Dipodidae (1), Geomyidae (3), and Sciuridae (11). Five species are shrews (Soricomorpha: Soricidae), and one, the pika, is a Lagomorph. Between 2002 and 2005, colleagues and I conducted field surveys along each gradient, amassing 2187 specimen records from 73 sites. At each site, sampling

occurred for a minimum of three nights, continuing until the species accumulation curve was asymptotic. Multiple trap types (e.g. Sherman live traps, museum specials, Victor rat-traps, gopher traps, and pitfall traps) were used to ensure that all members of the community were targeted. I supplemented these data with records from sixteen museum collections, using only localities with low spatial uncertainty (Rowe 2005). Most museum records were from comprehensive faunal surveys to document the mammal fauna of a particular area (Durrant 1952, Hall 1981), and are therefore suitable for use in defining species ranges. All museum localities were georeferenced using the same protocol (Wieczorek et al. 2004), and all elevations were derived from a 30 m digital elevation model (DEM) created in ArcInfo Workstation (ESRI, Redlands, CA, USA). In total, 9208 records from 741 localities were used to determine species richness patterns along these gradients (Rowe and Lidgard 2009). To document the richness-elevation relationship, each gradient was divided into 100 m elevational bands or bins (e.g. 1600 1699, 1700 1799 m etc.) with richness representing the number of species across all sites within a given 413

Table 1. The number of species, specimens, and localities for each mountain range. Locality numbers represent unique combinations of coordinates, spatial uncertainty, and year (Rowe 2005). Sampled domain represents the elevations of the lowest and highest sites along each gradient. Elevations were derived from a digital elevation model (DEM, 30 m resolution). Gradient extent spans from the lowest elevation along the veil line to the highest peak in the mountain range. Mountain range Uinta Mountains Wasatch Plateau Aquarius Plateau Markagunt Plateau

Species

Specimens

Localities

Sampled domain (m)

Gradient extent (m)

33 27 30 25

3123 1215 2041 907

268 109 115 124

1623 3671 1640 3167 1601 3399 1734 3426

1564 4118 1540 3440 1571 3448 1482 3449

band. Additionally, species were assumed to be present within a band if they were sampled within higher and lower elevational bands. This range-through assumption is warranted because the perceived absence of a species at this scale is likely attributed to sampling incompleteness rather than to a true gap in the distribution of the species (Grytnes and Vetaas 2002, Colwell et al. 2004). This is especially the case along a mountain where suitable habitat for a given species is available (even if patchy in distribution) in elevational bands bracketed by those in which it has been documented, thus indicating the potential for a resident population or dispersal event at any point in time or space along that gradient. The species ranges used in this study vary slightly from those presented in Rickart (2001); these discrepancies reflect results from recent field work, differences in how the lowland boundaries were delineated, and variation in sources used to compile occurrence records. Further details are available by Rowe and Lidgard (2009). Species richness is hump-shaped along each elevational gradient with a peak in richness occurring at about the midpoint of the sampled gradient (Fig. 2, Supplementary material, Appendix 1). The general shape of the richnesselevation relationship was not impacted by applying the range-through assumption (Rowe and Lidgard 2009). This study, examines the relative importance of several environmental and spatial factors in shaping the species richness patterns along these four neighboring gradients. These mountain ranges represent four of the six gradients discussed in Rowe and Lidgard (2009). Although the same pattern was documented for all mountain ranges, sampling concerns rendered the Fishlake and Tushar Mountain data sets not suitable for the analysis presented here. Local topography and accessibility in the Fishlake Plateau prohibited sampling sites in six of the eighteen 100 m elevational bands along this gradient. Assessing the processes driving richness patterns along this gradient would be problematic because climate data (see below) are extracted at the level of sites and therefore only twothirds of the gradient could be included in the analyses. Similarly, the lower portion of the Tushar Mountain gradient has a cluster of poorly sampled bins, containing only one site (Rowe and Lidgard 2009), thus minimizing the number of sites from which climate data can be averaged within and among bins along this gradient compared to others. Patterns of spatial autocorrelation in richness data can impact the results of statistical analyses, inflating type I errors or potentially biasing results toward certain processes (Diniz-Filho et al. 2003). To address this concern I tested for spatial autocorrelation in the underlying richness data for each gradient. Analyses were conducted in ArcGIS, 414

using Moranâ&#x20AC;&#x2122;s I to compare the location of sites and their richness values. Because richness is assessed using 100 m elevational bands, I used the mean center of sites within each 100 m bin for the location data. None of the richness values in the four gradients analyzed here were spatially autocorrelated (Supplementary material, Appendix 2). Climate and productivity Two climate variables were investigated with respect to species richness: air temperature and precipitation. Normals (1971 2000) for air temperature and precipitation were obtained for Utah from the PRISM Group (Oregon State Univ., /<http://www.prismclimate.org//>, created 4 Feb. 2004), at 4 km resolution. This resolution is fine enough to capture variability in climate for these regional elevation gradients where individual 100 m bands of elevation range in size from ca 7 to 2000 km2, with an average of 648 km2. More specifically, the following nine climate data sets were acquired: annual, January, and July values for total precipitation, maximum and minimum air temperature. As proxies for above-ground net primary productivity, I used two satellite-derived vegetation indices, the normalized difference vegetation index (NDVI) and enhanced vegetation index (EVI). Sixteen-day, 1 km composites of NDVI and EVI for June 2005 were derived from the Moderate-resolution Imaging Spectroradiometer (MODISTerra) and obtained via the Land Processes Distributed Active Archive Center /<http://lpdaac.usgs.gov/main.asp>. June imagery was selected as it corresponds to the middle of the growing season and is prior to peak monsoonal rain season, thus avoiding complications of additional cloud cover. A winter image was not obtained due to snow reflectance. NDVI is computed from the near infrared and red portions of the spectral bands. EVI is an improvement upon NDVI in accounting for residual atmospheric contamination (e.g. aerosols) and additional soil and canopy reflectance (Huete et al. 1997). Both indices permit reliable spatial and temporal comparisons of primary productivity, terrestrial photosynthetic activity, and canopy characteristics, such as leaf area index and above-ground biomass (Cramer et al. 1999, Serrano et al. 2000). In these mountains, the highest productivity sites were habitat mosaics dominated by cover types of spruce-fir, mountain-fir, pinyon-juniper, aspen, grassland, meadow, and mountain shrub. Sites with low productivity were predominantly barren or alpine. For each locality (109 268 depending on the gradient, Table 1), the corresponding value for each of the nine climate and two productivity variables was extracted using

mountain range, site-based climate data were then grouped into their corresponding 100 m elevational bands, and the mean value of all sites within each band was calculated for each variable. These mean values were used as the predictor variables in the analyses. This is necessary because the species richness data are summarized at the scale of 100 m elevational bands. Since multiple measures of each factor were gathered: precipitation (3), temperature (6), productivity (2), exploratory analyses were conducted to identify any significant correlations among factors, to determine which variable for each category would be represented in the multi-model analysis, and to examine whether the relationship of a given variable across each gradient was similar. Habitat heterogeneity Habitat heterogeneity was summarized using the Shannon diversity and evenness indices, both of which are commonly applied metrics at the landscape scale (Turner et al. 2001). In ArcGIS, I combined the Utah GAP land cover data (USGS, Biological Resources Division, /<http://earth.gis. usu.edu/index.html>) and a 30 m digital elevation model (DEM) of each mountain range to measure the area for each land-cover type in each 100 m elevational band (e.g. 1600 1699, 1700 1799 m etc.) per mountain range. The land cover data were modeled using Landsat 5 Thematic Mapper imagery from 1988 to 1989, generated at 30 m resolution. Cover type was assigned based on satellite imagery in combination with field observations, aerial photos, and existing maps (Scott et al. 1993). Thirtyeight land-cover types are defined and primarily reflect different classes of forest, woodland, shrubland, and herbaceous communities based on the principal associated species. Cloud, urban, and water land-cover types were excluded from analysis. Exploratory analyses were conducted on the two measures of habitat heterogeneity, the Shannon diversity and evenness indices, to identify whether they are correlated, to determine which variable would be represented in the multi-model analysis, and to examine whether the relationship of a given variable across each gradient was similar. Spatial variables Figure 2. The pattern of species richness along each elevation gradient. Richness values are based on the range-through assumption, where presences are inferred in all bins between the highest and lowest recorded localities for each species. Solid lines represent the second-order polynomial fit under a generalized linear model. Each GLM used a quasipoisson distribution and a log-link function (Supplementary material, Appendix 1).

Spatial analyst in ArcGIS. Site values for NDVI and EVI were excluded from analysis if the 1 km pixel associated with a given site was dominated by agricultural land or large bodies of water, which would create very high or low measures, respectively. This resulted in the removal of 40 outliers (6.5% of total sites). Regressions of the sitebased data for each variable across each gradient were conducted and the residuals were homogeneous. For each

To test the area hypothesis I calculated the amount of threedimensional surface area for each 100 m band of elevation on each mountain (e.g. 1600 1699, 1700 1799 m etc.). Surface area is measured along the slope of the landscape, therefore accounting for variation in the height of the surface. Calculations were done using 3D analyst in ArcGIS and were based on the 30 m DEM of each mountain range. Area was log-transformed for analysis. How best to test the MDE is a topic of much debate and the manner in which the hypothesis is tested can influence the estimated impact of domain boundaries on diversity patterns (Bellwood et al. 2005, McClain et al. 2007). Here, I test the fit of the mid-domain null model using both a Monte Carlo simulation program called Mid-domain null (McCain 2004) and by including MDE as a predictor variable in the multi-model analysis. The Mid-domain null 415

algorithm randomizes (without replacement) the empirical species’ ranges within the bounded domain of each mountain to generate a predicted species richness pattern under the null model (McCain 2004). Predicted values are computed for each 100 m band of elevation per mountain range and are based on the mean of 50 000 simulations. Observed and predicted richness values were compared in a regression analysis with goodness of fit assessed using both the coefficient of determination (r2) and by evaluating the corresponding model slope and intercept. The slope and intercept reflect the fit in shape and magnitude, respectively, between the observed pattern of richness and the mid-domain predictions (Jetz and Rahbek 2001). Using coefficients of determination has recently been criticized, in part, because the fit of the model is based on relative, not absolute magnitude (Colwell et al. 2004, Zapata et al. 2005), and evaluation of the intercept and slope is considered a more appropriate form of assessment (Jetz and Rahbek 2001). In multi-model analysis, I evaluate the relative contribution of MDE in conjunction with area and the environmental factors. In doing so, I use a different predictor variable for MDE than was used under the Mid-domain null program. In the multi-model analyses, I use distance (i.e. differences in elevation) from the domain boundaries as the predictor variable. For example, for all sites below the mid-domain I calculated the difference in elevation between each site and the lowest sampled site. Likewise, for sites about the mid-domain I calculated distances between each site and the highest sampled site. This approach is analogous to that used by Bellwood et al. (2005) in their study of Indo-Pacific coral reef biodiversity. In order to draw a comparison with the species richness data, for each gradient, I took the average of these distances among sites within each 100 m band of elevation. This resulted in a largely symmetrical monotonic decrease from the middomain to the gradient boundaries; a feature which is shared among all MDE models (Bellwood et al. 2005). Multi-model inference Six variables were included in the multi-model analyses, one from each general category: temperature, precipitation, productivity, habitat heterogeneity, area, and the middomain effect (MDE). Using only one variable from each category is necessary to reduce collinearity in multiple regression analyses. For each variable, I first evaluated the form of the relationship (linear or quadratic) with species richness. It is important to determine the appropriate form of the relationship because more complicated models tend to explain more variability in the data and thus provide a seemingly better fit (as estimated by r2 and p values), but may do so spuriously. As a result, a method for model selection must be adopted. For these model comparisons, I apply Akaike’s information criterion (AIC) which takes into account both the fit of the model (the residual sum of squares) and the complexity of the model (the number of parameters) (Akaike 1973, Burnham and Anderson 2002). Because of the small sample size (number of 100 m elevational bands in a gradient) relative to the number of 416

parameters in a model, when computing the score for each model I used the second-order AIC (AICc) which incorporates a bias-correction term (Hurvich and Tsai 1989, Burnham and Anderson 2002). I use the simpler linear model unless the quadratic model is a better fit as measured by an AICc score 4.159. This criterion follows Royall’s (1997) interpretation of likelihood ratios as evidence ratios, with an AICc score of 4.159 units (equivalent to an evidence ratio of 8) representing strong evidence supporting one hypothesis over another (Akaike 1973, Burnham and Anderson 2002). Multiple regression models were constructed for every combination of predictors (including each variable independently), resulting in a total of sixty-three models for each mountain range. All models were considered because all variables are of interest, they are not redundant (i.e. collinearity has already been reduced), and there is no a priori reason to exclude a given set of models from consideration. To evaluate the relative contributions of the different predictor variables in shaping patterns of species richness along each gradient, I again use AICc scores to assess model fit (Akaike 1973, Burnham and Anderson 2002). As mentioned above, in comparing the fit of a model to other models, this approach penalizes models that may fit the data slightly better but only because they have a greater number of parameters. In contrast to p values, Akaike weights are calculated for each model and provide the strength of measure of evidence for each model relative to the other models (Burnham and Anderson 2002, Hobbs and Hilborn 2006). Because determining a ‘‘best’’ model can be problematic, I calculated the impact factor for each predictor variable to assess their relative importance in structuring patterns of species richness along each gradient. Impact factors are a form of cross-model validation, and are simply the sum of the weights for each model containing the particular predictor (Burnham and Anderson 2002). As posterior probabilities over the set of hypotheses, impact factors ]0.80 were interpreted as strong evidence for the role of a given variable in shaping these diversity patterns. All regression analyses were carried out using SPSS (v.13.0; SPSS, Chicago, IL, USA).

Results Exploratory analysis Positive correlations were apparent among the different measures examined for each environmental variable: habitat heterogeneity, precipitation, temperature, and productivity. For habitat heterogeneity, Pearson correlation coefficients for the Shannon diversity and evenness indices ranged from 0.948 to 0.996 across gradients, each of which was significant at pB0.01. Similarly, for a given gradient, annual, January, and July values for precipitation were positively correlated (r 0.733 0.981, pB0.01 in all cases). Temperature measures also were positively correlated for each gradient (Tmax r 0.686 0.999, p B0.01 in all cases; Tmin r 0.536 0.999, pB0.05 in all cases), as were EVI and NDVI (r 0.790 0.976, pB0.01 in all cases). One measure for each category was chosen to illustrate the distribution of these environmental variables along each

Figure 3. The distribution of habitat heterogeneity, area, temperature, precipitation and productivity (EVI) along each elevation gradient (see Methods). The dotted line represents the elevation at maximum species richness.

elevational gradient (Fig. 3). Habitat heterogeneity is unimodal along each gradient, peaking at about mid-elevation. Although the placement and value of maximum habitat diversity are similar across gradients, the recorded minimum diversity varies. Productivity also is unimodal with elevation; however, the elevation at which productivity peaks varies

among gradients from low-mid elevations to mid-high elevations. Productivity values at high elevation tend to be much lower than those at low elevation because of the predominance of barren habitat or tundra. Precipitation increases linearly with increased elevation while temperature decreases linearly along the gradient. 417

When evaluated individually as predictors of richness, there is strong support for the importance of these environmental variables in shaping species richness patterns. Habitat heterogeneity was a significant linear predictor of species richness for three of the four gradients. The exception is the Markagunt Plateau where habitat diversity is relatively high and constant at low elevations, resulting in a great discrepancy between estimates of richness and habitat diversity for this portion of the gradient. The magnitude of effect (as measured by the p value) in predicting richness was similar under the Shannon diversity or evenness measures (Supplementary material, Appendix 3); only the Shannon diversity index was incorporated in the multi-model inference. A linear model was a better fit with richness for productivity, but not for precipitation and temperature. Analyses of productivity were significant for two gradients (Supplementary material, Appendix 3). Because EVI and NDVI were consistent in magnitude within a gradient and EVI is considered an improvement upon NDVI, only EVI was incorporated in the multiple regression analyses. For precipitation and temperature, multiple measures were not always consistent in the presence or magnitude of a significant fit within and among gradients (Supplementary material, Appendix 3). When the magnitude of the fit differed across precipitation variables, deviations favored annual or July measures. Similarly, for temperature, annual and July models were often equal in magnitude and a better fit than January models (Supplementary material, Appendix 3). July values for precipitation and temperature were therefore selected for use in the multiple regression analysis. To further reduce collinearity, I incorporated only maximum temperature measures. To investigate the potential seasonal effect in predictor strength, I conducted a second set of multiple regression analyses which incorporated January rather than July precipitation. This was not done with temperature, because the temperature experienced during one season is less likely to influence montane mammal populations into the next season in the same way as precipitation. As has been noted previously, area does not always decrease with increasing elevation (McCain 2007). In the Uinta Mountains, local patterns of topography result in a hump-shaped distribution of area with elevation (Fig. 3). Although initially counterintuitive, this pattern can result when low elevations do not occur along all sides of a mountain range and because they typically occupy valley floors or canyon mouths which are more uniform in relief and therefore smaller in area. In contrast, the three other gradients generally show a decrease in area with increasing elevation (Fig. 3). The form of the richness-area relationship also varied among gradients. The simpler linear model was a better fit for both the Uinta Mountain and Markagunt Plateau gradients, but a quadratic model was preferred in the Wasatch and Aquarius Plateaus. Under the Mid-domain null program, fit of the MDE varied with the method employed. Empirical richness values for each mountain range fell largely within the 95% prediction curves, and the r2 values from the linear regressions were high, suggesting good predictive ability for the MDE across all gradients (Table 2). Evaluation of the slope (fit in shape) and intercept (fit in magnitude) 418

Table 2. Results of the mid-domain analysis when considered as a single variable. Observed richness values were regressed against the mean predicted values generated after 50 000 simulations using the Mid-domain null program (McCain 2004). A linear relationship was used for each gradient. The coefficients of determination (r2) were significant (pB0.001) for each gradient. * Indicates a significant deviation from values expected under the null model (pB0.05) from one for the slope (S) and zero for the intercept (I). Mountain range Uinta Mountains Wasatch Plateau Aquarius Plateau Markagunt Plateau

0.83 0.77 0.64 0.60

1.40 1.99 7.73* 1.44

0.92 0.86 0.44* 0.87

indicates a good fit for only three of the four mountain ranges (Table 2). For the multiple regression analyses, the form of the richness-MDE relationship was not consistent across gradients. The simpler linear model was preferred for the Uinta Mountains, Aquarius Plateau and Markagunt Plateau gradients. A quadratic relationship fit best for the Wasatch Plateau. Multi-model inference The multi-model inference included one representative variable from each of the six categories: Shannon habitat diversity, EVI, precipitation (July or January), July maximum temperature, surface area, and MDE. The same sixtythree models were evaluated per mountain range. However, because the form of the richness-area and richness-MDE relationships (linear or quadratic) was not consistent across gradients, a model including the same variables can have a different number of parameters across model sets. Detailed results for each set of models are presented in Supplementary material, Appendix 4, 5. Impact factors, in general, are not consistent for a given predictor variable across gradients (Table 3a, b). For the set of models incorporating July precipitation, the Uinta Mountain model with the highest Akaike weight (wi 0.5646) includes both temperature and area (Supplementary material, Appendix 4), and both variables have high impact factors (]0.80), explaining the vast majority of the posterior probabilities (Table 3a). For the Wasatch Plateau the two models with the highest Akaike weights were MDE and area (wi 0.3656), and MDE and temperature (wi 0.3248). Only MDE has a high impact factor (0.94, Table 3a). The model incorporating both precipitation and MDE had the highest weight (wi 0.6349) in the Markagunt Plateau and impact factors are high for both variables (Table 3a). In the Aquarius Plateau, the ‘‘best’’ model includes precipitation and productivity (wi 0.3790), however, neither variable has a high impact factor: precipitation (0.74) and productivity (0.57); these measures are relatively weak and may vary given changes made to the underlying dataset. The impact factor for habitat heterogeneity was low for all gradients. The results of the second multi-model inference, replacing July with January precipitation, corroborate the first analysis for the Uinta Mountains and Wasatch Plateau (Table 3a, b). For the Wasatch Plateau, although the

Table 3. Impact factor for each variable across gradients. Impact factors are the posterior probabilities over the set of hypotheses, representing the sum of the Akaike weights for each model containing the particular predictor in the model set (see Burnham and Anderson 2002 for details). Values in bold (]0.80) are considered strong support. Variables are July or January precipitation (PPT), July maximum temperature (Tmax), enhanced vegetation index (EVI) as a proxy for productivity, habitat heterogeneity as measured by the Shannon diversity index (H’), the mid-domain effect (MDE), and surface area (AREA). All models and summary statistics can be found in Supplementary material, Appendix 4 (JulPPT) and 5 (JanPPT). (a) Analyses using July precipitation Mountain range Uinta Mountains Wasatch Plateau Aquarius Plateau Markagunt Plateau

JulPPT

Tmax

EVI

H’ *

MDE

AREA

0.21 0.18 0.74 0.87

0.81 0.42 0.24 0.12

0.12 0.07 0.57 0.10

0.10 0.04 0.26 0.10

0.19 0.94 0.13 0.99

1.00 0.42 0.12 0.10

JanPPT

Tmax

EVI

H’ *

MDE

AREA

0.02 0.69 0.01 0.53

1.00 0.16 0.90 0.40

0.10 0.30 0.21 0.10

0.09 0.04 0.58 0.17

0.11 0.83 0.19 0.99

1.00 0.17 0.15 0.09

(b) Analyses using January precipitation Mountain range Uinta Mountains Wasatch Plateau Aquarius Plateau Markagunt Plateau

*Calculations of H’ for the Uinta Mountains were restricted to the Utah GAP analysis dataset because the vegetation catgories for Wyoming differed. This effected 6 of the 22 bins and in each case the majority of area is within Utah (Fig. 1).

variables included in the models with the highest weights differ from the previous analysis, precipitation and MDE (wi 0.3696), and precipitation, productivity and MDE (wi 0.2691), MDE remains the only variable with a high impact factor (0.83, Table 3b), followed by weaker support for precipitation (0.69). In the Markagunt Plateau, the model containing precipitation and MDE still has the highest weight, although it is lower in magnitude (wi 0.3961) than in the previous analysis, and only MDE retains a high impact factor (0.99). In the Aquarius Plateau, changing the precipitation data results in strong support for temperature (0.90), and relatively weaker support (0.58) for habitat heterogeneity (Table 3b). This model combining temperature and habitat heterogeneity had the second highest weight in the previous analysis (JulPPT: wi 0.1206, JanPPT: wi 0.4594). Variation in the relative strength of a given environmental factor across gradients cannot be attributed to differences in aspect within and across elevational bins, as site-based analyses of aspect reveal a similar range of values within bins along each mountain range (data not shown).

Discussion No one factor or set of factors best explains the humpshaped pattern of species richness across these neighboring gradients. This dissimilarity among gradients is notable given commonality in biogeographic history, fauna, and gradient extent. These findings suggest that that there is no universal explanation for the processes shaping diversity and that local ecological factors may play a strong role. Factors identified as significant under individual regression analyses are not necessarily identified as contributors of richness patterns using multi-model inference. This is best illustrated by habitat heterogeneity which, when tested individually, was a strong predictor of richness (as measured

by a p value) along three of the four gradients (Supplementary material, Appendix 3), but was not identified as a critical factor for any gradient when multiple variables were considered in combination (Table 3a, b). This highlights the importance of considering factors in combination in order to develop a deeper understanding of how diversity patterns are structured. Although this study suggests that the effect of other factors is stronger than habitat heterogeneity along these gradients, it is possible that the role of habitat is scale dependent and would be more pronounced at the site level, or that the measure used here does not adequately reflect habitat use by small mammals. Although MDE has been identified as a primary driver of diversity along other spatial gradients, both elevational and larger (Bellwood et al. 2005, Kluge et al. 2006), this study finds an inconsistent role for MDE in shaping elevational patterns of species richness for non-flying small mammals. Under multi-model inference, MDE is identified as a critical driver of diversity along two of the four gradients (Table 3a, b). Given the similarities in domain extent and fauna among these gradients, this study suggests that variability in the role of MDE can not always be attributed to the scale of analysis (Dunn et al. 2007). In addition, because the veil line was defined in a consistent manner across gradients, it is unlikely that the impact of incorporating ‘‘soft’’ domain limits has affected fit of the null model as has recently been suggested (Zapata et al. 2005). In contrast to the multi-model inference, individual regression analyses generated from the Mid-domain null program suggest fit of the MDE for either three or four of the gradients, depending on the method employed (Table 2). Mismatch between the different forms of assessment has been noted previously, as high coefficients can be generated despite a marked discordance among the observed and expected patterns (Colwell et al. 2004, Zapata et al. 2005). Using data presented in Rickart (2001), McCain (2005) used the Mid-domain null program and coefficients of 419

determination to assess fit of MDE for the Uinta Mountains, Wasatch Plateau, and Aquarius Plateau. Those findings and the ones presented here vary with respect to the Uinta Mountains and likely reflect differences in the underlying data sets. Additionally, McCain (2007) suggests that when area is accounted for, MDE is no longer a strong predictor of richness along the Wasatch Plateau, a finding which is not corroborated here. Variation in fit of MDE among the three methods presented here underscores the difficulty of drawing comparisons across studies using alternate methods and ultimately the challenge of discerning the role of geometric constraints on diversity patterns. Overall, the findings presented here question the generality of MDE as the primary explanation for elevational patterns of species richness. The effect of area on diversity patterns is not consistent. Of the four gradients analyzed here, a strong richness-area relationship was found for only one gradient, the Uinta Mountains (Table 3). Variability in area effects for small mammal diversity patterns have been documented previously and differences largely attributed to the shape of the richness-elevation relationship, with area effects predominant among decreasing diversity patterns (McCain 2007). In the present case, each gradient has a hump-shaped pattern of richness with elevation (Fig. 1, Supplementary material, Appendix 1). However, an area effect is demonstrated in the one case where area also has a curvilinear relationship with elevation (Fig. 3, Table 3), thus meeting the expectation of a strong richness-area relationship in instances where both variables show concordant patterns along elevation. Results from the multi-model analysis differ from those of the individual regression analyses, where an area effect is found for three of the four gradients, thus emphasizing the need to consider variables in combination when testing for diversity patterns. Results from these individual regressions also vary from some of the findings reported in McCain (2007) using data presented in Rickart (2001). Where individual regressions presented here suggest an area effect in the Uinta Mountains, Wasatch Plateau and Aquarius Plateau (Supplementary material, Appendix 3), McCain (2007) demonstrates an area effect for two of these three gradients, the Uinta Mountains and Wasatch Plateau. Although results for individual gradients may vary between McCain (2005, 2007) and this study, both emphasize the variability in fit for particular variables across gradients. This study supports a climate-richness relationship, however, as seen with the other variables discussed, the effect is not universal. The particular climate factor(s) identified as primary drivers of richness varied across gradients and may vary seasonally for a given gradient (Table 3a, b). The analyses presented here suggest that the relative impact of July precipitation and temperature on diversity patterns varies with latitude. In the most northern of the mountain ranges, the Uinta Mountains, temperature is identified as a critical factor. In contrast, precipitation is significant in the Markagunt Plateau which is the most southern of the gradients (Table 3a, Fig. 1). In the remaining two gradients although neither temperature nor precipitation have very strong impact factors this trend is apparent, and the variables incorporated in the model(s) with the highest Akaike weights follow this pattern 420

(Supplementary material, Appendix 4, Results). Seasonal variation and discrepancies among gradients may, in part, be related to summer monsoonal rains moving in from the Gulf of Mexico which tend to be heaviest in the southern mountain ranges (i.e. the Markagunt and Aquarius gradients, Fig. 1). Variation in July precipitation across this region (Pearson correlation between latitude and minimum rainfall; n 6, r 0.903, pB0.05), supports this hypothesis, providing evidence for increased summer rainfall in southern ranges relative to northern ones. The influence of seasonal monsoonal rains is further supported by the absence of precipitation as a critical factor along these southern gradients when January values were incorporated (Table 3b). In contrast, for the northern mountain ranges, the factors shaping species richness did not vary under different measures of precipitation. Among environmental factors, primary productivity (a proxy for energy availability) is frequently cited as a fundamental determinant of biodiversity, although controversy remains as to the nature of the relationship (linear or curvilinear) and the potential underlying mechanisms (Mittelbach et al. 2001, Hawkins et al. 2003, Currie et al. 2004). Studies using satellite-derived vegetation indices as proxies for primary productivity have found significant productivity-richness relationships, both linear and unimodal, suggesting that such measures can be used to assess biodiversity patterns, albeit at different spatial scales depending on the taxonomic group in question (Hurlbert and Haskell 2003, Bailey et al. 2004, Seto et al. 2004). This study finds little support for productivity (as measured by EVI) as a primary driver in shaping patterns of species richness along elevational gradients (Table 3). Although both richness and productivity are unimodal with respect to elevation, the distribution of productivity with elevation does not necessarily correspond with richness for each gradient. This suggests that the relationships between productivity and elevation and between productivity and richness may be more complex than previously thought. Among elevational gradient studies, location of peak productivity is often qualitatively asserted based on inverse or largely incongruent trends in precipitation and temperature (Heaney 2001, McCain 2004). These results demonstrate that such assertions may be misleading. Poor support for productivity questions the energydiversity hypothesis for the small mammal fauna along these gradients. Recent studies suggest that energy input is only a strong predictor of diversity in the far northern portions of the globe, and that it is precipitation or the interaction between energy input and moisture which shapes large-scale biodiversity patterns worldwide (Hawkins et al. 2003). However, the results presented here provide weak support for this hypothesis as well, suggesting that the relationship between climate and species richness along smaller spatial gradients (i.e. a mountain range) may be incongruent with those found at larger spatial scales. It has been suggested that the richness patterns observed today are the result of historical processes (namely speciation, immigration, and extinction) operating across time and space and that the correlation with current environmental factors does not reflect causality (Ricklefs 2004). Although the role of historical processes cannot be tested directly, I have attempted to control for such processes by

examining multiple gradients within the same biogeographic region that share a common regional history, species pool, and general pattern of land use. The evolutionary histories of these gradients are likely to be more similar than they are different. If historical and regional processes were predominant forces in shaping richness patterns, then greater commonality in the impact factor of a variable across gradients would be expected. These analyses present a rigorous and comprehensive assessment of the relative contribution of the MDE, area, and several environmental factors in shaping species richness patterns for non-flying small mammals along four neighboring elevation gradients. The results indicate that no one factor or set of factors best explains patterns of species richness across all gradients. Rather, factors(s) identified as primary drivers of richness appear gradientspecific and may vary seasonally. Because these mountain ranges neighbor one another and share a common biogeographic history, it can be inferred that the observed differences in explanatory power of the causal factors are likely the result of local ecological rather than regional or historical forces. Overall, these results suggest that perceived commonality in the pattern of species richness across elevation gradients cannot be attributed to a simple universal explanation. Future comparative research projects will be needed to determine whether the results presented here hold under different sampling regimes, within different regions and differing taxonomic groups.

Acknowledgements I thank L. R. Heaney, S. Lidgard, M. A. Leibold, T. D. Price and E. A. Rickart for critical reviews of earlier versions of this work. N. Sanders and three anonymous reviewers provided detailed suggestions which have greatly improved the clarity and presentation of this work. I also thank J. A. Finarelli for assistance with data analysis and M. Bogan for generously allowing me to use capture records from his 2001 survey in the Markagunt Plateau. This work was funded by a STAR Fellowship from the Environmental Protection Agency (U91613701), the Chicago chapter of the ARCS foundation, the Univ. of Chicago Hinds Fund, an American Society of Mammalogists Grant-in-Aid of Research, and the Theodore Roosevelt Memorial Fund of the American Museum of Natural History. Fieldwork was conducted in accordance with the guidelines of the American Society of Mammalogists and field procedures were certified by the Univ. of Chicago (ACUP # 71190).

References Akaike, H. 1973. Information theory as an extension of the maximum likelihood principle. In: Petrov, B. N. and Csaki, F. (eds), Second International Symposium on Information Theory. Akademiai Kiado, Budapest, Hungary, pp. 267 281. Bailey, S. A. et al. 2004. Primary productivity and species richness: relationships among functional guilds, residency groups and vagility classes at multiple spatial scales. Ecography 27: 207 217. Bellwood, D. R. et al. 2005. Environmental and geometric constraints on Indo-Pacific coral reef biodiversity. Ecol. Lett. 8: 643 651. Bhattarai, K. R. et al. 2004. Fern species richness along a central Himalayan elevational gradient, Nepal. J. Biogeogr. 31: 389 400.

Burnham, K. P. and Anderson, D. R. 2002. Model selection and multi-model inference: a practical information-theoretic approach. Springer. Chase, J. M. and Leibold, M. A. 2002. Spatial scale dictates the productivity-biodiversity relationship. Nature 416: 427 430. Chase, J. M. and Leibold, M. A. 2003. Ecological niches: linking classical and contemporary approaches. The Univ. of Chicago Press. Colwell, R. K. and Hurtt, G. C. 1994. Nonbiological gradients in species richness and a spurious Rapoport effect. Am. Nat. 144: 570 595. Colwell, R. K. and Lees, D. C. 2000. The mid-domain effect: geometric constraints on the geography of species richness. Trends Ecol. Evol. 15: 70 76. Colwell, R. K. et al. 2004. The mid-domain effect and species richness patterns: what have we learned so far? Am. Nat. 163: E1 E23. Connell, J. H. 1961. The influence of interspecific competition and other factors on the distribution of the barnacle Chthamalus stellatus. Ecology 42: 410 423. Connor, E. F. and McCoy, E. D. 1979. The statistics and biology of the species-area relationship. Am. Nat. 113: 791 833. Cramer, W. et al. 1999. Comparing global models of terrestrial net primary productivity (NPP): overview and key results. Global Change Biol. 5: 1 15. Currie, D. J. et al. 2004. Predictions and tests of climate-based hypotheses of broad-scale variation in taxonomic richness. Ecol. Lett. 7: 1121 1134. Davies, R. G. et al. 2007. Topography, energy and the global distribution of bird species richness. Proc. R. Soc. B 274: 1189 1197. Diniz-Filho, J. A. F. et al. 2003. Spatial autocorrelation and red herrings in geographical ecology. Global Ecol. Biogeogr. 12: 53 64. Dunn, R. R. et al. 2007. When does diversity fit null model predictions? Scale and range size mediate the mid-domain effect. Global Ecol. Biogeogr. 16: 305 312. Durrant, S. D. 1952. Mammals of Utah: taxonomy and distribution. Univ. of Kansas Publications, Museum of Natural History. Fu, C. Z. et al. 2006. Elevational patterns of frog species richness and endemic richness in the Hengduan Mountains, China: geometric constraints, area and climate effects. Ecography 29: 919 927. Grytnes, J. A. and Vetaas, O. R. 2002. Species richness and altitude: a comparison between null models and interpolated plant species richness along the Himalayan altitudinal gradient, Nepal. Am. Nat. 159: 294 304. Hall, E. R. 1981. The mammals of North America. Wiley. Harper, K. T. and Reveal, J. L. (eds) 1978. Intermountain biogeography: a symposium. Great Basin Naturalist Memoir no. 2, Brigham Young Univ. Hawkins, B. A. et al. 2003. Energy, water, and broadscale geographic patterns of species richness. Ecology 84: 3105 3117. Heaney, L. R. 2001. Small mammal diversity along elevational gradients in the Philippines: an assessment of patterns and hypotheses. Global Ecol. Biogeogr. 10: 15 39. Herzog, S. K. et al. 2005. The elevational gradient in Andean bird species richness at the local scale: a foothill peak and a highelevation plateau. Ecography 28: 209 222. Hillebrand, H. 2004. On the generality of the latitudinal diversity gradient. Am. Nat. 163: 192 211. Hobbs, N. T. and Hilborn, R. 2006. Alternatives to statistical hypothesis testing in ecology: a guide to self teaching. Ecol. Appl. 16: 5 19.

421

Hortal, J. et al. 2008. Regional and environmental effects on the species richness of mammal assemblages. J. Biogeogr. 35: 1202 1214. Huete, A. R. et al. 1997. A comparison of vegetation indices over a global set of TM Images for EOS-MODIS. Remote Sens. Environ. 59: 440 451. Hurlbert, A. H. and Haskell, J. P. 2003. The effect of energy and seasonality on avian species richness and community composition. Am. Nat. 161: 83 97. Hurvich, C. M. and Tsai, C.-L. 1989. Regression and time series model selection in small samples. Biometrika 76: 297 307. Hutchinson, G. E. 1957. Concluding remarks. Cold Spring Harbor Symp. Quant. Biol. 22: 415 427. Jetz, W. and Rahbek, C. 2001. Geometric constraints explain much of the species richness pattern in African birds. Proc. Nat. Acad. Sci. USA 98: 5661 5666. Kluge, J. et al. 2006. What drives elevational patterns of diversity? A test of geometric constraints, climate and species pool effects for pteridophytes on an elevational gradient in Costa Rica. Global Ecol. Biogeogr. 15: 358 371. Kreft, H. and Jetz, W. 2007. Global patterns and determinants of vascular plant diversity. Proc. Nat. Acad. Sci. USA 104: 5925 5930. Lawton, J. H. 1999. Are there general laws in ecology? Oikos 84: 177 192. Lomolino, M. V. 2001. Elevation gradients of species-density: historical and prospective views. Global Ecol. Biogeogr. 10: 3 13. MacArthur, R. H. and Wilson, E. O. 1967. The theory of island biogeography. Monographs on population biology, no. 1. Princeton Univ. Press. McCain, C. M. 2004. The mid-domain effect applied to elevational gradients: species richness of small mammals in Costa Rica. J. Biogeogr. 31: 19 31. McCain, C. M. 2005. Elevational gradients in diversity of small mammals. Ecology 86: 366 372. McCain, C. M. 2007. Area and mammalian elevational diversity. Ecology 88: 76 86. McClain, C. R. et al. 2007. Challenges in the application of geometric constraint models. Global Ecol. Biogeogr. 16: 257 264. Mittelbach, G. G. et al. 2001. What is the observed relationship between species richness and productivity? Ecology 82: 2381 2396. Oommen, M. A. and Shanker, K. 2005. Elevational species richness patterns emerge from multiple local mechanisms in Himalayan woody plants. Ecology 86: 3039 3047. Preston, F. W. 1962. The canonical distribution of commonness and rarity. Ecology 43: 185 215, 410 432. Rahbek, C. 1997. The relationship among area, elevation, and regional species richness in neotropical birds. Am. Nat. 149: 875 902. Rahbek, C. 2005. The role of spatial scale and the perception of large-scale species-richness patterns. Ecol. Lett. 8: 224 239. Rickart, E. A. 2001. Elevational diversity gradients, biogeography and the structure of montane mammal communities in the intermountain region of North America. Global Ecol. Biogeogr. 10: 77 100.

Download the Supplementary material as file E5538 from / <www.oikos.ekol.lu.se/appendix>.

422

Rickart, E. A. et al. 1991. Distribution and ecology of small mammals along an elevational transect in southeastern Luzon, Philippines. J. Mammal. 72: 458 469. Ricklefs, R. E. 2004. A comprehensive framework for global patterns in biodiversity. Ecol. Lett. 7: 1 15. Rosenzweig, M. L. 1995. Species diversity in space and time. Cambridge Univ. Press. Rowe, R. J. 2005. Elevational gradient analyses and the use of historical museum specimens: a cautionary tale. J. Biogeogr. 32: 1883 1897. Rowe, R. J. and Lidgard, S. 2009. Elevational gradients and species richness: do methods change pattern perception?. Global Ecol. Biogeogr., in press. Royall, R. M. 1997. Statistical evidence: a likelihood paradigm. Chapman and Hall. Sanders, N. J. 2002. Elevational gradients in ant species richness: area, geometry, and Rapoportâ&#x20AC;&#x2122;s rule. Ecography 25: 25 32. Sanders, N. J. et al. 2007. Temperature, but not productivity or geometry, predicts elevational diversity gradients in ants across spatial grains. Global Ecol. Biogeogr. 16: 640 649. Scott, J. M. et al. 1993. Gap analysis a geographic approach to protection of biological diversity. Wildl. Monogr. 123: 1 41. Serrano, L. et al. 2000. Estimation of canopy photosynthetic and nonphotosynthetic components from spectral transmittance. Ecology 81: 3149 3162. Seto, K. C. et al. 2004. Linking spatial patterns of bird and butterfly species richness with Landsat TM derived NDVI. Int. J. Remote Sens. 25: 4309 4324. Shmida, A. and Wilson, M. V. 1985. Biological determinants of species diversity. J. Biogeogr. 12: 1 20. Storch, D. et al. 2006. Energy, range dynamics and global species richness patterns: reconciling mid-domain effects and environmental determinants of avian diversity. Ecol. Lett. 9: 1308 1320. Terborgh, J. 1977. Bird species-diversity on an Andean elevational gradient. Ecology 58: 1007 1019. Turner, M. G. et al. 2001. Landscape ecology in theory and practice: pattern and process. Springer. Vetaas, O. R. and Grytnes, J. A. 2002. Distribution of vascular plant species richness and endemic richness along the Himalayan elevation gradient in Nepal. Global Ecol. Biogeogr. 11: 291 301. Watkins, J. E. et al. 2006. Species richness and distribution of ferns along an elevational gradient in Costa Rica. Am. J. Bot. 93: 73 83. Whittaker, R. J. et al. 2001. Scale and species richness: towards a general, hierarchical theory of species diversity. J. Biogeogr. 28: 453 470. Wieczorek, J. et al. 2004. The point-radius method for georeferencing locality descriptions and calculating associated uncertainty. Int. J. Geogr. Inf. Sci. 18: 745 767. Williams, C. B. 1964. Patterns in the balance of nature. Academic Press. Zapata, F. A. et al. 2005. The mid-domain effect revisited. Am. Nat. 166: E144 E148.

Ecography 31: 306 315, 2008 doi: 10.1111/j.2008.0906-7590.05333.x # 2008 The Authors. Journal compilation # 2008 Ecography Subject Editor: Carsten Rahbek. Accepted 12 February 2008

Why do mountains support so many species of birds? Adriana Ruggiero and Bradford A. Hawkins A. Ruggiero (aruggier@crub.uncoma.edu.ar), Laboratorio Ecotono, Centro Regional Universitario Bariloche, Univ. Nacional del Comahue INIBIOMA-CONICET, Quintral 1250 (8400) Bariloche, Rio Negro, Argentina. B. A. Hawkins, Dept of Ecology and Evolutionary Biology, Univ. of California, Irvine, CA 92697, USA.

Although topographic complexity is often associated with high bird diversity at broad geographic scales, little is known about the relative contributions of geomorphologic heterogeneity and altitudinal climatic gradients found in mountains. We analysed the birds in the western mountains of the New World to examine the two-fold effect of topography on species richness patterns, using two grains at the intercontinental extent and within temperate and tropical latitudes. Birds were also classified as montane or lowland, based on their overall distributions in the hemisphere. We estimated range in temperature within each cell and the standard deviation in elevation (topographic roughness) based on all pixels within each cell. We used path analysis to test for the independent effects of topographic roughness and temperature range on species richness while controlling for the collinearity between topographic variables. At the intercontinental extent, actual evapotranspiration (AET) was the primary driver of species richness patterns of all species taken together and of lowland species considered separately. In contrast, within-cell temperature gradients strongly influenced the richness of montane species. Regional partitioning of the data also suggested that range in temperature either by itself or acting in combination with AET had the strongest ‘‘effect’’ on montane bird species richness everywhere. Topographic roughness had weaker ‘‘effects’’ on richness variation throughout, although its positive relationship with richness increased slightly in the tropics. We conclude that bird diversity gradients in mountains primarily reflect local climatic gradients. Widespread (lowland) species and narrow-ranged (montane) species respond similarly to changes in the environment, differing only in that the richness of lowland species correlates better with broad-scale climatic effects (AET), whereas mesoscale climatic variation accounts for richness patterns of montane species. Thus, latitudinal and altitudinal gradients in species richness can be explained through similar climatic-based processes, as has long been argued.

Biogeographers and ecologists are well aware of the tendency for mountains to harbour large numbers of species. The pattern has been documented for various taxa, especially in tropical regions where mountains usually are biodiversity hotspots due to a combination of evolutionary and ecological forces (e.g. plants and vertebrates: Myers et al. 2000; mammals: Simpson 1964; birds: Rahbek and Graves 2001, de Klerk et al. 2002, Hawkins and DinizFilho 2006, Davies et al. 2007; amphibians: Poynton et al. 2007, Lomolino 2001 for review). A two-fold effect of topography on species diversity is usually recognized. First, topographic complexity ( topographic heterogeneity) is believed to generate local geographic isolation that may promote speciation through allopatry and the drift of populations. Second, as complex topographies usually also define complex climatic gradients, this may also contribute to species richness and distribution patterns, promoting speciation through phenotypic divergence or ecological differentiation of parapatric populations (Orr and Smith 1998, Schluter 2000; see Graves 1985 for Andean birds). A recent resurgence of interest in the relationship between topography and species richness patterns at broad 306

geographic scales (Kerr and Packer 1997, Rahbek and Graves 2000, 2001, Davies et al. 2007, Rahbek et al. 2007) has stimulated debate concerning underlying mechanisms. However, although geographic isolation has been hypothesized to drive speciation and to generate diversification patterns in mountain faunas (e.g. Andean birds: Vuilleumier 1969, Graves 1985, 1988), the relative contributions of geomorphologic heterogeneity and altitudinally generated climate gradients to account for mountain species richness patterns have rarely been evaluated (but see Patton and Smith 1992 for Akodon mice and Garcı´aMoreno and Fjeldsa˚ 2000 for Andean birds). Most often topography has been measured by elevation range (Rahbek and Graves 2000, 2001, Jetz and Rahbek 2002, Hawkins et al. 2005, Davies et al. 2007, Rahbek et al. 2007), or more rarely by the standard deviation in elevation (Ruggiero and Kitzberger 2004). Although both measures are valid, they also implicitly represent different mechanisms. Elevation range is the difference between the highest and lowest elevations occurring within an area (cell), and although it is often claimed to be a measure of topographic heterogeneity, it

is in fact only a measure of overall strength of the gradient within the cell (McCarroll and Nesje 1996). Hence, the use of elevation range actually emphasizes the effect of altitudinal climatic gradients and the associated changes in habitat on species richness and distributions, and McCarroll and Nesje (1996) have shown that the standard deviation of the differences between adjacent elevations is a more meaningful measure of true topographic roughness than is elevation range. To our knowledge, no previous studies have attempted to disentangle these two effects of topography on broadscale patterns in species richness (i.e. the strength of mesoscale climatic gradients vs topographic heterogeneity per se), and thus the possible mechanisms underlying correlations between richness and elevation cannot be distinguished. Here we use the New World birds as model system to address this issue. Our goal is to analyse the richness patterns of birds inhabiting mountains within tropical and temperate latitudes in the New World to quantify the extent that variation in species richness is associated with changes in topographic heterogeneity or with altitudinal climatic gradients. If topographic heterogeneity drives species richness patterns, changes in species richness will be strongly associated with topographic roughness, estimated as the standard deviation of elevation. In contrast, if altitudinal climatic gradients are the main driver of species richness, changes in species richness across cells will correlate most strongly with the altitudinal range in mean annual temperature within cells. Throughout the analysis we use the range in temperature rather than elevation range per se. It is the temperature gradient up a mountain what determines its effectiveness as a barrier rather than its absolute range in elevation. Hence, the altitudinal range in temperature is biologically informative, whereas range in elevation is insensitive to the strong relationship of temperature with latitude, since the strength of the altitudinal climatic gradient in any place is mediated by the regional macroclimate (Janzen 1967, Turner and Hawkins 2004, Hawkins and Diniz-Filho 2006). Throughout the analysis, we distinguish species richness patterns of montane and lowland species (see Methods for detailed definition). Montane species have much narrower geographic distributions than lowland species (Hawkins and Diniz-Filho 2006). Recently, it has been argued that species richness patterns of widespread species are driven mainly by spatial variation of energy-water predictor variables, whereas, because narrow-ranged species are usually found in mountainous regions, topographic variability is a better predictor of their species richness (e.g. birds: Jetz and Rahbek 2002, Rahbek et al. 2007; mammals: Ruggiero and Kitzberger 2004). However, none of these studies compared the effects of topographic heterogeneity and altitudinal climatic gradients associated with elevation. Thus, we still do not know the extent to which widespread and narrow ranged species that coexist in regions of high topographic relief differ in their response to environmental structure.

Methods Area of study The study encompasses the mountain ranges that extend along western North and South America from Alaska to Tierra del Fuego (Fig. 1). We used GTOPO30 global DEM (USGS 1999) with a horizontal grid spacing of 30 arc-second to define our area study. We identified visually all mountain ranges running along the western portion of the New World (i.e. mainly the cordillera of the northwestern United States and southwestern Canada, the Rocky Mountains, the Sierra Madre Occidental, Oriental, del Sur and Eje Volca´nico Transversal in North America and the Andes in South America). We also included a buffer zone of ca 200 km around the mountains to ensure that we have included the full local altitudinal gradients in temperature. We excluded some isolated parts of the mountain ranges in North America e.g. extreme northern Alaska and Baja California to avoid confounding biogeographic effects that could complicate interpretation. Bird data We used the digitized breeding range maps of native bird species available in Ridgely et al. (2003). We classified species into two groups based on their geographic distributions: montane (n 1241) and lowland (n 1901). Montane species were defined as having 50% of their geographic distribution overlapping the mountain region. Actually, the vast majority of these species are narrowly distributed endemics restricted to the western mountain ranges in North and South America. However, because some of them also inhabit mountains outside the area included in our analysis (e.g. the Tepuis in Venezuela and the Sierras Pampeanas in Argentina) we do not use the term ‘‘endemic’’ to refer to this group. Lowland species were defined as those whose geographic distributions extend mostly in lowland areas although they may also inhabit some mountain regions. Species whose ranges did not overlap at all with the colored area illustrated in Fig. 1 were not included in the data set. Estimation of species richness The range maps were rasterized in ArcGIS 9.2 with respect to their distribution within the study area. Rasters were generated at two grain sizes using separate continental Lambert azimuthal equal-area grids of 50 50 km and 100 100 km, resulting respectively in 5952 and 1535 cells for statistical analysis. Coastal cells that included B50% of land surface were excluded. Species richness at each grain was estimated directly from the rasters. We created richness data and maps for all birds, montane species only, lowland species only, and the proportion of mountain species in each cell, at both grains.

307

Figure 1. Elevation map of the western New World. The colored region corresponds to the area included in the analysis.

Environmental predictors We generated datasets of environmental variables at each grain using consistent procedures in ArcGIS. We used 308

GTOPO30 global DEM (USGS 1999) to derive a measure of topographic roughness, estimated as the standard deviation in elevation based on all pixels within each cell (ca 2500 at the smaller grain). We did not use the standard

deviation of the differences in elevation among adjacent cells, as suggested by McCarroll and Nesje (1996), because we found a near perfect correlation (r 0.996) between the two measures based on the simulated data presented in their Table I (p. 966). Thus, we use standard deviation in elevation to ease calculation and interpretation. Range in temperature in each cell was estimated from the mean annual temperature layer with a spatial resolution of 1 km2 developed by Hijmans et al. (2005), available in the WorldClim database Bhttp://www.worldclim.org/current. htm . Of course, the temperature database is almost entirely interpolated, but this is irrelevant to our study as the temperatures in the database are estimated using very well established relationships between regional temperatures (themselves well known) and altitude. This allows us to estimate the overall temperature gradient in a cell after correcting for its latitudinal position. We do caution, however, about limitations inherent to this kind of climate surface. Mainly due to the low number of climatic stations and their frequently biased location in mountain areas (e.g. mainly at low elevations in high subtropical latitudes and at higher altitudes in the tropics) there is considerable uncertainty about local climatic values estimated in mountain areas (Hijmans et al. 2005). A consequence is that this climatic database makes it impossible to detect some local phenomena acting within mountain areas such as, for instance, climatic differences between humid slopes and warm/dry intermountain basins. At present, better resolution data do not exist for the New World, and we do not know the extent to which limitations in the data influence the results. Hence, the diversity-environment relationships we find should not be considered error free, and the patterns will need to be verified if higher quality, small scale databases become available in the future. On the other hand, such data may never become available. Because water-energy variables usually have strong relationships with bird species richness patterns at continental and intercontinental extents (Jetz and Rahbek 2002, Hawkins et al. 2005, 2007, Davies et al. 2007, Rahbek et al. 2007) we included annual potential evapotranspiration (AET) as a control variable in all our analyses. We estimated the maximum value of AET for each cell from the 30 30’ actual evapotranspiration dataset developed by Ahn and Tateishi (1994), available at Bhttp:// gcmd.gsfc.nasa.gov/ . Statistical analyses We generated a path model to summarize in a unique causal scheme the relationships among variables derived from a priori predictions. Our model proposes that the standard deviation in elevation and the altitudinal range in mean annual temperature are exogenous variables that may have direct effects on species richness. However, given that previous analyses have suggested that an interaction between macroclimate and topographic effects may ultimately drive the variation of species richness from the equator to the poles (Rahbek and Graves 2001), we also tested for indirect effects of topography on richness mediated through the effect of actual evapotranspiration (AET). We also assume

that the exogenous variables may be correlated with each other, as cells containing very rough mountains will have wider ranges of temperature than flatter cells. We standardized all variables for the analyses. We used path analysis (PROC CALIS in SAS 9.1) to test the hypothesized causal relationships. These analyses used the maximum likelihood method of parameter estimation (Hatcher 1994). We tested the path model first for all bird species, and then divided the data into montane, lowland and the proportion of mountain species in a cell. This allowed us to determine if the ‘‘effects’’ of topography on species richness differ between widespread (lowland) and narrowly ranged (montane) species. We also divided the data set spatially into four regional subsets: the part of North America covered by ice during the most recent Ice Age (Dyke et al. 2003), the unglaciated, extratropical part of North America, the tropics, and extratropical South America. We then reran the model for each species group separately for each region. Because of the greater disruptive effects of Pleistocene glaciation in northern temperate zone (Dynesius and Jansson 2000) we might expect the effects of topographic heterogeneity of mountains on bird species richness to be negatively correlated with latitude. Dividing the data into regions allowed us to control for this historical effect and evaluate if the relative magnitudes of topographic effects vs temperature effects have been influenced by Pleistocene climate change patterns. Analyses were conducted at both 50 50 km and 100 100 km grain, but the models were virtually identical. To reduce redundancy we report here only the results for the smaller gain.

Results Spatial patterns in species richness Species richness shows strong spatial structure (Fig. 2). The north-central Andes near the equator and the central Andes near 158S have the highest bird species richness (Fig. 2a). And as expected, montane and lowland species have strikingly different richness patterns within the Neotropics, the former being concentrated along the central Andes (Fig. 2b) and the later dominating lower altitudes to the east of the mountains and the eastern slope (Fig. 2c). Proportionally, montane species are more prevalent along the entirety of the Andes, representing more than half of all bird species in the tropics and subtropics of South America (Fig. 2d). In North America, there is also relatively high montane species richness in the transition zone between the Neotropical and Nearctic biogeographic regions, in clear association with the Sierra Madre Oriental, Occidental and Eje Volca´nico Transversal in Me´xico (Fig. 2b, d). The highly structured pattern in the richness of montane birds is clearest when the proportion of mountain species to total species is considered (Fig. 2c). There are very few montane birds in the northern Nearctic, and most birds in the northern Rockies are also widely distributed in lowlands (Fig. 2b, c).

309

Figure 2. Species richness patterns for (a) all birds, (b) lowland species, (c) montane species and (d) the proportion of montane species.

Relationships with topography at the hemispherical extent Across all data, the richness pattern of all species has a strong relationship with AET (Fig. 3a), reflecting the dominant influence of climate on overall diversity gradients at very broad extents. After controlling for this effect in the path model, the effect of topographic heterogeneity is twice as strong as the effect of the temperature gradients within cells, as indicated by the direct path coefficients (Fig. 3a).

However, this belies the fact that the model for overall richness comprises strongly contrasting relationships between macroclimate, topographic heterogeneity, and mesoscale climatic gradients for the lowland and montane components of bird faunas (cf. Fig. 3b, c). Thus, in both components the effect of the temperature gradient is actually much stronger than the effect of heterogeneity, but the relationships are in opposite directions. The path model for lowland species indicates that the richness of these birds increases strongly with increasing AET, decreases b) Lowland species (R2 = 0.67)

a) All species (R2 = 0.71)

SD Alt

0.28

0.25

-0.31

-0.31 0.92

AET

0.83

Richness

0.92

AET

Richness

0.40

0.40 Range in temp.

0.83

-0.12

-0.41

Range in temp.

c) Montane species (R2 = 0.56)

d) Proportion of montane spp. (R2 = 0.52) SD Alt

SD Alt

-0.35

0.01 -0.31 0.92

-0.31 AET

0.31

Richness

0.92

AET

Proportion

0.40

0.40 Range in temp.

-0.08

0.63

Range in temp.

1.03

Figure 3. Path models proposed to test for the effects of topography on the species richness variation of (a) all birds, (b) lowland birds, (c) montane birds and (d) the proportion of montane species. The standard deviation of elevation (SD Alt) and the altitudinal range in mean annual temperature (Range in temp.) are exogenous variables hypothesized to have direct effects on species richness. The model also tested for indirect effects of these variables through the effect of actual evapotranspiration (AET). Solid lines indicate positive effects, and dashed lines are negative effects. The curved line between SD Alt and Range in temp. indicates non-causal covariation between these variables.

310

moderately with stronger mesoscale temperature gradients (range in temperature), and increases less strongly with increasing topographic heterogeneity (the standard deviation of altitude), whereas montane bird richness increases most strongly with an increasing range of temperature in cells, increases less strongly in warmer and wetter macroclimates, and is independent of topographic heterogeneity. The very strong influence of mesoscale temperature gradients on the richness of bird faunas is most apparent in the path model for the proportion of montane species in each cell (Fig. 3d). It is important to note the strong collinearity between topographic heterogeneity and range in temperature in the New World (Fig. 3), which although controlled statistically in the path model can influence interpretation of the coefficients. Thus, we performed two further analyses to explore in more depth the effects of the interactions among the variables and to evaluate the extent to which collinearity influences our interpretation. First, we used partial regression to partition the source of species richness variation into four components, representing: 1) independent effects of topographic heterogeneity, 2) shared effects of heterogeneity and altitudinal temperature range, 3) independent effects of altitudinal temperature range, and 4) unexplained variation. This confirmed that ca one third of the variation in species richness (montane species: 38%; the proportion of montane species: 34%) was explained by the overlapping effects of topographic roughness and altitudinal temperature range. On the other hand, the independent effect of altitudinal temperature range was substantially stronger (montane species richness: 8.9%, proportional richness: 16%) than the independent effect of topographic heterogeneity (montane species richness: 0.1%, proportional richness: 1.7%). The analyses for all species taken together and the lowland species were not informative about the independent effects of

the topographic variables because richness in these two data sets is explained primarily by macroclimate (e.g. the independent effect of AET represents 94% of the variation in total bird richness explained by the path model). In the second analysis, we divided each path model into two submodels, each containing only one topographic variable (the topographic roughness submodel [Topo] and the temperature range submodel [Temp]). The coefficients based on Topo and Temp were qualitatively very similar to those in the models including both variables, although they differed slightly in the proportion of variance in richness explained (R2). Actual evapotranspiration continued to be the main driver of species richness in all species (Topo: bAET 0.82, R2 0.71; Temp: bAET 0.82, R2 0.70), and lowland species (Topo: bAET 0.80, R2 0.64; Temp: bAET 0.81, R2 0.66), whereas the effects of topographic roughness and altitudinal range in temperature were weak in all cases (between 0.15 and 0.14). For montane species the effect of altitudinal range in temperature on the richness of mountain species was greater than the effect of topographic roughness (Temp: bte 0.67, R2 0.56; Topo:, bto 0.59, R2 0.50), as was also the case for the proportion of mountain species (Temp: bte 0.71, R2 0.50; Topo: bto 0.60, R2 0.36). Subregional models The regional division of data to control for the potential influence of Pleistocene climate change gave the same qualitative results as those obtained at the intercontinental scale (Table 1). Notably, range in temperature, either by itself (the direct path coefficient) or acting in combination with AET (the total path coefficient), had the strongest effect on richness in 10 of the 16 path models, the direct effect of AET was strongest in 3 models (involving either all

Table 1. Direct (Dir), indirect (Ind) and total (Dir Ind) effects of temperature range (Temp), topographic roughness (Topo) and AET on the species richness of all birds, montane species, lowland species and the proportion of montane species, in four latitudinally defined parts of the New World. See Fig. 3 for the structure of the path models. Strongest effects in each model are highlighted in bold. Direct causal covariation (Dir) Temp

Indirect causal covariation (Ind)

Total effects (Dir Ind)

Model R2

Topo

AET

Temp

Topo

Temp

Topo

0.23 0.17 0.20 0.01

0.70 0.44 0.64 0.42

0.44 0.28 0.40 0.26

0.46 0.29 0.42 0.27

0.76 1.05 0.46 0.91

0.69 0.46 0.62 0.28

0.52 0.57 0.45 0.60

Extratopical North America non-glaciated All 0.44 0.18 Montane 0.70 0 Lowland 0.01 0.29 Proportion 0.69 0.06

0.08 0.07 0.06 0.11

0.04 0.03 0.03 0.05

0.01 0.01 0.01 0.02

0.40 0.73 0.04 0.74

0.19 0.01 0.30 0.08

0.35 0.46 0.07 0.46

Tropics All Montane Lowland Proportion

North America glaciated All 0.32 Montane 0.77 Lowland 0.06 Proportion 0.65

0.32 0.64 0.62 0.87

0.46 0.16 0.40 0.24

0.76 0.12 0.73 0.41

0.04 0 0.04 0.03

0.18 0.03 0.18 0.11

0.27 0.27 0.58 0.84

0.64 0.13 0.22 0.13

0.57 0.60 0.66 0.71

Extratropical South America All 0.49 Montane 0.73 Lowland 0.68 Proportion 1.05

0.34 0 0.33 0.30

0.65 0.33 0.54 0.03

0.11 0.06 0.09 0

0.01 0 0.01 0

0.60 0.67 0.77 1.05

0.35 0 0.34 0.30

0.50 0.58 0.51 0.60

311

species or lowland species), and topographic heterogeneity had the strongest effect in 1 (lowland species in central North America, although this model had a very low coefficient of determination [Table 1]). Thus, although the details of the models vary regionally, as a whole mesoscale temperature gradients have much stronger relationships with bird richness patterns than topographic heterogeneity per se, and the climatic history of the different latitudinal spans within the New World does not alter that result substantially.

Discussion We confirm the extent to which climatic gradients probably underlie the effect of topography on broad-scale species richness patterns. Our results are strongly consistent with the original proposition of Janzen (1967) that the association between elevation range, climatic diversity and the isolation of populations may promote the increase of species richness in regions of high topographic relief. On the other hand, local topographic roughness, as measured by the standard deviation in elevation, had smaller effects on richness variation. Thus, at the scale of resolution of the present analysis, topographic heterogeneity per se can not be considered an important predictor of broad-scale bird species richness patterns in the mountains of the New World. Janzen (1967) argued cogently that it is the temperature gradient across a mountain that determines its effectiveness as a barrier rather than its absolute range in elevation. He proposed that ‘‘mountain passes are higher in the tropics’’ because a lowland tropical species that attempts to cross a mountain pass will face more extreme environmental conditions than it normally experiences, whereas this is less likely to be true for a lowland temperate species. Although Janzen (1967: 244) explained clearly that he was not offering an explanation for tropical diversity, his idea makes it reasonable to predict that the narrower physiological tolerance of tropical organisms and the lack of overlap in the thermal regimes over tropical altitudinal gradients set up conditions that favor reduced dispersal and overlap in species’ distributions across elevation and high rates of allopatric speciation, hence promoting the coexistence of a higher number of species within the tropics (see Ghalambor et al. 2006 for detailed discussion and examples). Also along these lines, recent analyses of bird species richness patterns at continental to global scales have reinforced the hypothesis that latitudinal variation in terrestrial species richness is influenced by a interaction between climate and topography (Rahbek and Graves 2001, Davies et al. 2007, Hawkins et al. 2007). An important issue with respect to our results and those of other studies that include topographic variables is identifying possible causal explanations underlying correlations between species richness and range in elevation. Range is elevation has been interpreted to measure, among other things, ‘‘topographic heterogeneity’’ (Rahbek and Graves 2001, Jetz and Rahbek 2002), ‘‘habitat heterogeneity’’ (Kerr and Packer 1997, Kreft and Jetz 2007, Davies et al. 2007), ‘‘habitat diversity’’ (Kalmar and Currie 2006), and

312

‘‘climatic variability’’ (O’Brien et al. 2000), and topographic relief (interpreted as a proxy for habitat heterogeneity) is often considered an explanation for diversity gradients alternative to those proposing that climate directly or indirectly drives diversity (Kerr and Packer 1997, Hawkins et al. 2003, Davies et al. 2007, Rahbek et al. 2007). However, our results suggest that range in elevation is not a different hypothesis from climate, but is the same hypothesis operating at smaller spatial scales (see also O’Brien et al. [2000] and Hawkins and Diniz-Filho [2006]). Bird diversity gradients in mountains appear to be responding most strongly to local climatic gradients, and the only real difference between lowlands and mountains is that in the former diversity varies gradually due to broad, shallow climate gradients, whereas in the latter diversity varies sharply in response to strong climate gradients across altitudes. This is also consistent with the right-skewed range size frequency distributions and the predominance of small median range size found in the Andes that contrast with the left-skewed distributions and predominance of wide-ranging species found over much of South America (Graves and Rahbek 2005). And although range in elevation almost certainly correlates with habitat diversity, this only reflects that local climates influence vegetation structure, which then influences animal distributions. This is conceptually identical to the version of the energy hypothesis that assumes that climate drives broad-scale diversity via indirect effects on vegetation (the productivity hypothesis). Thus, both the productivity hypothesis and the habitat heterogeneity hypothesis assume that climate drives vegetation, which then drives animal diversity. Therefore, correlations of richness with range in elevation vs. broad-scale climate variables (rainfall, temperature, AET, etc.) are not necessarily alternative explanations for diversity gradients; rather they indicate that explanations must be sensitive to the spatial scale of the pattern being studied. This is also related to a recent trend to compare diversity patterns based on the range sizes of constituent species. Various workers have emphasized the influence of geographic range on richness environment relationships (Jetz and Rahbek 2002, Ruggiero and Kitzberger 2004, Kreft et al. 2006, Rahbek et al. 2007), leading to the belief that climate models explain richness patterns of widespread species but not small-ranged species. In general, the high richness of bird species with restricted geographic distributions found in the tropical mountains of South America has been associated with factors assumed to be decoupled from measures of contemporary climate at large scales (see e.g. Rahbek et al. 2007 for discussion). We do not discount the role of such historical factors (see next paragraph), but it is important to realize that contemporary climate in fact has strong empirical relationships with species richness patterns of small-ranged, montane birds. The difference between large- and small-ranged species is that correlations for the former are strongest for climate variables measured at broad scales (AET, Fig. 3b), whereas for the latter finer grained climate variables correlate more strongly (e.g. range in temperature, Fig. 3c). Although we find that range in temperature is more strongly associated with bird richness than topographic

heterogeneity throughout the New World, dividing the hemisphere into regions based on Pleistocene climate history suggests that deep-rooted historical effects are also operating. Notably, most of the path coefficients for the total effects of heterogeneity shifted from negative to positive moving from glaciated North America southward, which probably reflects a signal of speciation by isolation of small populations in places that remained stable during the Pleistocene (Fjeldsa˚ 1994, Fjeldsa˚ et al. 1999). At local scales within tropical Andean regions, the diversity of montane, endemic species peaks in places that are climatically more stable than adjacent slopes supporting many widespread, lowland species; these stable sites have predictable cloudiness and mist formation and offer topographic protection against polar winds that occasionally freeze the region (Fjeldsa˚ et al. 1999). It has also been suggested that these stable sites may have acted as minirefuges for the persistence of species as well as centers of evolution of new species, most likely during Pleistocene glacial periods (Fjeldsa˚ et al. 1999). High topographic roughness may be associated with these sites, because they are often found in places where mountain chains make sharp bends, and where rain-capturing ridges are adjacent to warm, dry mountain basins (Fjeldsa˚ and Rahbek 2006). It should be noted that these effects are not independent of climate, but they should not necessarily be correlated with the contemporary range of temperature found in an area, thus providing a statistically ‘‘independent’’ contribution of heterogeneity to richness. Actually, both birds (Graves 1988, Garcı´a-Moreno and Fjeldsa˚ 2000) and mammals (Patton and Smith 1992) suggest that speciation in tropical mountains has occurred mainly by population fragmentation and horizontal differentiation within altitudinal zones; therefore, the strong positive effect of altitudinal temperature range on the richness of montane bird species may be the signal of a secondary process of redistribution and stacking of species in different altitudinal belts driven by climatic gradients per se. This involves the well known pattern of range linearity and the altitudinal replacement of congenerics, mainly along the eastern Andean slopes of tropical South America, which has been considered the result of the adaptation of birds to local climatic conditions, interespecific competition, and the altitudinal zonation of habitats (for examples and general discussion, see Terborgh and Weske 1975, Terborgh 1985, Graves 1988, Fjeldsa˚ 1994, Patterson et al. 1998). We also find a commonly reported pattern that the richness of widespread (i.e. lowland) species more strongly defines the overall richness gradient than the richness of small-ranged (i.e. montane) species (Jetz and Rahbek 2002, Lennon et al. 2004). The correlation between total richness and lowland species is much stronger than the correlation between total richness and montane species (r 0.933 and r 0.522, respectively). Consequently, the richness patterns of both lowland species and all species taken together tend to be explained by similar models in which a broad-scale climate variable (AET) dominates. Montane species, on the other hand, generate a very different path model, but the distinction between the effects of altitudinal temperature range and topographic roughness allow us to identify

the climatic component that underlies the effects of topography most often invoked to account for the richness patterns in narrow-ranged species (Jetz and Rahbek 2002, Davies et al. 2007, Rahbek et al. 2007). Despite an inherently strong collinearity between the two topography variables, it is possible to partition the greater independent positive influence of contemporary climatic effects on the richness of mountain species compared to the independent effects of topographic roughness. To what extent can the species richness-topography relationships found for birds can be generalized to other taxa? New World birds have been used to develop much of the theory concerned with the origin and maintenance of biological diversity at large scales. However, not all taxonomic groups have the same altitudinal patterns (see Rahbek 1995, McCain 2005, 2007 for review). In the tropical Andes, the elevational zonation and altitudinal richness patterns of birds may represent an intermediate situation between the patterns found, for instance, in bats and rodents (Patterson et al. 1998). Bats show a strong monotonic altitudinal decrease in species richness in humid tropical mountains in the New World that is strongly correlated with temperature (McCain 2007). Bats inhabiting tropical Andean mountains also differ from birds by showing little evidence of discrete zonation and the lack of a highland endemic fauna (Patterson et al. 1998). In contrast, the species richness of non-volant mammals tends to peak at intermediate elevations (McCain 2005). In the case of rodents inhabiting the Eastern Versant of the Peruvian Andes, the pattern coincides with the altitudinal replacement of three distinct endemic faunas at low, high and intermediate elevations that result in no steady decrease in species richness with altitude and non-nested elevational richness patterns (Patterson et al. 1998). Additional studies on mammals are needed to determine if their different distributional patterns may produce a stronger signal of topographic roughness on richness patterns or whether results we find for birds can be generalized to other endotherms. In sum, we find that mesoscale climate gradients moving up mountain slopes are strong predictors of species richness variation and thus constitute an important part of the explanation of why mountains support so many birds. If this is true, this allows us to reunite explanations for latitudinal and altitudinal gradients in species richness using a climatic template. Both of these gradients have long been thought to represent homologous patterns, although recently there has been a tendency to treat them as being independently generated. For birds at least, they both appear to be manifestations of very similar processes, albeit operating over different spatial scales.

Acknowledgements We appreciate the useful comments of C. Rahbek and an anonymous reviewer. This project was developed during the visit of A.R. to the Dept of Ecology and Evolutionary Biology at Univ. of California, Irvine. A.R. thanks the people at UCI for their hospitality, and especially the Fulbright Commission for financial support. A.R. also receives support from CONICET,

313

ANPCyT (PICT 38148 BID 1728/OC-AR) and Universidad Nacional del Comahue.

References Ahn, C. H. and Tateishi, R. 1994. Development of a global 30minute grid potential evapotranspiration data set. Photogram. Rem. Sens. 33: 12 21. Davies, G. D. et al. 2007. Topography, energy and the global distribution of bird species richness. Proc. R. Soc. B 274: 1189 1197. de Klerk, H. M. et al. 2002. Patterns of species richness and narrow endemism of terrestrial bird species in the Afrotropical region. J. Zool. (Lond.) 256: 327 342. Dyke, A. S. et al. 2003. Deglaciation of North America. Geological Survey of Canada Open File 1574. Dynesius, M. and Jansson, R. 2000. Evolutionary consequences of changes in species’ geographical distributions driven by Milankovitch climate oscillations. Proc. Nat. Acad. Sci. USA 97: 9115 9120. Fjeldsa˚, J. 1994. Geographical patterns for relict and young species of birds in Africa and South America and implications for conservation priorities. Biodiv. Conserv. 3: 207 226. Fjeldsa˚, J. and Rahbek, C. 2006. Diversification of tanagers, a species rich bird group, from lowlands to montane regions of South America. Integr. Comp. Biol. 46: 72 81. Fjeldsa˚, J. et al. 1999. Correlation between endemism and local ecoclimatic stability documented by comparing Andean bird distributions and remotely sensed land surface data. Ecography 22: 63 78. Garcı´a-Moreno, J. and Fjeldsa˚, J. 2000 Chronology and mode of speciation in the Andean avifauna. In: Rheinwald, G. (ed.), Isolated vertebrate communities in the tropics. Bonn. Zool. Monogr. 46, pp. 25 46. Ghalambor, C. K. et al. 2006. Are mountain passes higher in the tropics? Janzen’s hypothesis revisited. Integr. Comp. Biol. 46: 5 17. Graves, G. R. 1985. Elevational correlates of speciation and intraspecific geographic variation in plumage in Andean forest birds. Auk 102: 556 579. Graves, G. R. 1988. Linearity of geographic range and its possible effect on the population structure of Andean birds. Auk 105: 47 52. Graves, G. R. and Rahbek, C. 2005. Source pool geometry and the assembly of continental avifaunas. Proc. Nat. Acad. Sci. USA 102: 7871 7876. Hatcher, L. 1994. A step by step approach to using SAS system for factor analysis and structural equation modeling. SAS Inst., Cary, NC. Hawkins, B. A. and Diniz-Filho, J. A. F. 2006. Beyond Rapoport’s rule: evaluating range size patterns of New World birds in a two-dimensional framework. Global Ecol. Biogeogr. 15: 461 469. Hawkins, B. A. et al. 2003. Productivity and history as predictors of the latitudinal diversity gradient of terrestrial birds. Ecology 84: 1608 1623. Hawkins, B. A. et al. 2005. Water links the historical and contemporary components of the Australian bird diversity gradient. J. Biogeogr. 32: 1035 1042. Hawkins, B. A. et al. 2007. Climate, niche conservatism, and the global bird diversity gradient. Am. Nat. 170: S16 S27. Hijmans, R. J. et al. 2005. Very high resolution interpolated climate surfaces for global land areas. Int. J. Climatol. 25: 1965 1978.

314

Janzen, D. H. 1967. Why mountain passes are higher in the tropics. Am. Nat. 101: 233 249. Jetz, W. and Rahbek, C. 2002. Geographic range size and determinants of avian species richness. Science 297: 1548 1551. Kalmar, A. and Currie, D. J. 2006. A global model of island biogeography. Global Ecol. Biogeogr. 15: 72 81. Kerr, J. T. and Packer, L. 1997. Habitat heterogeneity as a determinant of mammal species richness in high-energy regions. Nature 385: 252 254. Kreft, H. and Jetz, W. 2007. Global patterns and determinants of vascular plant diversity. Proc. Nat. Acad. Sci. USA 104: 5925 5930. Kreft, H. et al. 2006. The significance of geographic range size for spatial diversity patterns in Neotropical palms. Ecography 29: 21 30. Lennon, J. J. et al. 2004. Contribution of rarity and commonness to patterns of species richness. Ecol. Lett. 7: 81 87. Lomolino, M. V. 2001. Elevation gradients of species-density: historical and prospective views. Global Ecol. Biogeogr. 10: 3 13. McCain, C. 2005. Elevational gradients in diversity of small mammals. Ecology 86: 366 372. McCain, C. 2007. Could temperature and water availability drive elevational species richness patterns? A global case study for bats. Global Ecol. Biogeogr. 16: 1 13. McCarroll, D. and Nesje, A. 1996. Rock surface roughness as an indicator of degree of rock surface weathering. Earth Surface Processes and Landforms 21: 963 977. Myers, N. et al. 2000. Biodiversity hotspots for conservation priorities. Nature 403: 853 858. O’Brien, E. M. et al. 2000. Climatic gradients in woody plant (tree and shrub) diversity: water-energy dynamics, residual variation, and topography. Oikos 89: 588 600. Orr, M. R. and Smith, T. B. 1998. Ecology and speciation. Trends Ecol. Evol. 13: 502 506. Patterson, B. D. et al. 1998. Contrasting patterns of elevational zonation of birds and mammals in the Andes of southeastern Peru. J. Biogeogr. 25: 593 607. Patton, J. L. and Smith, M. F. 1992. mtDNA phylogeny of Andean mice. A test of diversification across ecological gradients. Evolution 46: 174 183. Poynton, J. C. et al. 2007. Amphibian diversity in East African biodiversity hotspots: altitudinal and latitudinal patterns. Biodiv. Conserv. 16: 1103 1118. Rahbek, C. 1995. The elevational gradient of species richness: a uniform pattern? Ecography 18: 200 205. Rahbek, C. and Graves, G. R. 2000. Detection of macroecological patterns in South American hummingbirds is affected by spatial scale. Proc. R. Soc. B 267: 2259 2265. Rahbek, C. and Graves, G. R. 2001. Multiscale assessment of patterns of avian species richness. Proc. Nat. Acad. Sci. USA 98: 4534 4539. Rahbek, C. et al. 2007. Predicting continental-scale patterns of bird species richness with spatially explicit models. Proc. R. Soc. B 274: 165 174. Ridgely, R. S. et al. 2003. Digital distribution maps of the birds of the western hemisphere, ver. 1.0. NatureServe, Bhttp:// www.natureserve.org . Ruggiero, A. and Kitzberger, T. 2004. Environmental correlates of mammal species richness in South America: effects of spatial structure, taxonomy and geographic range. Ecography 27: 401 416. Schluter, D. 2000. The ecology of adaptive radiation. Oxford Univ. Press.

Simpson, G. G. 1964. Species density of North American recent mammals. Syst. Zool. 13: 57 73. Terborgh, J. 1985. The role of ecotones in the distribution of Andean birds. Ecology 66: 1237 1246. Terborgh, J. and Weske, J. S. 1975. The role of competition in the distribution of Andean birds. Ecology 56: 562 576. Turner, J. R. G. and Hawkins, B. A. 2004. The global diversity gradient. In: Lomolino, M. V. and Hearny, L. R. (eds),

Frontiers of biogeography: new directions in the geography of nature. Sinauer, pp. 171 190. USGS 1999. GTOPO30 Documentation. Bhttp://edcwww. cr.usgs.gov/landdaac/gtopo30/README.html . Vuilleumier, F. 1969. Pleistocene speciation in birds living in the high Andes. Nature 223: 1179 1180.

315

ECOGRAPHY 27: 401 /416, 2004

Environmental correlates of mammal species richness in South America: effects of spatial structure, taxonomy and geographic range Adriana Ruggiero and Thomas Kitzberger

Ruggiero, A. and Kitzberger, T. 2004. Environmental correlates of mammal species richness in South America: effects of spatial structure, taxonomy and geographic range. / Ecography 27: 401 /416. Although some consensus exists regarding the positive synergism between energy and heterogeneity in increasing species diversity, the role of environmental variability remains controversial. We examine how these factors interact to explain spatial variation in mammal species richness in South America. After taking into account the effects of spatial autocorrelation and area, elevation variability and energy mainly drive spatial variation in mammal species richness. The effect of environmental variability is less important. When different taxonomic groups of mammals are analyzed separately, three ways emerge whereby energy and heterogeneity interact to promote species richness. Heterogeneity may have no effect on species richness, habitat heterogeneity and energy availability contribute independently to species richness, or heterogeneity increases in importance with an increase in energy availability. The partition of species into range size quartiles shows that habitat heterogeneity and temporal instability in the resource supply account for the species richness pattern in the narrowest- ranging species. Habitat heterogeneity is significant also for intermediate ranging species but not for the widest-ranging species. Energy alone drives the species richness pattern in the latter species. The interplay between ecology and biogeographic history may ultimately explain these differences given that narrow- and wide-ranging species show distinct biogeographic patterns, and different taxonomic groups also unequally represent them. A. Ruggiero, (aruggier@crub.uncoma.edu.ar) Laboratorio Ecotono, Centro Regional Universitario Bariloche, Universidad Nacional del Comahue-CONICET, Quintral 1250, (8400) Bariloche, Rı´o Negro, Argentina.

Recent years have brought some consensus regarding the primary roles of hypotheses based on climate and energy (e.g. species-energy hypothesis: Wright 1983, waterenergy dynamic theory: O’Brien 1993, 1998, O’Brien et al. 2000) and historical contingency (i.e., glaciation effects, differences in dispersal and speciation rates) in explaining much of the variance in species richness from the equator to the poles (Whittaker et al. 2001, see also Gaston 2000, Kerr 2001, Hawkins et al. 2003). The area of biomes (Rosenzweig 1992, 1995, Blackburn and Gaston 1997), topographic heterogeneity (Simpson 1964, Kerr and Packer 1997, Rahbek and Graves 2001, Jetz and Rahbek 2002), levels of precipitation (Rahbek

and Graves 2001, Hawkins et al. 2003), and several measures of the availability of energy in local environments, such as mean and maximum temperature, net primary productivity, actual and potential evapotransporation, and solar radiation received per unit area (see e.g. Currie 1991, Blackburn and Gaston 1996a, Kerr and Packer 1997, Kerr and Currie 1999, Currie et al. 1999, Van Rensburg et al. 2002, Hawkins et al. 2003) appear to be the best candidates to explain latitudinal variation in species diversity in both hemispheres (see also Pianka 1966, Begon et al. 1986, Rohde 1992, Ricklefs and Schluter 1993, Brown 1995, Rosenzweig 1995, Currie et al. 1999, Gaston 2000, Willig 2000, Gaston and

Accepted 12 January 2004 Copyright # ECOGRAPHY 2004 ISSN 0906-7590 ECOGRAPHY 27:4 (2004)

401

Blackburn 2000, Kerr 2001, Whittaker et al. 2001, and chapters in the ‘‘species richness’’ section in Blackburn and Gaston 2003 for reviews). Despite the obvious importance of such conceptual progress, some issues still await comprehensive understanding. One issue is related to the kind of spatial heterogeneity that is inherent to ecological systems. Large-scale physical processes occurring over the earth surface, history and ecological interactions among species can lead the environmental variables, the species richness and their interactions to have complex spatial structures; as a consequence, environmental variables and species richness are very often spatially autocorrelated, with this implying that two sites located near one another are unlikely to be independent from each other (e.g. Legendre 1993, Boone and Krohn 2000, Lichstein et al. 2002, Van Rensburg et al. 2002, DinizFilho et al. 2003). This complicates standard statistical testing of hypotheses because it can create false positive / or negative / results (see e.g. Legendre 1993, Diniz-Filho et al. 2003). Moreover, multi-scale relationships between species richness and environmental variables are increasingly well recognized to affect gradients of species diversity, which complicate further ecological interpretation (e.g. Lyons and Willig 1999, Rahbek and Graves 2001, Arita and Rodrı´guez 2002, Blackburn and Gaston 2002, Jetz and Rahbek 2002, Diniz-Filho et al. 2003). On the other hand, dominant factors associated with richness may shift with latitude at continental and global scales (e.g. Kerr and Packer 1997, Rahbek and Graves 2001, Hawkins et al. 2003). However, they may also differentially affect species with different body sizes, belonging to different trophic or spatial guilds (e.g. Andrews and O’Brien 2000), having different dispersal capacities (Kerr and Currie 1999), geographic range sizes (e.g. Jetz and Rahbek 2002) or evolutionary histories (e.g. Lathman and Ricklefs 1993). As a consequence, complex latitudinal diversity gradients may emerge from the effect of processes interrelated at different scales of time and space. In the present paper, we address some of these issues using South American mammals as a model system. We incorporate the effects of spatial structuring of ecological variables into tests of three non-mutually exclusive hypotheses which are most often proposed to account for spatial patterns of variation in mammal species richness: species-energy, environmental stability and habitat-heterogeneity hypotheses. Note, however, that we will not address the possible role of stochastic factors within bounded domains in affecting gradients in mammal species richness in South America. Although the ‘‘mid-domain effect’’ (e.g. Colwell and Hurtt 1994, Willig and Lyons 1998, Colwell and Lees 2000) has been considered a credible hypothesis to explain the latitudinal diversity gradients during the past decade, recent work suggests that the model is flawed in its assumptions 402

and explains very little variation in species richness at macro-scale (see Hawkins and Diniz-Filho 2002, Zapata et al. 2003). A critical assessment of mid-domain models on the species richness patterns of south american mammals is out of the scope of the present analysis (see however: Willig and Lyons 1998, Willig 2000). In the present analysis, we conduct tests separately for different taxonomic groups of mammals (i.e. marsupials, edentates, primates, rodents, artiodactyls, carnivores and bats) to explore different ways whereby two of these hypotheses (energy and heterogeneity) may interact to explain variation in mammal species richness at the continental scale. We also partition the total number of mammal species into range size quartiles to give a more complete picture of species’ distributions. Species with small geographic ranges may be more likely limited by environmental factors that vary on a local or regional scale (e.g. topography); in contrast, widespread species should be less affected by such regional factors (e.g. Brown and Maurer 1989, Brown 1995, Jetz and Rahbek 2002). Hence, in the present paper, we analyse whether determinants of overall species richness patterns are also representative for groups of species with different distributions (e.g. for narrow-ranging and widespread species, as in Jetz and Rahbek 2002). We also show that differences in range size are not independent of taxonomy in this species assemblage.

Hypotheses The species-energy hypothesis proposes that richness is limited by the total or average amount of energy entering into an ecosystem. It predicts that high-energy availability promotes the persistence of high species richness (Wright 1983, see also Brown 1988, Brown and Lomolino 1998, Currie 1991, Gaston and Blackburn 2000). Here, we follow Hawkins et al. (2003) and distinguish between two versions of the energy-hypothesis (i.e. productivity vs ambient-energy versions): The productivity hypothesis (Wright 1983, see also Hawkins et al. 2003) states that the level of resource production in an ecosystem limits animal species richness. In this paper, we test this hypothesis using measures of energy closely related to the level of primary production in a region: annual evapotranspiration (AET) and annually Integrated Normalized Difference Vegetation Index (INDVI). AET is the amount of water that actually evaporates or is transpired from an area, depending on the joint availability of energy and water (see e.g. Currie and Paquin 1987, Currie 1991, Andrews and O’Brien 2000). INDVI provides an integrated index of ecosystem function through its strong correlation with aboveground net primary productivity and absorbed photosynthetically active radiation (Kerr and Ostrovsky 2003). INDVI shows a non-linear relationship with the ECOGRAPHY 27:4 (2004)

leaf-area index that defines canopy structure (Waring and Running 1998), however, it remains the most commonly used and most intensively studied vegetation index (see Kerr and Ostrovsky 2003 for discussion). Both AET and INDVI can be considered indicators of the availability of energy (in chemical form) for primary consumers (see Whittaker et al. 2001). We predict a general positive relationship between AET, INDVI and mammal species richness at continental scale. The ambient-energy hypothesis (see Turner et al. 1996, Hawkins et al. 2003 and other references therein) proposed that the animal species richness of a region is directly controlled by the total or average energy available. Given that endotherms usually maintain themselves at a higher temperature than that of the environment, the lower the ambient temperature the more these animals must spend to maintain body temperatures, and consequently the less energy they devote to growth and reproduction. All other things being equal, higher temperatures will promote faster growth of individuals and populations; this greater biomass will, in turn, promote greater species richness (see Brown 1988, Blackburn and Gaston 1996b, Turner et al. 1996, Hawkins et al. 2003 for more detailed discussion). In this paper, we use potential evapotranspiration (PET) and minimum temperature of the coldest month (TMIN) to test this hypothesis. PET is the amount of water that evaporates from a saturated surface, depending mainly on the amount of energy available to evaporate water and, to a lesser degree, on the relative humidity (Currie 1991, see also Andrews and O’Brien 2000). TMIN is a partial index of the environmental energy regime that describes the minimum degree of heat at each particular site (e.g. Andrews and O’Brien 2000). It also relates the ambient-energy hypothesis to the original climatic freezing-hypothesis of von Humboldt (Hawkins 2001, Hawkins et al. 2003). We predict general positive relationships between PET, TMIN and mammal species richness. The environmental variability hypothesis proposes that temporally less variable environments usually permit a greater number of species to coexist because species are able to specialize more and to evolve narrower ecological niches. In contrast, more variable environments are expected to have low species richness as a consequence of fewer species being able physiologically to tolerate the stressful conditions of varying environments (for theoretical discussions see e.g. Pianka 1966, MacArthur 1972, Rohde 1992, Turner et al. 1996, Brown and Lomolino 1998, Whittaker et al. 2001). Empirical analyses of mammal species richness patterns have offered equivocal evidence in support of this explanation (see e.g. Kerr 1999, Andrews and O’Brien 2000). We recognize the variability in resource supply (e.g. Connell and Orians 1964) and climate as different versions of the ECOGRAPHY 27:4 (2004)

environmental variability hypothesis. We used phenological seasonality (SEAS) and interannual variability or instability (INST) in INDVI to test the resource supply-stability hypothesis (i.e., intra- and interannual variability in resource availability). The difference in temperature between the warmest and coldest month i.e., yearly temperature amplitude (AMPLIT), is used to test the climatic-stability hypothesis. We predict that mammal species richness increases as climatic and energy supply stability increases. The habitat-heterogeneity hypothesis proposes that high spatial heterogeneity promotes the persistence of high species richness because the limiting resources can be more readily subdivided in complex habitats. This promotes greater specialization and the coexistence of a great number of species (see Simpson 1964, Pianka 1966, MacArthur 1972, McCoy and Connor 1980, Rohde 1992, Kerr and Packer 1997, Brown and Lomolino 1998). In the Americas, the high habitat complexity and strong elevation gradient associated with the presence of mountains in the west of the continent are associated with high diversity in mammals (Simpson 1964, Patterson 1994, Patterson et al. 1996, Kerr and Packer 1997). In the present analysis, we use the spatial heterogeneity in NDVI (HETER) and elevation variability (EVAR) as two measures of habitat-heterogeneity. We predict that the number of mammal species increases as HETER and EVAR increase.

Methods Data on South American mammals Distributional information was compiled for 825 species of the following taxa: Marsupialia (MAR; N /61), Edentata (EDE; N /28), Chiroptera (CHI; N /189), Primates (PRI; N /78), Rodentia (Hystricognathi, HYS; N /155, Sciurognathi, SCI; N /247), Artiodactyla (ART; N /20), Carnivora (CAR; N /42), Perissodactyla (N /3), Lagomorpha (N /2). The basic data are the same used in Ruggiero (1994, 1999) and Ruggiero et al. (1998). These previous papers should be consulted for details of the main sources of distributional information used and discussion of the potential problems associated with these data. In the present paper we added data on Sciurognathi and updated the original information for all mammal species based on the maps published in Eisenberg (1989), Redford and Eisenberg (1992), and Eisenberg and Redford (1999). The geographic range of each species was drawn onto a cylindrical equal area (Peters’) projection map of South America, overlaid by a grid of 170 squares, each cell approximately covering 123 000 km2. The spatial resolution is the same used in Ruggiero (1999) and was maintained fixed throughout the present analysis. The presence (1) or absence (0) of each species was recorded 403

in each cell of the grid map. Given the coarse-scale resolution of our analysis, we adopt a conservative criterion to delimit species’ distributions; a species was assigned to be ‘‘present’’ when its geographic range covers at least 25% of any cell. Individual species grids were entered and further processed in IDRISI ver. 2.0 (Eastman 1997). Species richness throughout the present analysis was the total number of species coexisting in each cell of a grid map. We analyse whether similar (or different) environmental determinants account for species richness patterns in narrow-ranging, intermediate-ranging and widespread species. Range size was estimated simply by counting the number of cells occupied by each species. Given the coarse-scale resolution of our analysis, it was not possible to settle an exact limit for each range size quartile. Thus, each quartile actually proportionally represented 27.6% (Q1: narrow-ranging species), 23.1% (Q2: intermediate-ranging species), 24.2% (Q3: intermediate-ranging species) and 24.9% (Q4: widespreadranging species) of the total of 825 mammal species.

Environmental data Remotely sensed vegetation data We used the NOAA/NASA Pathfinder AVHRR NDVI dataset (1981 /1999) sub-setted for South America (838W to 338W; 13.28N to 578S) (Agbu and James 1994). The Normalized Difference Vegetation Index (NDVI) measures the proportion of the photosynthetically absorbed radiation and vegetation structure. It is calculated from atmospherically corrected reflectance (R) from the visible and near-infrared AVHRR channels. Data of NDVI originally acquired at 1 /1 km resolution were resampled into 8 /8 km (4.35 /4.35?) using the equal-area Goode Interrupted Homolosine Projection (Agbu and James 1994). Data were derived from compositing daily-derived images over 10-d periods to minimize cloudiness and smoke and/or fog effects. Tenday maximum NDVI composites retained considerable amount of anomalous reflectance caused by cloud and/ or smoke contamination. Thus, ten-day maximum NDVI images were composited to increasingly larger periods until clouds and/or smoke effects disappeared. Because complete loss of these effects was attained with six-month maximum NDVI composites, semester NDVI composites were obtained for the periods October /March (OMi) and April /September (ASi). Unacceptable quality in images from the second semester of 1993 through the first semester of 1995, the OM 93 /94, OM 94 /95 and AS 94 NDVI images precluded their inclusion in the general dataset. Thus, the total number of images over which analyses were performed was n /15 and n /16 for OMi and ASi, respectively. Each semester was averaged over the years 404

to obtain mean six-month NDVI (OM and AS). We used six-month composite NDVI images to estimate the following variables: annually Integrated Normalized Difference Vegetation Index (INDVI), phenological seasonality (SEAS), mean annual integrated NDVI (INDVI), interannual variability or instability (INST), and spatial heterogeneity in NDVI (HETER). We calculated phenological seasonality (SEAS) as: SEAS jOM ASj Annual integrated NDVI (INDVIi) for each of the 15 years was obtained as: INDVIi (OMi ASi )=2 and X INDVI INDVIi =15 The degree of interannual variability or for instability (INST) in INDVI in each pixel had was estimated as the coefficient of variation: INST 100 (1=2) X =15 =INDVI (INDVIi INDVI)2 Spatial heterogeneity in NDVI (HETER) was quantified by calculating Shannon’s diversity index over a 7 /7 pixel (ca 56 /56 km) moving window on the INDVI image: X HETER p log p where p is the proportion of each of 128 INDVI classes in the 7 /7 window. NDVI cover-class diversity was preferred over cover class richness as it accounts for the evenness of cover classes. Because of slight registration errors in the multitemporal dataset, contamination by pixels corresponding to ocean reflectance occurred. By applying a distance operator to the ocean mask, all pixels closer than 2 pixels (diagonals included) from water were excluded from pixel aggregation process for the estimation of INDVI, SEAS and INST. Similarly, to avoid the moving window that estimated heterogeneity to be contaminated by pixels corresponding to water, all pixels closer than 3 pixels (diagonals included) from water were excluded from pixel aggregation process for the estimation of HETER. Other physical variables We estimated elevation variability (EVAR) as the standard deviation in elevation calculated over a 20 /20 pixel (ca 370 /370 km) window on the 10 /10? Clark FNOC Elevation, Terrain, and Surface Characteristics Global Dataset (Clark 1992). Minimum temperature of the coldest month (TMIN) and yearly temperature amplitude (AMPLIT; the difference in temperature between the warmest and coldest month) were obtained from the 30 /30? IIASA database for mean monthly ECOGRAPHY 27:4 (2004)

values of temperature, precipitation, and cloudiness on a global terrestrial grid (Leemans and Cramer 1992). Annual actual (AET) and potential evapotranspiration (PET) were obtained from the 30 /30? Ahn and Tateishi monthly potential and actual evapotranspiration and water balance dataset (Ahn and Tateishi 1994). To quantify variation in species richness produced in coastal areas due to reductions in the amount of land, the proportion of (continental) land in each cell (PLAND) was calculated and incorporated as an independent variable. Geographic rectification and resampling All images corresponding to the environmental variables were projected to the Peters projection of the mammal dataset by rubbersheeting on 20 uniformly distributed ground control points applying quadratic transformation and nearest neighbor resampling to a new resolution that was a multiple of the resolution of the mammal dataset. Finally, images were contracted to the resolution of mammal dataset by pixel averaging over fixed windows corresponding geographically to the cells of the mammal dataset. All operations were performed with IDRISI ver. 2.0 (Eastman 1997).

Analyses A number of methods have been proposed to assess the effects of spatial structuring of variables in statistical modeling (see Legendre 1993, Lichstein et al. 2002, Diniz-Filho et al. 2003). We adopted the so-called ‘‘raw data’’ approach, whereby the effects of environmental variables on species richness variation are examined by partial regression analysis; the effect of space in this kind of analysis is partitioned out by site variables as in trend surface analysis (see Borcard et al. 1992, Legendre 1993, Boone and Krohn 2000, Lichstein et al. 2002, Van Rensburg et al. 2002 for detailed explanation). We partialled out the spatial variation in mammal species richness into for components, representing the effects of a) local (‘‘non-spatial’’: sensu Borcard et al. 1992, Legendre 1993) environmental variation, b) regional (‘‘spatially structured’’: sensu Borcard et al. 1992, Legendre 1993) environmental variation, c) spatial variation of mammal species richness that is not shared by the environmental variables; this component could reflect the effect of other unknown biotic or abiotic processes that are also spatially structured (e.g. species interactions within communities, social aggregation, etc.), interactions between variables considered in the present study that were not included in the statistical model (see Borcard et al. 1992, Boone and Krohn 2000), and d) unexplained variation; this component represents the local effect of other unknown environmental factors. ECOGRAPHY 27:4 (2004)

Multicollinearity among predictor variables may introduce serious distortions in standard multiple regression analyses (see e.g. Chattrjee and Price 1977). All environmental and spatial predictors used in the present analyses are significantly correlated (all rs /0.7, pB/0.0001), with the highest Spearman correlation coefficients (rs) obtained among energy descriptors. This complicates selection of the best environmental and spatial predictors of mammal species richness (see Diniz-Filho et al. 2003 for discussion). To avoid final models with highly redundant and multicollinear data structures, we applied a forward selection procedure with the proviso that a tolerance threshold of 0.6 was used when entering a new variable into the model. Tolerance is equal to (1 /R2i ), where R2i is the squared multiple correlation coefficient from the regression of the ith predictor variable on all other independent variables in the regression equation. Once the best environmental and spatial predictors have been selected by this procedure, we combined their effects in a standard multiple linear regression analysis. This allows us to evaluate the significance of the best environmental predictors in the presence of the best spatial predictors. In the section of results that follows, we report parameter estimates exclusively from these final regression models. Partial regression analyses performed in the present study involved several steps: 1) Forward elimination procedure, with a tolerance threshold of 0.6 to overcome the problems generated by multicollinearity, was used to select the best environmental predictors for each taxonomic group or quartile of species considered. The coefficient of determination from this step (r21) measured the proportion of the total variation of mammal species richness explained by the local and regional spatially structured environmental variation (i.e. fraction (a /b) below). 2) A FORTRAN program (SpaceMaker: compiled version for DOS available at B/http://www.fas.umontreal.ca/biol/legendre/ /) was used to describe the spatial structure of data by a third order polynomial function of the geographic coordinates of the cells in the grid map. To fit the trend surface to data on mammal species richness, we regressed species richness on all polynomials terms: z b0 b1 x b2 y b3 x2 b4 xy b5 y2 b6 x3 b7 x2 y b8 xy2 b9 y3 Forward elimination procedure, with a threshold tolerance of 0.6, was used to select the individual terms of the spatial polynomial to be considered the best predictors of species richness for each mammal group or quartile of mammal species. The coefficient of determination from step 2 (r22) measured the proportion of the total variation of mammal species richness explained by regional spatially structured environmental variation and spatial variation that is not shared by the environmental variables (i.e. fraction (b /c) below). 3) A standard 405

multiple linear regression analysis of species richness applied on the previously selected (i.e. in steps 1 and 2) environmental and spatial variables allowed to estimate the total proportion of variance of species richness explained (r23) by all (spatial and environmental) variables considered for each group of species (fraction (a / b /c) below). 4) As in Boone and Krohn (2000), the different sources of variation were calculated as: b (a b) (b c) (a b c) a (a b) b; c (b c) b; d 1 (a b c) Note that partial regression analysis, as applied in the present study, is only an exploratory way of examining the extent to which spatial autocorrelation may be influencing the effects of environmental factors operating on species richness because it only accounts for broad-scale spatial trends; however, even after controlling for broad-scale spatial patterns in the variables examined, the residuals in the fitted model may still be autocorrelated, indicating evidence of spatial dependence (see Lichstein et al. 2002, Diniz-Filho et al. 2003). We performed all the analyses twice. Preliminary analyses were based on raw (untransformed) variables. Then, we performed all analyses based on transformed variables because the assumption of a linear relationship between species richness and some of the predictor variables (EVAR, AMPLIT, SEAS, INST) was better conformed when these variables were log10-transformed; we also applied a square root transformation on the dependent variable to ensure normality and constancy of error variance (see e.g. Chattrjee and Price 1977). We evaluated qualitatively the extent to which the use of transformations changed our results, i.e. whether different predictor variables were selected based on transformed or raw models. Throughout the present analysis, the tests of hypotheses were performed first for all non-flying mammal species taken together. Then, we repeat all the analyses separately for each mammal group or taxon considered and for each range size quartile. Note, however, that the sample size in each analysis is the number of grid squares, rather than the number of species, and so is fairly constant across all analyses: Marsupialia (MAR; N /154), Edentata (EDE; N /155), Chiroptera (CHI; N /166), Primates (PRI; N /124), Rodentia (Hystricognathi, HYS; N /165, Sciurognathi, SCI; N /165), Artiodactyla (ART; N /165), Carnivora (CAR; N /166), and range size quartiles (Q1 4; N /165).

Results Tests of hypotheses Results based on transformed and raw data differ slightly. Given that the main qualitatively trends are 406

found in both analyses, we only report here the results of the analyses based on raw data. The species-energy hypotheses These hypotheses predict a general positive relationship between productivity (AET, INDVI), ambient energy (TMIN, PET) and mammal species richness at the continental scale. When all non-flying mammal species are analysed together, the productivity hypothesis is supported; actual evapotranspiration (AET) is selected as the best energy-based predictor of mammal species richness (Table 1); the significant positive effect of AET on species richness is maintained after including the effect of spatial structuring of data into the analysis (see Table 1 and Appendix 1). When eight mammal taxa are analysed separately, both the productivity and ambient energy versions of the energy hypotheses are supported. After including the effect of spatial structuring of data into the analyses, 4 out of 8 taxa show AET as a significant and positive predictor of mammal species richness. Similarly, the species richness of edentates increases as INDVI increases, and so do the richness of bats with TMIN and artiodactyls with PET (Table 1 and Appendix 1); given the nature of multiple comparisons, and setting a pB/0.05, 5% of the taxa would show a significant effect of the environmental variables considered by chance alone. However, the probability that 4 out of 8 taxa would show AET as a significant predictor of species richness (binomial p(4,8,0.05) /0.000014) indicates that the possibility of committing an overall Type I error is low (for INDVI, TMIN and PET, binomial p(1,8,0.05) /0.057). The partition of species into range size quartiles shows that the positive effect of energy on mammal species richness does not change substantially with range size. However, the narrowest-ranging and widest-ranging species show distinct biogeographic patterns (Fig. 1). As a consequence, the productivity version of the hypothesis is supported by narrow- and intermediateranging species, with AET explaining just under 50% of mammal species richness variation. The ambient-energy hypothesis is supported only by the widest-ranging species, with TMIN explaining a greater proportion (ca 70%) of species richness variation (Table 1; Appendix 1). The environmental variability hypothesis This hypothesis predicts that mammal species richness will increase as climatic and energy supply stability increase, with this translating into a general negative relationship between phenological seasonality (SEAS), interannual variability in NDVI (INST), and yearly temperature amplitude (AMPLIT) and mammal species richness. When the effect of spatial structuring of data is included into the analyses only the primates support this ECOGRAPHY 27:4 (2004)

Table 1. Summary of key results obtained throughout the present analysis. Significant (p B/0.05) positive or negative relationships between variables after including the effect of spatial structure. One relationship that turned non-significant (p /0.05, in marsupials) after including the effect of spatial structure is within parentheses. TOT /all mammal species excluding bats, MAR /marsupials, EDE /edentates, CHI /bats (Chiroptera), PRI /primates, HYS /hystricognath rodents, CAR /carnivores, ART /artiodactyls, SCI /sciurognath rodents. Q1, Q2, Q3 and Q4 are quartiles as defined in the main text. See Appendix 1 for estimated coefficients and exact probability values. HYPOTHESES Species-energy

Environmental stability

Habitat

Land

Productivity

Ambient energy

Resource supply

Climatic

Heterogeneity

Area

AET

TMIN

SEAS

AMPLIT

HETER

EVAR

PLAND

/ ( /)

/ /

INDVI

Taxonomic groups TOT / MAR / EDE / CHI PRI HYS / CAR / ART SCI / Quartiles / Q1 Q2 / Q3 / Q4

PET

/ /

/ / /

prediction, showing a significant decrease in species richness as the yearly temperature amplitude (AMPLIT) increases (Table 1; Appendix 1). Although inter-annual variability in NDVI (INST) is selected as a significant predictor of mammal species richness variation for the narrowest ranging species (Q1), the positive regression coefficient (b /0.19; Appendix 1) is against our original prediction. A region of high interannual instability (INST) may explain this intriguing relationship, previously observed along the eastern slopes of the tropical Andes (see e.g. Fig. 4 in FjeldsaË&#x161; et al. 1999); in the present analysis, this region of high instability (INST) is included within cells where peaks in the richness of narrow- ranging species occurred.

The habitat heterogeneity hypothesis This hypothesis predicts that the number of mammal species will increase as spatial heterogeneity in NDVI (HETER) and elevation variability (EVAR) increase. When the effect of spatial structuring of data is included into the analyses, all non-volant mammal species analysed together show a significant positive effect of EVAR on mammal species richness (Table 1; Appendix 1). This positive effect is also shown in 3 out of the 8 taxa analysed separately (EVAR: binomial p(3,8,0.05) /0.00037). In contrast, the effect of spatial heterogeneity in vegetation (HETER) is less important and against our original prediction: it has a negative effect only on the species richness of primates, suggesting ECOGRAPHY 27:4 (2004)

INST

that species richness decreases in locally heterogeneous habitats (see Table 1 and Appendix 1). The partition of mammal species into range size quartiles confirms that elevation variability is an important predictor of mammal species richness for narrow- and intermediate- ranging species but not for the widest-ranging species. The proportion of species richness variation explained by elevation variability decreases moderately with range size, from 48% in Q1 to 34% in Q3.

Effects of spatial structure About 50% of the total mammal species richness variation at the continental scale is explained by the regional spatially structured component of environmental variation. However, there are clear differences across taxa in the proportion of variance explained by this component (e.g. between bats: 74% and primates or rodents: ca 20 /30%; Fig. 2). Not surprisingly, this component accounts for a greater proportion of species richness variation in the widest-ranging species (Q4) rather than in intermediate- or narrow- ranging species (Fig. 2). Effects of taxonomy and range size are also detectable on the proportion of mammal species richness variance explained by local environmental effects. In general, such effects explain B/30% of mammal species richness variation. However, local effects of environment are higher for primates (42%) and rodents (hystricognath rodents: 22%, sciurognath rodents: 21%) and almost 407

Fig. 1. Geographic pattern of mammal species richness variation after species were assigned to range size quartiles. (a) Q1: first quartile, species’ range sizes covering from 1 to 3 cells in the grid map; (b) Q2: second quartile, species’ range sizes covering from /3 to 9 cells; (c) Q3: third quartile, species’ range sizes covering from /9 to 31 cells; (d) Q4: fourth quartile, species’ range sizes covering /31 cells.

undetectable for edentates (0.04%), bats (0.04%) and artiodactyls (0.05%) (Fig. 2). In general, local environmental effects on mammal species richness tend to decrease with range size (Fig. 2). The proportion of variance explained by component c) (spatially structured processes not considered in the present analysis) is B/2% in all the analyses performed (Fig. 2). This suggests that the spatial descriptors selected in the best-fit models adequately describe the spatial structure in these ecological data. The proportion of variance that remained unexplained by the environmental factors considered in the present analysis is between 10 and 50% depending upon the taxonomic group considered; the partition of species into range size quartiles shows that a greater proportion of variance remains unexplained for narrow- and intermediate408

ranging species than for widespread species (Fig. 2). These trends suggest local effects of other unknown environmental descriptors on the mammal species richness pattern.

Discussion Environmental effects Mammals in South America support the productivity version of the species-energy hypothesis, and suggest actual evapotranspiration (AET) is the main energy predictor. This favors the idea that it is not direct energy per se but its transformation into different levels of resources available to mammals that is the mechanism ECOGRAPHY 27:4 (2004)

Fig. 2. Proportion (%) of species richness variance explained by environment in each mammal taxon (top) or range size quartile (bottom). Env. (local): variation in species richness explained by fine-scale effects of environment. Env. (regional): variation in species richness explained by regional spatially structured variation of environment. Space: spatial variation of mammal species richness that is not shared by the environmental variables (unexplained spatially structured variation). Unexpl.: unexplained variation ( /local effects of unknown environmental factors).

underlying the species-energy relationship at the continental scale (see also Van Rensburg et al. 2002 and other references cited therein). The ambient energy hypothesis receives less support throughout the present analysis. Our study contradicts previous evidence suggesting that potential evapotranspiration (PET) is the most important and least ambiguous predictor of species richness variation at continental (e.g. Currie 1991, Kerr and Packer 1997, Hawkins et al. 2003) or global scales (e.g. Whittaker et al. 2001). We also show that the effect of energetic constraints / as expressed by the minimum temperature of the coldest month (TMIN) / is detectable only for bats and/or for the widest-ranging species. Indeed, bats contribute with a great proportion of species to the widest-ranging quartile (see details below), which explains the parallel trend. Neotropical bats are generally considered to be eurytopic with their distributional limits being generally poorly constrained by phytogeographic zones (e.g. Willig and Mares 1989). Although most bats are insectivores, they have become adapted to obtain food from a variety of sources / e.g. insects, fish, fruit, nectar, small vertebrates, blood, and because of their specific capacities for echolocation and maneuverable flight they have been able to develop nocturnal habits (see e.g. Neuweiler 1989, Arita and Fenton 1997). In general, Neotropical bats are poor ECOGRAPHY 27:4 (2004)

thermo- regulators although the capacity to ensure effective temperature regulation varies among different taxa depending upon food habits (Fig. 3 in McNab, 1982). The subtropics (between 308 and 358S) represent the southernmost limit of distribution for several taxa (Stenodermatinae, Carollinae, Phyllostominae, Sturnirinae, Desmodontinae: McNab 1982), which also suggests their sensitivity to cold. As a consequence, our finding of a significant positive effect of TMIN on bat species richness patterns is not surprising and could explain the strong latitudinal gradient in species diversity previously reported for this taxon. Indeed, latitude alone was previously reported to account for ca 85% of the variation in bat species density (e.g. Willig and Selcer 1989, Kaufman and Willig 1998). The present analysis confirms that continental patterns of variation in mammal species richness are explained by the synergism between energy (as expressed by AET or INDVI) and habitat heterogeneity. Elevation variability is the main heterogeneity predictor, mainly due to the presence of the Andes running along the west of South America (see also Kerr and Packer 1997, Rahbek and Graves 2001). The present analysis suggest that elevation variability is more influential on the species richness patterns of rodents (representing ca 48% of analysed species) and artiodactyls, and this may be explained by the biogeographic history of these taxa being clearly associated to the Andean habitats (see e.g. Franklin 1982, Reig 1986, Eisenberg 1987, Marquet 1989, Smith and Patton 1993 for discussion). Conceptually, we recognize three main ways whereby energy and heterogeneity can interact to promote species richness (Fig. 3, top). 1) Heterogeneity may have no effect and species richness is principally controlled by energy availability. 2) Habitat heterogeneity and energy availability may contribute independently to species richness (additive effects), or 3) Heterogeneity may increase in importance with the increase in energy availability (multiplicative effect; Fig. 3 top). Empirical evidence in the literature provides evidence for the last case (Kerr and Packer 1997). Model 3 is also supported for all nonflying mammals in our study. However, when high order taxa are analyzed separately, it is clear that different groups display different energy-heterogeneity synergism. Edentates and Chiroptera conform to model 1, with richness relatively independent of heterogeneity. Primates and hysticognath rodents conform to model 2. Here, heterogeneity increases richness in both high and low energy regions. Note, however, that primates span a narrower range of energy variation (mainly tropical and subtropical) than the other taxa. Finally, carnivores, artiodactyls, sciurognath rodents, and marsupials all conform to model 3, showing that heterogeneity-richness relationships become markedly steeper when energy availability increases (Fig. 3). Clearly, the main behavior 409

Fig. 3. Synergism between energy availability and habitat heterogeneity. Species richness is principally controlled by energy availability (no effect of heterogeneity; top left), habitat heterogeneity and energy availability contribute additively to species richness (top center), and heterogeneity increases in importance with the increase in energy availability (multiplicative effect; top right). Bottom graphs represent best fits to quadratic surfaces of mammal species richness (TOT /non-flying mammals, and each taxon) as a function of elevational variability (EVAR) and annually integrated NDVI (INDVI).

410

ECOGRAPHY 27:4 (2004)

for all non-flying species is heavily influenced by these last 4 taxa, representing ca 58% of the analyzed species. The climatic variability hypothesis is not supported throughout the present study, thus confirming a previous observation for North American mammals (Kerr 1999). In contrast, we find a positive effect of inter-annual variability in resource supply on the richness of narrow ranging species. The species richness pattern in the narrow- ranging species shows peaks in species richness associated with the Andean mountain ranges. The high ecoclimatic instability observed in the eastern slopes of the tropical Andes and coastal deserts of the western slopes may be partly the response to the dominant signal of El NinË&#x153;o Southern Oscillation (ENSO) which causes major dislocations in the dominant circulation that mostly affect topographically controlled rainfall regimes. Continental-scaled hydrological studies indicate strongest responses to ENSO in the tropical eastern slopes and Andean highlands where dry and wet periods correspond to warm and cold phases of ENSO, respectively (Dettinger et al. 2000). Paleoecological and archeological evidence also point to the eastern tropical Andes as a climatically sensitive region at both the inter-annual and century scale (Paulsen 1976, Thompson et al. 1992). Over these tropical mountain ridges high spatial and temporal variability in rainfall may induce sharp changes in vegetation. Over long time scales these environmental changes may have differentially affected speciation patterns of narrow-ranged mammals that have little migration potential in tropical mountain ranges (Janzen 1967). At this point, however, we must emphasize that the definition of regional climatic variability most often used in analyses of species diversity (e.g. Kerr 1999, Andrews and Oâ&#x20AC;&#x2122;Brien 2000) refers exclusively to intraannual climatic variability, or seasonality, in temperature and precipitation. In reality, climatic variability can affect speciesâ&#x20AC;&#x2122; distributions / and hence species richness patterns /at different temporal scales. Although in the present analysis we have tried in part to overcome this problem by estimating a measure of inter-annual variability in resource supply (INST), our window of time is still well within an ecological scale of analysis (15 yr). Other processes acting at longer time scales could underlie the spatial variation in species richness at the continental scale, although they may remain unnoticed at the relatively short-time scale of our present analysis (see Brown 1995, Rohde 1996, Dynesius and Jansson 2000, Clarke and Crame 2003 for examples and discussions).

Effects of range size, taxonomy and spatial structure Mammals in South America suggest that differences in the range sizes of species may affect our perception of ECOGRAPHY 27:4 (2004)

environmental determinants of species richness patterns at the continental scale. Different environmental descriptors account for the species richness patterns in the narrowest- and widest-ranging species, as previously observed by Jetz and Rahbek (2002) for the sub-Saharan avifauna. However, a difference between Africa and South America is clear. In Africa, the importance of habitat heterogeneity in accounting for variation in bird species richness increases markedly with decreasing range size, and the opposite is true for the effect of productivity (Jetz and Rahbek 2002). In South America, determinants of overall species richness (i.e., elevation variability and productivity) consistently explain about the same proportion of variance across most of the range size classes considered (i.e. from Q1 3). Only the widestranging species (Q4) show a different pattern, as they support exclusively the ambient-energy hypothesis (TMIN). The stronger effect of elevation variability to account for species richness patterns in South American mammals is, in part, explained by the greater distributional extent and complexity of the Andean mountain ranges compared to African mountains. Another reason is that species richness patterns of narrow- and intermediate-ranging species are rather similar in South America. Both groups of species suggest an increase in mammal species richness toward the west, in tropical Andean regions. In contrast, the richness of the widestranging mammal species peak in tropical lowlands (Fig. 1). Clearly, the main determinants of mammal species richness differ markedly only between those range size quartiles that show distinct biogeographic patterns. Mammals in South America confirm that species richness variation at continental scale is driven mainly by broad-scale spatially structured components of environmental variation, as previously reported by Boone and Krohn (2000) and Van Rensburg et al. (2002). However, South American mammals allow more detailed interpretation of the extent to which differences in taxonomy and range size affect the proportion of species richness variation explained by local and regional environmental effects. Major taxonomic groups differ in the proportion of variance explained by fine-scale effects of the environment. The partition of mammal species into range size quartiles clearly shows that local environmental effects account for a greater proportion of species richness variation for narrow- and intermediateranging species, whereas regional effects increase markedly in importance for the widest-ranging species. However, differences in range size are not independent of differences in taxonomy in this species assemblage (x2 /282.52, p(21) B/0.001; Fig. 4). Only marsupials and artiodactyls contribute with equal proportions of species to each size range size quartile (marsupials: x2 /0.907, artiodactyls: x2 /4.49, p(3) /0.05). p(3) /0.05; The other taxonomic groups contribute with a significant greater proportion of species to either the 411

Fig. 4. Proportional representation of species for each taxonomic group in each range size quartile.

narrow-ranging quartiles (primates: x2 /16.26, p(3) B/ 0.01; Hystricognathii: x2 /32.1, p(3) B/0.01; Sciurognathii: x2 /60.79, p(3) B/0.01) or to the widest-ranging ones (edentates: x2 /11.39, p(3) B/0.05; bats: x2 / 127.34, p(3) B/0.001; carnivores: x2 /29.25, p(3) B/0.01) (Fig. 4). This complicates distinguishing between ecological and historical components of mammal species richness variation at continental scale. However, one possible interpretation, on ecological grounds, is that the influence of range size on species richness patterns may be related to the spatial scale or ‘‘grain’’ at which different mammal taxa perceive the environment, according to their body size and/or dispersal capabilities. Suggestively, in the present study, the species richness patterns in the high mobile bats and in the large bodysized terrestrial artiodactyls were better explained by the regional spatially structured components of environmental variation. The proportion of species richness variance explained by local effects of environment is very low in bats (B/1%), approaching values previously reported for birds in South Africa ( B/2%: Van Rensburg et al. 2002). In contrast, the highest proportion of variance explained by fine-scale environmental effects is found in primates (ca 40%), and in the smaller and less vagile rodents (ca 20%). Clearly, only further analyses that control for functional groups and evolutionary and biogeographic history will provide a definite answer around these issues. Acknowledgements / Insightful comments and suggestions were provided by T. M. Blackburn, S. L. Chown, J. A. F. Diniz-Filho, A. G. Farji-Brenner, B. A. Hawkins, P. A. Marquet, A. Premoli and K. Roy, and several other colleagues attending the Friday Afternoon Seminars at Laboratorio Ecotono. CONICET (PEI: no. 317/98) and Universidad Nacional del Comahue supported this project.

References Agbu, P. A. and James, M. E. 1994. The NOAA/NASA Pathfinder AVHRR land data set user’s manual. / Goddard Distributed Active Archive Center, NASA, Goddard Space Flight Center, Greenbelt, MD, B/http://eosdata.gsfc. nasa.gov/data/dataset/AVHRR/02_Cd_Rom/index.html /.

412

Ahn, C. H. and Tateishi, R. 1994. Development of a Global 30minute grid potential Evapotranspiration Data Set. Journal of the Japan Soc. / Photogram. Rem. Sens. 33: 12 /21. Andrews, P. and O’Brien, E. M. 2000. Climate, vegetation, and predictable gradients in mammal species richness in southern Africa. / J. Zool. 251: 205 /231. Arita, H. T. and Fenton, M. B. 1997. Flight and echolocation in the ecology and evolution of bats. / Trends Ecol. Evol. 12: 53 /58. Arita, H. T. and Rodrı´guez, P. 2002. Geographic range, turnover rate and the scaling of species diversity. / Ecography 25: 541 /550. Begon, M., Harper, J. L. and Townsend, C. R. 1986. Ecology. Individuals, populations and communities. / Sinauer. Blackburn, T. M. and Gaston, K. J. 1996a. Spatial patterns in the species richness of birds in the New World. / Ecography 19: 369 /376. Blackburn, T. M. and Gaston, K. J. 1996b. A sideways look at patterns in species richness, or why there are so few species outside the tropics. / Biodiv. Lett. 3: 44 /53. Blackburn, T. M. and Gaston, K. J. 1997. The relationship between geographic area and the latitudinal gradient in species richness in New World birds. / Evol. Ecol. 11: 195 / 204. Blackburn, T. M. and Gaston, K. J. 2002. Scale in macroecology. / Glob. Ecol. Biogeogr. 11: 185 /189. Blackburn, T. M. and Gaston, K. J. (eds) 2003. Macroecology: concepts and consequences. / The 43rd Ann. Symp. Brit. Ecol. Soc. British Ecol. Soc. Boone, R. B. and Krohn, W. B. 2000. Partitioning sources of variation in vertebrate species richness. / J. Biogeogr. 27: 457 /470. Borcard, D., Legendre, P. and Drapeau, P. 1992. Partialling out the spatial component of ecological variation. / Ecology 73: 1045 /1055. Brown, J. H. 1988. Species diversity. / In: Myers, A. A. and Giller, P. S. (eds), Analytical biogeography. Chapman and Hall, pp. 57 /89. Brown, J. H. 1995. Macroecology. / Univ. of Chicago Press. Brown, J. H. and Maurer, B. A. 1989. Macroecology: the division of food and space among species in continents. / Science 243: 1145 /1150. Brown, J. H. and Lomolino, M. V. 1998. Biogeography, 2nd ed. / Sinauer. Chatterjee, S. and Price, B. 1977. Regression analysis by example. / Wiley. Clark, K. 1992. FNOC Global elevation, terrain, and surface characteristics. Digital raster data on a 10-minute cartesian orthonormal geodetic (lat/long) 1080 /2160 grid. / In: Anon. (ed.), Global ecosystems database ver. 2.0. Boulder, CO: NOAA National Geophysical Data Center, B/http://www.ngdc.noaa.gov/seg/eco/cdros/gedii_a/dataset/ a13/fnoc.htm#top /. Clarke, A. and Crame, J. A. 2003. The importance of historical processes in global patterns of diversity. / In: Blackburn, T. M. and Gaston, K. J. (eds), Macroecology: concepts and consequences. The 43rd Ann. Symp. Brit. Ecol. Soc. British Ecol. Soc, pp. 130 /151. Colwell, R. K. and Hurtt, G. C. 1994. Nonbiological gradients in species richness and a spurious Rapoport effect. / Am. Nat. 144: 570 /595. Colwell, R. K. and Lees, D. C. 2000. The mid-domain effect: geometric constraints on the geography of species richness. / Trends Ecol. Evol. 15: 70 /76. Connell, J. H. and Orians, E. 1964. The ecological regulation of species diversity. / Am. Nat. 98: 399 /414. Currie, D. J. 1991. Energy and large-scale patterns of animaland plant-species richness. / Am. Nat. 137: 27 /49. Currie, D. J. and Paquin, V. 1987. Large-scale biogeographic patterns of species richness of trees. / Nature 329: 326 /327. Currie, D. J., Francis, A. P. and Kerr, J. T. 1999. Some general propositions about the study of spatial patterns of species richness. / E`coscience 6: 392 /399. ECOGRAPHY 27:4 (2004)

Dettinger, M. D. et al. 2000. Multiscale streamflow variability associated with El Nin˜o/Southern Oscillation. / In: Diaz, H. F. and Markgraf, V. (eds), El Nin˜o and the Southern Oscillation; multiscale variability and global and regional impacts. Cambridge Univ. Press, pp. 113 /147. Diniz-Filho, J. A. F., Bini, L. M. and Hawkins, B. A. 2003. Spatial autocorrelation and red herrings in geographical ecology. / Glob. Ecol. Biogeogr. 12: 53 /64. Dynessius, M. and Jansson, R. 2000. Evolutionary consequences of changes in species’ geographical distributions driven by Milankovitch climate oscillations. / Proc. Natl. Acad. Sci. USA 97: 9115 /9120. Eastman, J. R. 1997. IDRISI ver. 2.0 for Windows Geographic Analysis System. / The Clark Labs for Cartographic Technology and Geographic Analysis. Eisenberg, J. F. 1987. The evolutionary history of Cervidae with special reference to the South American radiation. / In: Wemmer, C. M. (ed.), Biology and management of the Cervidae. Res. Symp. Nat. Zool. Park. Smithsonian Inst., pp. 60 /64. Eisenberg, J. F. 1989. Mammals of the Neotropics. The Northern Neotropics. Vol. 1. / Univ. of Chicago Press. Eisenberg, J. F. and Redford, K. H. 1999. Mammals of the Neotropics. The Central Neotropics. Vol. 3. / Univ. of Chicago Press. Fjeldsa˚, J., Lambin, E. and Mertens, B. 1999. Correlation between endemism and local ecoclimatic stability documented by comparing Andean bird distributions and remotely sensed land surface data. / Ecography 22: 63 /78. Franklin, W. L. 1982. Biology, ecology and relationship to man of the South American camelids. / In: Mares, M. A. and Genoways, H. H. (eds), Mammalian biology in South America. Vol. 6. Spec. Publ. Ser. Pymatuning Lab. Ecol. Univ. of Pittsburgh, pp. 457 /489. Gaston, K. J. 2000. Global patterns in biodiversity. / Nature 405: 220 /227. Gaston, K. J. and Blackburn, T. M. 2000. Patterns and process in macroecology. / Blackwell. Hawkins, B. A. 2001. Ecology’s oldest pattern? / Trends Ecol. Evol. 16: 470. Hawkins, B. A. and Diniz-Filho, J. A. F. 2002. The mid-domain effect cannot explain the diversity gradient of Nearctic birds. / Glob. Ecol. Biogeogr. 11: 419 /426. Hawkins, B. A., Porter, E. E. and Diniz-Filho, J. A. F. 2003. Productivity and history as predictors of the latitudinal diversity gradients of terrestrial birds. / Ecology 84: 1608 / 1623. Janzen, D. H. 1967. Why mountain passes are higher in the tropics. / Am. Nat. 101: 233 /249. Jetz, W. and Rahbek, C. 2002. Geographic range size and determinants of avian species richness. / Science 297: 1548 / 1551. Kaufman, D. M. and Willig, M. R. 1998. Latitudinal patterns of mammalian species richness in the New World: the effects of sampling method and faunal group. / J. Biogeogr. 25: 795 /805. Kerr, J. T. 1999. Weak links: ‘‘Rapoport’s rule’’ and largescale species richness patterns. / Glob. Ecol. Biogeogr. 8: 47 /54. Kerr, J. T. 2001. Global biodiversity patterns: from description to understanding. / Trends Ecol. Evol. 16: 424 /425. Kerr, J. T. and Packer, L. 1997. Habitat heterogeneity as a determinant of mammal species richness in high-energy regions. / Nature 385: 252 /254. Kerr, J. T. and Currie, D. J. 1999. The relative importance of evolutionary and environmental controls on broad-scale patterns of species richness in North America. / E`coscience 6: 329 /337. Kerr, J. T. and Ostrovsky, M. 2003. From space to species: ecological applications for remote sensing. / Trends Ecol. Evol. 18: 299 /305. Lathman, R. E. and Ricklefs, R. E. 1993. Global patterns of tree species richness in moist forests: energy-diversity theory ECOGRAPHY 27:4 (2004)

does not account for variation in species richness. / Oikos 67: 325 /333. Leemans, R. and Cramer, W. P. 1992. IIASA Database for mean monthly values of temperature, precipitation, and cloudiness on a global terrestrial grid. / In: Anon. (ed.), Global Ecosystems Database ver. 2.0. Boulder, CO: NOAA National Geophysical Data Center, B/http:// www.ngdc.noaa.gov/seg/eco/cdroms/gedii_a/datasets/a03/lc. htm#top /. Legendre, P. 1993. Spatial autocorrelation: trouble or new pardigm? / Ecology 74: 1659 /1673. Lichstein, J. W. et al. 2002. Spatial autocorrelation and autorregressive models in ecology. / Ecol. Monogr. 72: 445 /463. Lyons, S. K. and Willig, M. R. 1999. A hemispheric assessment of scale dependence in latitudinal gradients of species richness. / Ecology 80: 2483 /2491. MacArthur, R. H. 1972. Geographical ecology. Patterns in the distribution of species. / Princeton Univ. Press. Marquet, P. A. 1989. Paleobiogeography of South American cricetid rodents: a critique to Caviedes and Iriarte. / Rev. Chil. Hist. Nat. 62: 193 /197. McCoy, E. D. and Connor, E. F. 1980. Latitudinal gradients in the species diversity of North American mammals. / Evolution 24: 193 /203. McNab, B. K. 1982. The physiological ecology of South American mammals. / In: Mares, M. A. and Genoways, H. H. (eds), Mammalian biology in South America. Vol. 6. Spec. Publ. Ser. Pymatuning Lab. Ecol. Univ. Pittsburgh, pp. 187 /208. Neuweiler, G. 1989. Foraging ecology and audition in echolocating bats. / Trends Ecol. Evol. 4: 160 /166. O’Brien, E. M. 1993. Climatic gradients in woody plant species richness: towards an explanation based on an analysis of Southern Africa’s woody flora. / J. Biogeogr. 20: 181 /198. O’Brien, E. M. 1998. Water-energy dynamics, climate, and prediction of woody plant species richness: an interim general model. / J. Biogeogr. 25: 379 /398. O’Brien, E. M., Field, R. and Whittaker, R. J. 2000. Climatic gradients in woody plant (tree and shrub) diversity: waterenergy dynamics, residual variation, and topography. / Oikos 89: 588 /600. Patterson, B. D. 1994. Accumulating knowledge on the dimensions of biodiversity: systematic perspectives on Neotropical mammals. / Biodiv. Lett. 2: 79 /86. Patterson, B. D., Pacheco, V. and Solari, S. 1996. Distributions of bats along an elevational gradient in the Andes of southeastern Peru. / J. Zool. 240: 637 /658. Paulsen, A. C. 1976. Environment and empire: climatic factors in pre-historic Andean culture change. / World Archaeol. 8: 121 /132. Pianka, E. R. 1966. Latitudinal gradient in species diversity: a review of concepts. / Am. Nat. 100: 33 /46. Rahbek, C. and Graves, G. R. 2001. Multiscale assessment of patterns of avian species richness. / Proc. Natl. Acad. Sci. USA 98: 4534 /4539. Redford, K. H. and Eisenberg, J. F. 1992. Mammals of the Neotropics. Vol 2. The Southern Cone. / Univ. of Chicago Press. Reig, O. A. 1986. Diversity patterns and differentiation of high Andean rodents. / In: Vuilleumier, F. and Monasterio, M. (eds), High altitude tropical biogeography. Oxford Univ. Press, pp. 404 /439. Ricklefs, R. E. and Schluter, D. 1993. Species diversity in ecological communities. Historical and geographical perspectives. / Univ. of Chicago Press. Rohde, K. 1992. Latitudinal gradients in species diversity: the search for the primary cause. / Oikos 65: 514 /527. Rohde, K. 1996. Rapoport’s rule is a local phenomenon and cannot explain latitudinal gradients in species diversity. / Biodiv. Lett. 3: 10 /13. Rosenzweig, M. L. 1992. Species diversity gradients: we know more and less than we thought. / J. Mammal. 73: 715 /730.

413

Rosenzweig, M. L. 1995. Species diversity in space and time. / Cambridge Univ. Press. Ruggiero, A. 1994. Latitudinal correlates of the sizes of mammalian geographical ranges in South America. / J. Biogeogr. 21: 545 /559. Ruggiero, A. 1999. Spatial patterns in the diversity of mammal species: a test of the geographic area hypothesis in South America. / coscience 6: 338 /354. Ruggiero, A., Lawton, J. H. and Blackburn, T. M. 1998. The geographic ranges of mammalian geographical ranges in South America. / J. Biogeogr. 25: 1093 /1103. Simpson, G. G. 1964. Species density of North American recent mammals. / Syst. Zool. 13: 57 /73. Smith, M. F. and Patton., J. L. 1993. The diversification of South American murid rodents: evidence from mitochondrial DNA sequence data for the akodontine tribe. / Biol. J. Linn. Soc. 50: 149 /177. Thompson, L. G., Mosley-Thompson, E. and Thompson, P. A. 1992. Reconstructing interannual climate variability from tropical and subtropical ice-core records. / In: Diaz, H. F. and Markgraf, V. (eds), El NinË&#x153;o. Historical and paleoclimatic aspects of the southern oscillation. Cambridge Univ. Press, pp. 295 /322. Turner, J. R. G., Lennon, J. and Greenwood, J. J. D. 1996. Does climate cause the global biodiversity gradient? / In: Hochberg, M. E., Clobert, J. and Barbault, R. (eds), Aspects of the genesis and maintenance of biological diversity. Oxford Univ. Press, pp. 199 /220.

414

Van Rensburg, B. J., Chown, S. L. and Gaston, K. J. 2002. Species richness, environmental correlates and spatial scale: a test using South African birds. / Am. Nat. 159: 566 /577. Waring, R. H. and Running, S. W. 1988. Forest ecosystems. Analysis at multiple scales, 2nd ed. / Academic Press. Whittaker, R. J., Willis, K. J. and Field, R. 2001. Scale and species richness: toward a general hierarchical theory of species diversity. / J. Biogeogr. 28: 453 /470. Willig, M. R. 2000. Latitude, trends with. / In: Levin, S. (ed.), Encyclopedia of biodiversity. Academic Press, pp. 701 /714. Willig, M. R. and Mares, M. A. 1989. A comparison of bat assemblages from phytogeographic zones of Venezuela. / In: Morris, D. W. et al. (eds), Patterns in the structure of mammal communities. Special Publication of the Museum of Texas Technology Univ., no. 28, pp. 59 /67. Willig, M. R. and Selcer, K. W. 1989. Bat species gradients in the New World: a statistical assessment. / J. Biogeogr. 16: 189 /195. Willig, M. R. and Lyons, S. K. 1998. An analytical model of latitudinal gradients of species richness with an empirical test for marsupials and bats in the New World. / Oikos 81: 93 /98. Wright, D. H. 1983. Species-energy theory: an extension of species-area theory. / Oikos 41: 496 /506. Zapata, F. A., Gaston, K. J. and Chown, S. L. 2003. Mid-domain models of species richness gradients: assumptions, methods and evidence. / J. Animal. Ecol. 72: 677 /690.

ECOGRAPHY 27:4 (2004)

415

EDE

b r2 T p

b r2

TMIN

PET

INST

0.31 0.61 0.39 0.0004

Best environmental predictors B 0.71 0.60 r2 0.72 0.71 T 0.28 0.29 p B/0.0001 B/0.0001

INDVI

AET

MAR

CHI

0.37 0.67 0.32 B/0.0001

Parameters of the full regression models N 165 154 155 166 0.77 0.63 0.56 0.90 R2 p B/0.0001 B/0.0001 B/0.0001 B/0.0001

TOT 124 0.72 B/0.0001

PRI

0.59 0.69 0.31 B/0.0001

165 0.57 B/0.0001

HYS

0.55 0.59 0.41 B/0.0001

166 0.67 B/0.0001

CAR

0.32 0.80 0.20 0.01

165 0.49 B/0.0001

ART

0.50 0.55 0.45 B/0.0001

165 0.55 B/0.0001

SCI

0.19 0.40

0.28 0.48 0.52 B/0.0001

165 0.65 B/0.0001

0.20 0.47 0.53 0.003

165 0.63 B/0.0001

0.37 0.47 0.53 B/0.0001

165 0.64 B/0.0001

0.65 0.76 0.24 B/0.0001

165 0.90 B/0.0001

Appendix 1. Test of energy, heterogeneity and stability hypotheses. Regression coefficients are given exclusively for environmental variables that remained significant in the presence of the best spatial descriptors. Significant spatial descriptors, and regression coefficients for environmental variables without considering the effect of spatial structuring of data, are not shown. N /number of valid cells in the grid map for each taxa or quartile; b /standardized (beta) regression coefficient; R2 / Coefficient of determination of the full regression models, including the effects of both environment and space; r2 /partial coefficient of determination; it indicates the relative importance of each selected variable in the presence of the other selected environmental and/or spatial descriptors; T /tolerance coefficient for each environmental descriptor in the presence of other environmental/spatial descriptors; P /error probability.

416

ECOGRAPHY 27:4 (2004)

b r2 T p

PLAND

0.24 0.19 0.80 B/0.0001

b r2 T p

EVAR

0.26 0.34 0.66 B/0.0001

/0.47 0.15 0.85 B/0.0001

PRI

b r2 T p

CHI

HETER

EDE

/0.42 0.12 0.87 B/0.0001

MAR

AMPLIT b r2 T p

T p

TOT

Appendix 1. (Continued).

0.31 0.19 0.81 B/0.0001

0.23 0.28 0.72 0.0003

HYS

0.31 0.42 0.57 B/0.0001

CAR

0.22 0.14 0.86 0.0005

ART

0.64 0.48 0.51 B/0.0001

SCI

0.58 0.42 0.58 B/0.0001

0.60 0.001

0.56 0.40 0.59 B/0.0001

0.38 0.34 0.66 B/0.0001

ECOGRAPHY 25: 25–32, 2002

Elevational gradients in ant species richness: area, geometry, and Rapoport’s rule Nathan J. Sanders

Sanders, N. J. 2002. Elevational gradients in ant species richness: area, geometry, and Rapoport’s rule. – Ecography 25: 25 – 32. Studying the distributions of plants and animals along environmental gradients can illuminate the factors governing and maintaining species diversity. There are two general predictions of how species richness and elevation are related: either species richness decreases monotonically with increasing elevation or richness peaks at mid-elevations. Several processes might contribute to this pattern. In this paper, I examine patterns in ant species richness along elevational gradients in three states in the western US: Colorado, Nevada, and Utah. I test for the effects of available area and the geometric constraints model on species richness patterns. I also test Rapoport’s rescue hypothesis, which relates the extent of species’ elevational ranges to patterns in species richness. In each state, species richness peaked at mid-elevations. Area explained more variation in species richness than the geometric constraints model in Colorado and Utah, but not in Nevada. Area and geometric constraints together explained 90%, 99%, and 57% of the variation in species richness in Colorado, Nevada, and Utah, respectively. Even though there were peaks at midelevations, I still found a strong Rapoport effect. This work suggests that the influences of area and geometric constraints cannot be overlooked when examining patterns in species richness along environmental gradients. N. J. Sanders (njs12@humboldt.edu), Dept of Biological Sciences, 371 Serra Mall, Stanford Uni6., Stanford, CA 94305 -5020, USA (present address: Dept of Biological Sciences, Humboldt State Uni6., Arcata, CA 95521, USA).

A frequently documented ecological pattern is the relationship between species richness and elevation. Two general patterns emerge: a monotonic decrease in species richness with increasing elevation (e.g., MacArthur 1972, Stevens 1992) or a humped-shaped relationship, with a peak in richness at intermediate elevations (see Rahbek 1995 for a detailed review). Both patterns have been documented in a variety of habitats and taxa (e.g., Terborgh 1977, Stevens 1992, Brown 1995, Rahbek 1995, Rosenzweig 1995, Brown and Lomolino 1998), but Rahbek (1995, 1997) and others (Lees et al. 1999, Colwell and Lees 2000) have pointed out that perhaps mid-elevational peaks are more common. If peaks in species richness at mid-elevations are a common pattern, then the next step is to understand the mechanisms contributing to the pattern. Traditionally,

such explanations have focused on relating species richness to elevation via productivity. That is, productivity varies along elevational gradients, and productivity is the driving force behind patterns in richness; elevation merely serves as a surrogate for productivity. Species richness can be related to productivity in at least two ways: 1) as productivity increases, species richness increases monotonically (Hutchinson 1959, Preston 1962a, b, Connell and Orians 1964, MacArthur 1965, 1969, 1972, Brown 1988, Brown and Lomolino 1998) or 2) as productivity increases, species richness increases, peaks at mid-levels of productivity, and then decreases at high productivities (Tilman 1982, Rosenzweig and Abramsky 1993, Rosenzweig 1995). However, hard data demonstrating that elevation is a fair surrogate for productivity remain elusive (Rahbek 1997).

Like productivity, another important influence on species richness is area. In fact, one of ecology’s few laws is that large areas often support more species than do smaller areas (Schoener 1976). But few studies have examined the relationship among area, elevation, and species richness. For insects feeding on bracken in Britain, there was no relationship between richness and elevation after the effects of area were removed (Lawton et al. 1987). Rahbek (1997) found that when not accounting for the effect of area, there was a monotonic decrease in neotropical bird diversity with increasing elevation. However, when the influence of area was factored out, the relationship between diversity and elevation became hump-shaped with peaks at mid-elevations (Rahbek 1997). Colwell and Lees (2000) have suggested another hypothesis, the mid-domain effect, which seems to be very robust among different taxa. A mid-domain peak in richness is generated when there is increasing overlap of species ranges toward the center of the domain because the extent of the elevational ranges of species are bounded by the highest and lowest elevation possible in the region (Colwell and Hurtt 1994). Thus, regardless of variation in climatic variables such as productivity, a peak in species richness at mid-elevations may be due simply to the limits imposed by geographic boundaries. Another pattern along elevational gradients is a positive correlation between elevation and the elevational range of species; this pattern has been called Rapoport’s rule (Stevens 1992) or effect (Blackburn and Gaston 1996). It posits that climates at higher elevations are more variable, so species at higher elevations can tolerate more variability and therefore have larger elevational ranges. Richness is inflated at low elevations because of the proximity of nearby core areas for these ‘‘low elevation’’ species, but these species cannot persist at higher elevations. As a result, species richness decreases monotonically with elevation. This pattern has been documented in butterflies (Fleishman et al. 1998), trees, mammals, reptiles, some amphibians, and grasshoppers (references cited in Stevens 1992). For insects, there is considerable empirical evidence for both peaks in species richness at low elevations (e.g., Wolda 1987, Fernandes and Price 1988, McCoy 1990, Kearns 1992, Stevens 1992, Olson 1994, Sparrow et al. 1994) and peaks in species richness at intermediate elevations (Janzen 1973, McCoy 1990, Olson 1994, Sanchez-Rodriguez and Baz 1995, Fleishman et al. 1998). However, to my knowledge, only Lawton et al. (1987) have explored the relationship between insect species richness and elevation when area is controlled for and Fleishman et al. (1998) are the only researchers to have tested for Rapoport’s rule in insects. A few studies of ants have examined the effects of elevation on species richness, with differing conclu26

sions. Several studies demonstrated that there are fewer species at higher elevations than at lower elevations, or that there are no species above a certain elevation (Weber 1943, Brown 1973, Janzen 1973, Janzen et al. 1976, Collins 1980, Atkin and Proctor 1988). Species richness of leaf litter ants in a Malaysian rainforest decreased exponentially with increasing elevation (Bru¨ hl et al. 1999). Similarly, studies performed in a Panamanian rainforest (Olson 1994) and in Madagascar (Fisher 1996) showed a monotonic decrease in ant species richness with increasing elevation. In contrast, Fisher (1998), in another study in Madagascar, detected peaks in species richness at midelevations in leaf litter ants, and Samson et al. (1997) reported a peak in species richness at mid-elevations in forests in the Philippines. To date, no studies have examined the relationship among elevation and ant species richness while considering the effects of area and the mid-domain effect. To examine the relationship between elevation and species distributions, this study considers the distributions of ant species in the western US. The data on ant distributions come from extensive faunistic surveys of three states in the western US: Colorado (Gregg 1963), Nevada (Wheeler and Wheeler 1986), and Utah (Allred 1982). Although there are unknown biases in sampling effort, species identifications, and locality information with these three studies, these references include a comprehensive review of richness in these states and trustworthy data on locality from which most species were collected. I use the data on ant distributions from each of these states as a sample, so when I refer to the Colorado ants, I simply mean the data from Gregg’s monograph on Colorado ants; the ants do not respond differently just because they are from a particular state. Since the range of altitudes (ca 150–4400 m) in the three states is so great, it is possible to get a fair estimate of the elevational range size of each species. Furthermore, the influences of abiotic factors on species richness patterns are not confounded by latitudinal gradients because the states are at roughly the same latitudes (McCoy 1990). Here, I test the generality of patterns in ant species distributions along elevational gradients in two ways. First, I use data from intensive faunistic surveys of three states (Colorado, Nevada, and Utah). Second, I compare my results across subfamilies in the Formicidae. If the same pattern of distributions along elevational gradients exists among the three states and the most common subfamilies, then they are likely to be general mechanisms. Specifically, this study asks: 1) What is the influence of available area and geometric constraints on ant species richness patterns along elevational gradients in Colorado, Nevada, and Utah? 2) Do the patterns in richness differ among states? 3) Are the results consistent with Rapoport’s rule? ECOGRAPHY 25:1 (2002)

Methods Ant taxonomy has been extensively revised since the publication of these regional surveys I used. I elevated every subspecies listed in each of the three sources to the appropriate species-level taxon using Bolton (1994). In total, 226 species were identified and over 11 700 ant specimens were recorded in the three states. To examine the relationship between species richness and elevation, I divided the range of elevations into 100 m bands and found the total number of species in each band in each state. I assumed that each species was present in all bands between its highest and lowest reported elevations. I determined the relationship between area and elevation in each state by calculating the number of square kilometers in an elevational band. To calculate the area at each elevational band in Colorado, I used a Digital Elevation Model (DEM) from ESRI’s Arcview GIS Data and Maps 1999. For the Nevada data set, I used a 1:100 000 scale DEM from the Nevada Div. of Wildlife. To determine the area at each elevation in Utah, I used a DEM contour map with intervals at 500 ft and calculated the area at each interval. I used Colwell’s RangeModel software (Colwell 2000) to generate a null distribution predicted by the mid-domain effect based on 1000 runs of Ni species, where Ni is the number of species in each state i. The output of the simulation is the expected number of species at each elevation band. To examine the influence of area and the mid-domain effect on ant species richness patterns in each state, I determined for each state the relationship between species and area by log transforming the number of species in each elevational band and the area at each elevational band; this is the species-area curve for each state. I also log transformed the number of species predicted by the geometric constraints model. I used simple linear regressions with area and the predicted number of species alone as independent variables. I then used both area and the predicted number of species from the mid-domain effect in a multiple regression to examine their combined effects on species richness patterns. All p values reported here are simply indices of relative fit of the dependent and independent variables. They are not really ‘‘tests of statistical significance’’ because elevational bins are not independent (Colwell pers. comm.). To examine the relationship between the extent of the elevational range size of ants and elevation, I calculated the elevational range of each species in each state by subtracting the lowest elevation at which a species was collected from the highest elevation at which it was collected for each species collected at two or more elevations. I assumed that a species was present at all intervals between its highest and lowest recorded elevational distributions. To overcome statistical non-indeECOGRAPHY 25:1 (2002)

pendence of spatial data, I used the ‘‘midpoint method’’ as a measure of central tendency (Rohde et al. 1993); a midpoint for each species was calculated as the mean of the highest elevation and lowest elevation at which a species was collected. I used correlation analysis with a =0.05 to test for associations between elevation and elevational range size. In Figs 3 – 5, best fit lines are only to aid the reader’s eye; they are not meant to suggest causality. To test if the strength of the relationship between elevation and range size was independent of state and subfamily, I performed a test similar to a chi-square analysis using x2 = S(ni − 3)(zi − zw)2, where ni is size of sample i, zi is Fisher’s transformation (Fisher 1915), and zw is the weighted mean correlation coefficient for all samples. I also performed post hoc comparisons using Tukey tests to test for differences between each sample (Zar 1999).

Results Species richness for each state peaked at mid-elevations (Fig. 1). The total richness from the three states combined also peaked at mid-elevations. The numbers of species in each subfamily also peaked at mid-elevations, so there is likely to be no phylogenetic effect on ant distributions. The area in an elevational band was generally greatest at mid-elevations (Fig. 1). The species-area curves for each state are plotted in Fig. 2. Area explained a significant amount of the variation in each state (Colorado: y = 0.8969x− 1.8546, r2 = 0.75, p B0.0001; Nevada: y = 0.4028x+0.1222, r2 = 0.71, p B 0.0001; Utah: y=0.5234x− 0.5014, r2 = 0.53, p =0.0003). The patterns in species richness for each state are consistent with the mid-domain effect (Colorado: y = 0.261x+ 0.768, r2 = 0.13, p =0.04; Nevada: y = −1.30+ 1.72, r2 = 0.91, p B 0.0001; Utah: y = −1.21+ 1.70, r2 = 0.37, p =0.004). The combined effects of area and geometric constraints explained a considerable amount of the variation in species richness in each state (Colorado: r2 = 0.90, p B 0.0001; Nevada: r2 = 0.99, p B0.0001; Utah: r2 = 0.57, p =0.0008). The elevational extent of species tended to increase with increasing elevation, as Rapoport’s rule predicts (n =364, r = 0.453, p B0.001) (Fig. 3). Even though there is considerable scatter around the best fit line, there is a positive correlation between the elevational range size and the midpoint of the range size for most taxa; species at higher elevations had broader ranges. For each subfamily, elevational extent tended to increase with increasing elevation, as Rapoport’s rule 27

predicts (Dolichoderinae: n =18, r = 0.536, p B 0.05; Formicinae: n = 196, r =0.442, p B 0. 001; Myrmicinae: n =145, r =0.343, p B0.001) (Fig. 4). The responses for each taxon were not significantly different

Fig. 2. Species-area curves for each state. Shown are the log transformed numbers of species and number of km2 in each elevational band in each state.

from one another (x2 = 1.57, DF =2, p\ 0.20), so there is probably little or no phylogenetic effect. The elevational extent of species in each state increased with increasing range size (Colorado: n = 137, r = 0.622, p B 0.001; Nevada: n =134, r =0.400, p B 0.001; Utah: n =93, r= 0.569, p B0.001) (Fig. 5), but states differ in the strength of the correlation between elevational range size and the midpoint of the range (x2 = 6.46, DF = 2, p B 0.05). The Rapoport effect for Colorado ant distributions is significantly different from that of Nevada (‘‘Tukey’’ test, q = 3.50, p B 0.05), but neither Colorado and Utah (q =0.84, p \ 0.05) nor Nevada and Utah (q =2.31, p \ 0.05) were significantly different.

Fig. 1. The relationship between species richness, area, the mid-domain effect, and elevation. In each figure, the filled circles show the observed number of species present, the open circles show the expected number of species predicted by the mid-domain effect, and the solid line shows the amount of available area in that elevational band.

Fig. 3. Rapoport effect on elevational ranges of ants from all three states combined. The line is the least squares linear regression line (y =0.5554x +60.916). ECOGRAPHY 25:1 (2002)

Fig. 4. Rapoport effect on elevational ranges of the three most abundant subfamilies, the Dolichoderinae, Formicinae, and Myrmicinae. The line is the least squares linear regression line. (Dolichoderinae: y = 1.1739x−681.79, Formicinae: y = 0.5697x + 111.44, Myrmicinae: y =0.3744x+240.01.)

Discussion Ant species richness peaked at mid-elevations in each state (Fig. 1). In this study, over 90% of the variation in species richness along elevational gradients in Colorado and Nevada was explained by available area and geometric constraints, and 57% of the variation in species richness in Utah was explained by area and geometric constraints.

Fig. 5. Rapoport effect on elevational ranges in each state. The line is the least squares linear regression line. (Colorado: y = 1.0112x − 1038.8, Nevada: y = 0.3802x+ 440.1, Utah: y = 0.8931x − 408.39.) ECOGRAPHY 25:1 (2002)

Reports of mid-elevation peaks in richness are common in the literature (e.g., Wolda 1987, Fernandes and Price 1988, McCoy 1990, Kearns 1992, Stevens 1992, Olson 1994, Sparrow et al. 1994, see Rahbek 1995 for a review), and such mid-elevation peaks are probably the rule rather than the exception. Several hypotheses have been suggested to explain mid-elevation peaks in richness. The ‘‘ends are bad’’ hypothesis states that distributions are limited by climatic severity and reduced availability of resources at upper elevations, and by climatic severity and predation at lower elevations (McCoy 1990). A second hypothesis is the ‘‘middle is good’’ hypothesis which posits that productivity is highest at mid-elevations because daytime temperatures allow for higher photosynthesis rates, and cool evenings allow for lower plant respiratory rates (Janzen 1973, Janzen et al. 1976). A third hypothesis is that lower elevations receive more disturbance, thereby reducing species diversity at lower elevations. A final hypothesis directly relates species richness along elevational gradients to productivity: as productivity increases, diversity first increases then declines, giving a hump-shaped pattern (Rosenzweig and Abramsky 1993 and references therein). Many of these hypotheses are difficult to test. But authors have tested them, and results are equivocal. For example, some studies on the effects of habitat disturbance on ant species richness have shown that disturbance reduces richness or diversity (e.g., Greenslade and Greenslade 1977), while others have demonstrated little or no effect of disturbance (Room 1975, Torres 1984). It is unclear what the relationship between productivity and elevation is (Rahbek 1997), and studies to date on the relationship between productivity and ant species richness are equivocal. Desert ant species richness is positively correlated with productivity in North America (Davidson 1977) and negatively correlated in Australia (Morton and Davidson 1988) and South America (Medel 1995). In the most thorough study to date on the relationship between productivity and ant species richness, Kaspari et al. (2000) showed that ant species richness is positively correlated with productivity in 15 habitats throughout North America. The results reported in this study point to two other mechanisms creating hump-shaped patterns in species richness along elevational gradients: area and the mid-domain effect. At the regional scale in Colorado, Nevada, and Utah, there is more area at midelevations (Fig. 1). And the hard boundaries of the highest mountain tops and lowest elevations in states limit the range sizes of ant species, thereby generating a peak in species richness at mid-elevations. Rosenzweig (1995) notes that if you sample a bigger area, you will find more species. He also lists several processes that contribute to this pattern. For 29

the ants of Colorado, Nevada, and Utah, it seems likely that larger areas simply have higher habitat diversities, though I did not test this hypothesis. Several recent studies (see Colwell and Lees 2000) have shown that the mid-domain effect is common among many different taxa. Area explained more of the variation in species richness than did the middomain effect in Colorado and Utah but not Nevada. However, the peak in species richness for Colorado and Utah was at a lower elevation than the peak predicted by the null model for each the ants in each state, and species richness was generally lower at higher elevations than the null model predicted (Fig. 1). These results are similar to Rahbek’s (1997) result on tropical bird richness and suggest that factors other than the mid-domain effect, such as available area, influence patterns in species richness along elevational gradients. In Nevada, the peaks in the observed number of species and predicted number of species occurred at the approximately the same elevation (Fig. 1). Many authors have equated the latitudinal gradient with the elevational gradient in species richness and argued that the underlying mechanisms are the same for both (Stevens 1989, 1992). Rahbek (1995) points out that though species richness patterns might be the same along elevational and latitudinal gradients, the underlying mechanisms need not be the same. However, both available area (e.g., Rosenzweig 1995) and the mid-domain effect (e.g., Colwell and Lees 2000) are important influences on the latitudinal gradient, just as they are on elevational gradients in ant distributions reported in this study. The elevational range sizes of ants from Colorado, Nevada, and Utah increase with increasing elevation; this agrees with Rapoport’s rule. Distributions of ants in the three most common subfamilies, Myrmicinae, Formicinae, and Dolichoderinae, show this pattern. Interestingly, range sizes of ants from different states respond differently to increased elevation, though all are positive. Why does Rapoport’s rule not generalize across three samples? Differences among states in the strength of the Rapoport effect undoubtedly result from differences in sampling intensities, variation in the structures of the habitat types sampled, and different levels of disturbance among sampling sites, or a combination of factors. Recall that Rapoport’s elevational rule, according to Stevens (1992), relates to the rescue effect and is presented as an explanation for monotonic decreases in species richness with increasing elevation. But the ants from Colorado, Nevada, and Utah all show peaks at mid-elevations. How can these apparently conflicting patterns be reconciled? Colwell and Hurtt (1994) and Rahbek (1995, 1997) have pointed out that the data Stevens (1989, 1992) presented in sup30

port of his version of Rapoport’s rule actually show a peak at mid-elevations (or latitudes for the latitudinal version of the rule). The distributions of ants reported here probably suffer from geometrical limits, and show a Rapoport effect by default (Colwell and Hurtt 1994), though I did not explicitly test this hypothesis. Perhaps Rapoport’s elevational rule describes a spurious effect, or, if true, helps to explain peaks in species richness at mid-elevations rather than monotonically decreasing richness with increasing elevation. So far, there have been too few studies on Rapoport’s rule for ecologists to make general conclusions about its generality or applicability (Gaston and Blackburn 1999). But, if anything, Rapoport’s rule is not general (Rohde et al. 1993, Rohde 1996). There is a lack of information on the elevational distributions of plants and animals, especially for such ecologically important organisms as ants. Most studies have been performed in the tropics along relatively short elevational gradients, and sampling regimes have varied considerably. This study shows that ant species richness is highest at mid-elevations in three states in the Western US as a result of more area at mid-elevations and the mid-domain effect. Other studies, on both ants and other taxa, have found very different results from those reported here. It could be very interesting to explore further the relationship between elevation and species richness for other taxa when the influence of area and the effects of geometric constraints are considered, though it is likely that area and geometry will be among the most important influences on species richness along elevational gradients. Acknowledgements – Conversations with Rob Colwell, Liz Hadly, Aaron Hirsh, Charley Knight, Taylor Ricketts, Michael Rosenzweig, and Chris Wheat were very enlightening. Comments from Deborah Gordon, Nicole Heller, Melodie McGeoch, Henrique Pereira, Veronica Volny, Chris Wheat and especially Liz Hadly and Carsten Rahbek greatly improved the text. John Fay shared his wealth of GIS knowledge and deserves a big thanks. Bill Case and Randall Phillips kindly shared data on area and elevation. A special thanks goes to Deane Bowers and Virginia Scott for introducing me to the ants (and other insects) of Colorado.

References Allred, D. M. 1982. Ants of Utah. – The Great Basin Nat. 42: 415 – 511. Atkin, L. and Proctor, J. 1988. Invertebrates in litter and soil on Volcan Barva, Costa Rica. – J. Trop. Ecol. 4: 307 – 310. Blackburn, T. M. and Gaston, K. J. 1996. Spatial patterns in the geographic range sizes of bird species in the New World. – Philos. Trans. R. Soc. Lond. B 351: 897 – 912. Bolton, B. 1994. Identification guide to the ant genera of the world. – Harvard Univ. Press. Brown, J. H. 1988. Species diversity. – In: Myers, A. A. and Giller, P. S. (eds), Analytical biogeography: an ECOGRAPHY 25:1 (2002)

integrated approach to the study of animal and plant distribution. Chapman and Hall, pp. 57 – 89. Brown, J. H. 1995. Macroecology. – Univ. of Chicago Press. Brown, J. H. and Lomolino, M. V. 1998. Biogeography. – Sinaur. Brown, W. L. 1973. A comparison of the Hylean and Congo-West African rain forest ant faunas. – In: Meggers, B. J., Ayensu, E. S. and Duckworth, W. D. (eds), Tropical forest ecosystems in Africa and South America: a comparative review. Smithsonian Inst. Press, pp. 161 – 185. Bru¨ hl, C. A., Mohamed, M. and Linsenmair, K. E. 1999. Altitudinal distribution of leaf litter ants along a transect in primary forest on Mount Kinabalu, Sabah, Malaysia. – J. Trop. Ecol. 15: 265 –267. Collins, N. M. 1980. The distribution of soil macrofauna on the west ridge of Gunung Mulu. – Oecologia 44: 263 – 275. Colwell, R. K. 2000. RangeModel a Monte Carlo simulation tool for assessing geometric constraints on species richness. Ver. 3. – User’s guide and application published at: http://viceroy.eeb.uconn.edu/rangemodel . Colwell, R. K. and Hurtt, G. C. 1994. Nonbiological gradients in species richness and a spurious Rapoport effect. – Am. Nat. 144: 570 –595. Colwell, R. K. and Lees, D. C. 2000. The mid-domain effect: geometric constraints on the geography of species richness. – Trends Ecol. Evol. 15: 70 –76. Connell, J. H. and Orians, E. 1964. The ecological regulation of species diversity. – Am. Nat. 98: 399 – 414. Davidson, D. W. 1977. Species diversity and community organization in desert seed-eating ants. – Ecology 58: 711 – 724. Fernandes, G. W. and Price, P. W. 1988. Biogeographical gradients in galling species richness. – Oecologia 76: 161 – 167. Fisher, B. L. 1996. Ant diversity patterns along an elevational gradient in the Reserve Naturelle Integrale d’Andringitra, Madagascar. – Fieldiana Zool. 85: 93 – 108. Fisher, B. L. 1998. Ant diversity patterns along an elevational gradient in the Reserve Special d’Anjanaharibe Sud and on the western Masoala Peninsula, Madagascar. – Fieldiana Zool. 90: 39 –67. Fisher, R. L. 1915. Frequency distributions of the values of the correlation coefficient in samples from an indefinitely large population. – Biometrika 10: 507 – 521. Fleishman, E., Austin, G. T. and Weiss, A. 1998. An empirical test of Rapoport’s rule: elevational gradients in montane butterfly communities. – Ecology 79: 2472 – 2483. Gaston, K. J. and Blackburn, T. M. 1999. A critique for macroecology. – Oikos 84: 353 –368. Greenslade, P. J. M. and Greenslade, P. 1977. Some effects of vegetation cover and disturbance on a tropical ant fauna. – Insectes Soc. 24: 163 –182. Gregg, R. E. 1963. The ants of Colorado. – Univ. of Colorado Press. Hutchinson, G. E. 1959. Homage to Santa Rosalia or why are there so many kinds of animals? – Am. Nat. 93: 145 – 158. Janzen, D. H. 1973. Sweep samples of tropical foliage insects: effects of seasons, vegetation types, elevation, time of day, and insularity. – Ecology 54: 687 –708. Janzen, D. H. et al. 1976. Changes in the arthropod community along an elevational transect in the Venezuelan Andes. – Biotropica 8: 193 –203. Kaspari, M., O’Donnell, S. and Kercher, J. R. 2000. Energy, density, and constraints to species richness: ant assemblages along a productivity gradient. – Am. Nat. 155: 280 – 293. Kearns, C. A. 1992. Anthophilous fly distribution across an elevation gradient. – Am. Midl. Nat. 127: 172 – 182. ECOGRAPHY 25:1 (2002)

Lawton, J. H., MacGarvin, M. and Heads, P. A. 1987. Effects of altitude on the abundance and species richness of insect herbivores on bracken. – J. Anim. Ecol. 56: 147 – 160. Lees, D. C., Kremen, C. and Andriamampianina, L. 1999. A null model for species richness gradients: bounded range overlap of butterflies and other rainforest endemics in Madagascar. – Biol. J. Linn. Soc. 67: 529 – 584. MacArthur, R. H. 1965. Patterns of species diversity. – Biol. Rev. 40: 510 – 533. MacArthur, R. H. 1969. Patterns of communities in the tropics. – Biol. J. Linn. Soc. 1: 19 – 30. MacArthur, R. H. 1972. Geographical ecology. – Princeton Univ. Press. McCoy, E. D. 1990. The distribution of insects along elevational gradients. – Oikos 58: 313 – 332. Medel, R. G. 1995. Convergence and historical effects in harvester ant assemblages of Australia, North America, and South America. – Biol. J. Linn. Soc. 55: 29 – 44. Morton, S. R. and Davidson, D. W. 1988. Comparative structure of harvester ant communities in arid Australia and North America. – Ecol. Monogr. 58: 19 – 38. Olson, D. M. 1994. The distribution of leaf litter invertebrates along a neotropical altitudinal gradient. – J. Trop. Ecol. 10: 129 – 150. Preston, F. W. 1962a. The canonical distribution of commonness and rarity. – Ecology 43: 185 – 215. Preston, F. W. 1962b. The canonical distribution of commonness and rarity. – Ecology 43: 410 – 432. Rahbek, C. 1995. The elevational gradient of species richness: a uniform pattern? – Ecography 18: 200 – 205. Rahbek, C. 1997. The relationship among area, elevation, and regional species richness in neotropical birds. – Am. Nat. 149: 875 – 902. Rohde, K. 1996. Rapoport’s rule is a local phenomenon and cannot explain latitudinal gradients in species diversity. – Biodiv. Lett. 3: 10 – 13. Rohde, K., Heap, M. and Heap, D. 1993. Rapoport’s rule does not apply to marine teleosts and cannot explain latitudinal gradients in species richness. – Am. Nat. 142: 1 – 16. Room, P. M. 1975. Diversity and organization of the ground foraging ant faunas of forest, grassland and tree crops in Papua New Guinea. – Aust. J. Zool. 23: 71 – 89. Rosenzweig, M. L. 1995. Species diversity in space and time. – Cambridge Univ. Press. Rosenzweig, M. L. and Abramsky, Z. 1993. How are diversity and productivity related? – In: Ricklefs, R. and Schluter, D. (eds), Species diversity in ecological communities: historical and geographical perspectives. Univ. of Chicago Press, pp. 52 – 65. Samson, D. A., Rickart, E. A. and Gonzales, P.C. 1997. Ant diversity and abundance along an elevational gradient in the Philippines. – Biotropica 29: 349 – 363. Sanchez-Rodriguez, J. F. and Baz, A. 1995. The effects of elevation on the butterfly communities of a Mediterranean mountain, Sierra de Javalmbre, central Spain. – J. Lepidopterists’ Soc. 49: 192 – 207. Schoener, T. W. 1976. The species-area relation within archipelagoes: models and evidence from island land birds. – In: Firth, H. J. and Calaby, J. H. (eds), Proc. of the XVI Int. Ornithol. Congr. Australian Acad. of Sci., pp. 629 – 642. Sparrow, H. R. et al. 1994. Techniques and guidelines for monitoring neotropical butterflies. – Conserv. Biol. 8: 800 – 809. Stevens, G. C. 1989. The latitudinal gradient in geographical range: how so many species coexist in the tropics. – Am. Nat. 133: 240 – 256. Stevens, G. C. 1992. The elevational gradient in altitudinal range: an extension of Rapoport’s latitudinal rule to altitude. – Am. Nat. 140: 893 – 911.

Terborgh, J. 1977. Bird species diversity on an Andean elevational gradient. – Ecology 58: 1007 –1019. Tilman, D. 1982. Resource competition and community structure. – Princeton Univ. Press. Torres, J. A. 1984. Niches and coexistence of ant communities in Puerto Rico: repeated patterns. – Biotropica 16: 284 – 295.

Weber, N. 1943. The ants of the Imatong Mountains, Anglo-Egyptian Sudan. – Bull. Mus. Comp. Zool. 93: 261 – 354. Wheeler, G. C. and Wheeler, J. 1986. The ants of Nevada. – Los Angeles County Mus. Nat. Hist. Wolda, H. 1987. Altitude, habitat and tropical insect diversity. – Biol. J. Linn. Soc. 30: 313 – 323. Zar, J. H. 1999. Biostatistical analysis. – Prentice-Hall.

ECOGRAPHY 25:1 (2002)

Ecography 32: 133 142, 2009 doi: 10.1111/j.1600-0587.2008.05507.x # 2009 The Authors. Journal compilation # 2009 Ecography Subject Editor: Jens-Christian Svenning. Accepted 19 June 2008

Relative importance of climate vs local factors in shaping the regional patterns of forest plant richness across northeast China Xiangping Wang, Jingyun Fang, Nathan J. Sanders, Peter S. White and Zhiyao Tang X. Wang, J. Fang (jyfang@urban.pku.edu.cn) and Z. Tang, Dept of Ecology, and Key Laboratory of Earth Surface Processes of the Ministry of Education, Peking Univ., Beijing 100871, China. N. J. Sanders, Dept of Ecology and Evolutionary Biology, Univ. of Tennessee, Knoxville, TN 37996, USA. P. S. White, Dept of Biology, Univ. of North Carolina, Chapel Hill, NC 27599-3280, USA.

Northeast (NE) China covers three climatic zones and contains all the major forest types of NE Asia. We sampled 108 forest plots in six nature reserves across NE China to examine the influence of climate and local factors (canopy seasonality, successional stage, topography and forest structure) on geographic patterns of plant richness. We analyzed the relative effects of different factors at two spatial scales: the regional scale (across both latitude and altitude) and the local scale (along the altitudinal gradient within site). Our results showed that the relative importance of climate vs local factors differed remarkably depending on scale and functional group. While total and tree species richness were mainly limited by climate, herb and shrub richness was more related to local factors (especially at the local scale). In the climatic factors, heat sum was the major correlate of tree, shrub and total species richness, while herb richness was more associated with winter coldness. Precipitation was not a limiting factor for forest plant richness in NE China. Climate accounted for 34 76% of variation in richness at the regional scale, but explained only 0 44% at the local scale. Among the local factors, shrub species richness was sensitive to seasonal canopy openness, with higher richness in deciduous forests than in the evergreen needle-leaf forest. On the other hand, herb richness was sensitive to forest successional stage, with higher richness in middle- successional forests than in the early and late-sucessional forests. Local topography (aspect and position on slope) and forest structure (tree density) also showed remarkable influence on species richness. Our results suggest the importance of including local factors when examining large scale diversity gradient (especially for understory species), and the necessity of comparing diversity patterns among functional groups at different spatial scales.

Understanding the underlying causes of geographic diversity gradients, especially in hyper-diverse ecoregions, is important for the conservation and sustainable use of biodiversity (Gaston 2000, Grytnes 2003). It is widely observed that species richness varies systematically with both latitude and altitude for a variety of taxa (Hillebrand 2004, Rahbek 2005). Though many mechanisms for these changes have been suggested, climate is widely reported to be closely correlated with species richness, and thus some authors suggested climate as the first-order predictor for species richness (Oâ&#x20AC;&#x2122;Brien 1993, Francis and Currie 2003, Hawkins et al. 2003). The relative importance of particular climatic factors might, however, vary geographically. For instance, temperature generally accounts for more variation in richness in cold regions, while water availability is more important for areas with high energy inputs (Oâ&#x20AC;&#x2122;Brien 1993, Hawkins et al. 2003, Kreft and Jetz 2007). Temperate East Asia has long been a focus in broadscale diversity studies (Latham and Ricklefs 1993, Qian and Ricklefs 2000). Significantly higher taxonomic richness in this area than other temperate regions in the world

has been repeatedly reported at different taxon levels. This Asian bias is hypothesized to be related to: 1) a longer history of direct connection with rich tropical species pools; 2) greater environmental heterogeneity; 3) higher beta diversity; and 4) less extinction during historical glaciations (Qian and Ricklefs 2000, Qian et al. 2005). However, most studies on broad-scale diversity patterns in East Asia have been conducted only for species richness within large grain sizes (i.e. areas 10 km2; Qian and Ricklefs 2000, Ricklefs et al. 2004). Data at more local scales, especially from the rich floras of temperate China, are urgently needed for comparisons with other temperate regions (Ricklefs et al. 2004). Diversity gradients may differ remarkably among plant groups, reflecting differences in the determinants for different groups (Pausas 1994, Austin et al. 1996, Leathwick et al. 1998). Comparisons among ecological or taxonomic groups are necessary to gain a more comprehensive understanding of diversity patterns (Lomolino 2001, Pausas and Austin 2001). Diversity at plot scale (e.g. 1 m2 1 ha) is affected by a variety of local factors in

133

addition to climate, such as disturbance, canopy openness, topography, forest structure and biotic interactions (Pausas 1994, Saha 2003, Quigley and Platt 2003, Laughlin et al. 2005). These local factors are particularly important for understory plants (Svenning and Skov 2002, Laughlin et al. 2005, Laughlin and Grace 2006). However, the relative roles of local environments have seldom been quantified systematically along both latitudinal and altitudinal gradients (but see Austin et al. 1996, Harrison et al. 2006). A major objective of this study was to examine the relative importance of climate vs local factors in shaping the geographic richness gradients, emphasizing the differences among tree, shrub and herb species. The processes determining diversity are scale dependent, and thus it is crucial to examine the relative roles of different factors at different scales (Rahbek 2005, Harrison et al. 2006). Accordingly, we examined the effects of various variables at two spatial scales: 1) the regional scale of northeast China (along both latitudinal and altitudinal gradients); 2) the local scale within study site (along the altitudinal gradient). In this analysis, we sampled 108 plots from six nature reserves across northeast (NE) China to examine the following three specific questions: 1) what are the relative roles of different climatic and local factors in limiting species richness in NE China? 2) Are patterns different

between the regional and the local scales? 3) Are patterns similar for tree, shrub and herb growth forms?

Materials and methods Study area Northeast China (115837? 13585?E, 38843? 53834?N) consists of Heilongjiang, Jilin, Liaoning provinces and the eastern part of Inner Mongolia Autonomous Region, covering an area of 1 240 000 km2. The climate is controlled by high latitude East Asia monsoons, changing from warm temperate, temperate to cool temperate zones from south to north, and from humid, semi-humid to semiarid zones from east to west. These climatic gradients drive variation in vegetation types, from deciduous broadleaf forest over needle-leaf and broadleaf mixed forest to boreal forest and from forest over forest steppe to steppe, respectively (Zhou 1997). Geographically, NE China is characterized by plains separated by three major mountain systems (ChangbaiZhangguangcai Mountains, Xiaoxing’an Mountains, and Daxing’an Mountains; Fig. 1). With these great gradients in climate and topography, NE China is a major biodiversity center in East Asia (Chen 1998), and thus provides an ideal region for examining large scale diversity patterns across many taxa and their underlying causes.

Figure 1. Location of study sites in NE China, together with two climatic diagrams (Mt. Changbai and Mt. Baikalu). The study sites include six nature reserves: Mt. Changbai, Mt. Datudingzi, Mt. Mao’er, Liangshui, Mt. Baikalu and Genhe. The ‘‘(30 30)’’ in the climatic diagrams means that the temperature and precipitation were both means of 30 yr records.

134

Data collection The study sites were located in six well-protected nature reserves across the forest region of NE China: Mt. Changbai in Changbai Mountains, Mt. Mao’er and Mt. Datudingzi in Zhangguangcai Mountains, Liangshui in Xiaoxing’an Mountains, and Genhe and Mt. Baikalu in Daxing’an Mountains (Table 1, Fig. 1). These nature reserves cover all the ten major forest types of NE China, including: 1) Betula ermanii forest, a typical timberline forest in humid NE Asia; 2) Larix olgensis forest in the Changbai Mountains; 3) Picea jezoensis forest and 4) Abies nephrolepis forest in subalpine and upper montane of humid NE Asia; 5) Pinus koraiensis forest and 6) Pinus koraiensis and broadleaf mixed forest, the zonal forests of temperate NE Asia; 7) Betula platyphylla forest and 8) B. platyphylla-Populus davidiana mixed forest that occur widely; 9) deciduous broadleaf forest in the lower montane zone, a forest type dominated by Quercus mongolica, Tilia amurensis, Fraxinus mandshurica, Tilia mandshurica etc.; and 10) Larix gmelini forest, the zonal vegetation of cooltemperate NE Asia (Zhou 1997). For details on the species composition of these forests, see Wang et al. (2006b). In each of the six sites, we placed plots along the altitudinal gradient to cover all the vertical forest zones, with an altitudinal interval between plots of 50 100 m. A total of 108 plots were sampled during the summers of 2000 2003. The number of plots at each study sites differed depending on forest types and altitudinal ranges (Table 1). In each plot (20 30 m), vascular plant species were recorded and divided into tree, shrub and herb species. Nomenclature follows Fu (1995). Diameter at breast height (DBH) and tree height were measured for trees with DBH ]3 cm. Latitude, longitude, altitude, aspect, position on the slope (POS), and slope for each plot were measured in situ. Aspect was measured in degrees to real north, and transformed into the nine categories in Table 2. POS was recorded as four categories (Table 2). Climatic variables We estimated climatic variables for each of the plots using a well-established method (Tang and Fang 2006, Wang et al. 2006a). Monthly mean temperature and precipitation were estimated based on linear models using latitude, longitude and altitude as predictors. Details for the models and model validation have been described in an earlier study and thus not presented here (Wang et al. 2006a). Climatic indices were calculated for each plot using the estimated monthly temperature and precipitation. Heat sum, water supply and winter coldness are commonly

suggested as crucial climatic limitation for plant biological activity, and thus, may be important in affecting species richness (Woodward and Rochefort 1991, Hawkins et al. 2003). Accordingly, the following three indices were selected for this study: 1) warmth index (WI, 8C month), an index of growing season heat sum. WI is defined as the sum of mean monthly temperatures above 58C for the months with a temperatures higher than 58C (Kira 1945); 2) mean temperature for the coldest month (MTCM, 8C), a surrogate for absolute minimum temperature (Prentice et al. 1992); and 3) annual precipitation (AP, mm). Other indices, including mean temperature for the warmest month (MTWM), cold index (Kira 1945), annual potential and actual evapotranspiration (PET and AET, respectively), and the moisture index of Thornthwaite (1948), were also estimated for each plot. However, they were highly correlated with one of the three indices mentioned above (e.g. r 0.99 between WI, PET and MTWM). Thus they were excluded from subsequent statistical analyses to avoid collinearity. PET was equal to AET for all our plots, suggesting no water deficit in our study region (Thornthwaite 1948, Francis and Currie 2003). Data analysis Total, tree, shrub and herb species richness and 12 explanatory variables were used in data analyses, including geographic, climatic, topographical, forest structural and forest type variables (Table 2). Tree density and biomass were used as forest structural parameters (Pausas 1994). Biomass for each plot was estimated with DBH and tree height using site and species specific allometric relationships (Wang et al. 2008). We classified the 10 forest types into four categories to examine the effect of seasonal canopy openness on species richness: evergreen need-leaf forest, needle-leaf and broadleaf mixed forest, deciduous needleleaf forest and deciduous broadleaf forest. We also classified the 10 forest types into early, middle and late successional forests to examine the influence of successional stages (Table 2). Species richness, slope and tree density were square-root transformed, while biomass and altitude were log-transformed prior to analyses. WI, AP, MTCM and latitude were not transformed, for transformation could not reduce skewness and kurtosis (Table 2). For each model, we further used diagnostic plots to check the normality and homoscedasticity of residuals. The effects of climatic and local factors on species richness were analyzed with general linear models (GLMs) and F-tests, using sequential (type-I) sums of squares

Table 1. Geographic, climatic and vegetation outlines of the six sites sampled in northeast China. Abbreviations: AMT, annual mean temperature; AP, annual precipitation. Site Mt. Changbai Mt. Datudingzi Mt. Mao’er Liangshui Mt. Baikalu Genhe

Latitude (8N)

Longitude (8E)

Altitude (m)

AMT (8C)

41o23? 42836? 44824? 45820? 45825? 47807? 47814? 51836? 50849? 50851?

126855? 12980? 128812? 127830? 127834? 128848? 128856? 123804? 121830? 121831?

500 2691 350 1669 300 805 280 707 450 1460 780 1142

7.3 4.9 2 4 2.8 0.3 5.6 1.2 5.4

AP (mm) 600 1340 550 650 724 680 360 500 450 550

Number of plots 65 8 6 13 10 6

135

136 Table 2. Descriptive statistics for species richness and explanatory variables used in the study. For variables that were square-root (*) or log (**) transformed in subsequent analysis, skewness and kurtosis were given for transformed data. Abbreviations: SD, standard deviation. WI, warmth index; AP, annual precipitation; MTCM, mean temperature for the coldest month; POS, position on slope. Max

Min

72 23 17 49

9 1 1 2

38.7 8.7 8.6 21.7

14.18 4.93 4.15 9.73

0.23 0.41 0.66 0.32

0.24 0.80 0.11 0.74

51.9 1945

42.1 344

44.5 964.6

3.35 464.82

1.21 0.10

0.02 1.08

Climate WI (8C month 1) AP (mm) MTCM (8C)

59.5 1132.4 16.6

17.8 469.6 29.2

40.3 787.6 20.6

13.52 190.00 3.67

-0.15 0.07 1.19

1.52 0.98 0.49

Forest structure Tree density (hm 2)* Biomass (mg hm 2)**

3116.7 663.7

383.3 56.1

1250.9 260.4

548.07 122.56

0.26 -0.19

0.15 0.32

30.0 0.0 5.7 7.36 Nine categories: N, NE, E, SE, S, SW, W, NW and flat land with no aspect Four categories: bottom, lower, middle and upper slope

0.50

-0.72

Richness (/plot)* Total Tree Shrub Herb Geographic variables Latitude (8N) Altitude (m)**

Topography Slope (8)* Aspect POS Forest type Canopy seasonality Successional stage

Mean

Skewness

Kurtosis

Four categories: evergreen need-leaf, needle-leaf and broadleaf mixed, deciduous needle-leaf, and deciduous broadleaf forest Three categories: early, middle and late successional forest 2

1 Evergreen need-leaf forests: Picea jezoensis forest and Abies nephrolepis forest; needle-leaf and broadleaf mixed forests: Pinus koraiensis forest and P. koraiensis and broadleaf mixed forest; deciduous needle-leaf forest: Larx olgensis forest and L. gmelini forest; and deciduous broadleaf forest: other forest types. 2 Early successional forest: Betula platyphylla forest and B. platyphylla-Populus davidiana mixed forest; middle successional forest: deciduous broadleaf forest; and late successional forest: other forest types (Zhou 1997).

(Schmid et al. 2002). We first analyzed the influence of different factors on regional-scale richness pattern (along both latitude and altitude gradients). The variables entering the model were selected by a forward stepwise procedure (Pausas 1994, Austin et al. 1996). We did not include interactions between variables and the quadratic terms of continuous variables (Austin et al. 1996, Kreft and Jetz 2007) in our final analyses, because many of them were not significant and only explained a very small amount of variation. In a second analysis, we entered site (the six study site) into GLMs before other variables to examine the roles of climatic and local factors in explaining the within-site variation of species richness (Schmid et al. 2002). In the present study, the within-site variation of richness was mainly caused by the altitudinal gradient, for we sampled plots along altitude in each site. We used this method to examine whether the within-site effects of the predictors were different from that at the regional scale. The variables entering the model were selected by the same procedure described above, except that this time we started forward selection from a model including site (instead of a null model in the first analysis) and added variables after site. Our plots were samples from six spatially separated study sites (Fig. 1). Hence, we also analyzed the data with linear mixed models, using site as random effect. However, the results were nearly the same as the GLMs fitting site before other variables (Supplementary material, Appendix S1 and S4 S6), and thus we used GLMs in the final analyses. Spatial autocorrelation in geographic diversity data can inflate type I errors in statistical analyses (Lennon 2000, Diniz-Filho et al. 2003). Hence, for bivariate relationships, we tested the significance using the modified t-test of Dutilleul et al. (1993). For multivariate models, Dutilleulâ&#x20AC;&#x2122;s method was used to correlate the observed and estimated richness for each model to test the overall statistical significance of the model (Hawkins et al. 2007). We also calculated Moranâ&#x20AC;&#x2122;s I values for both species richness data and the residuals of the models, to examine how the spatial autocorrelation in species richness was explained by the predictor variables (Hawkins and Porter 2003, Diniz-Filho et al. 2003). Since the model residuals did not contain significant positive spatial autocorrelation at the short distances, the influence of spatial autocorrelation on F-tests was very slight (Hawkins and Porter 2003). The analyses were conducted with the software R 2.6 (R Development Core Team 2007), SAM 3.0 (Rangel et al. 2006) and the Mod_t_test program (Legendre 2000).

Results Regional patterns of species richness A total of 421 species were recorded in the 108 plots, belonging to 72 families and 205 genera, of which 137 were woody species (25 families and 55 genera), and 284 were herb species (50 families and 149 genera). The 137 woody species included 50 tree species (13 families and 22 genera) and 87 shrub species (18 families and 36 genera). A correlation analysis showed that, species richness generally decreased with increasing altitude and latitude (Table 3). However, shrub richness did not change remarkably with latitude, while herb richness was not significantly correlated with altitude. Relative effects of climate vs local factors at the regional scale We first examined the influence of different factors on species richness at the regional scale (along both latitude and altitude). Results of both correlation analysis and GLMs suggested that (Table 3 and 4, Supplementary material, Appendix S2), growing season warmth (WI) was the major correlate of total, tree and shrub species richness, which accounted for 66, 72 and 34% of variation, respectively (Fig. 2). However, winter coldness (MTCM) was more powerful in explaining herb richness (34%). Precipitation showed only a weak effect on species richness. Climatic factors accounted for 73 76% of variation in total and tree richness, and ca 34% for shrub and herb species (Table 4). Local factors together explained only 13 15% of variation in total and tree species richness, but explained 36 38% for shrub and herb species. Relative effects of climate vs local factors at the local scale By fitting site before other variables, we examined the effects of climate and local factors on the within-site variation of species richness (i.e. along altitudinal gradient) (Table 4, Supplementary material, Appendix S3). WI was the only climatic variable entering the models, and explained 35, 44 and 17% of variation in total, tree and shrub species richness, respectively. Interestingly, no climatic variable entered the model for herb species. Local factors together accounted for 16, 5, 16 and 41% of variation in total, tree, shrub and herb species richness,

Table 3. Correlations between geographic and climatic variables and species richness. Adj. p, p values corrected for spatial autocorrelation, WI, warmth index; AP, annual precipitation; MTCM, mean temperature for the coldest month. Richness

Total Tree Shrub Herb

Latitude

Altitude

MTCM

Adj. p

0.40 0.44 0.19 0.46

B0.001 0.010 0.090 0.021

0.57 0.56 0.68 0.22

0.040 0.039 0.008 0.303

0.81 0.85 0.60 0.48

0.002 0.001 0.021 0.025

-0.09 -0.09 -0.53 0.18

0.719 0.741 0.016 0.417

0.66 0.71 0.08 0.58

B0.001 B0.001 0.321 0.001

137

1.0 4.5 0.0 4.0 7.1 0.0 8.1 9.8 11.3 3.6 13.8 18.7 7.3 1.1 0.0 25.8 1.3 1.0 0.0 9.1 1.2 3.6 23.7 3.7 16.4 5.0 15.7 41.2

1.2 2.5 7.6 3.0

Local Region Local Region Local

Canopy seasonality explained 24 and 8% of variation in shrub richness at the regional and local scales, respectively (Table 4). However, it showed only weak effects on herb richness. To visualize the influence of canopy seasonality, we used the residuals of richness fitted to climatic variables (Supplementary material, Appendix S2) to analyze the difference among forest groups (Schmid et al. 2002). The results showed that, shrub richness was significantly lower in evergreen forest than in the deciduous forests when the effect of climate had been accounted (Fig. 3), suggesting a remarkable effect of seasonal canopy opening. However, no significant difference was observed for other species groups (p 0.05 in all cases). Successional stage accounted for 9 and 26% of variation for herb richness (Table 4). However, it explained little variation (B1.1%) for tree and shrub species. Middle successional forest had a significantly higher richness for total and herb species than the late and early successional forests (when the effect of climate had been accounted). However, this difference among successional stages was not observed for woody (tree and shrub) species at p B0.05 (Fig. 3). Local topographical factors (aspect and POS) accounted for 7 19% of variation in total, shrub and herb species richness, but had only weak effect on tree species. Slope did not show any explanatory power for species richness (Table 4, Supplementary material, Appendix S2 S3). Tree density showed a positive effect on tree richness but a negative effect on herb and total species richness (Supplementary material, Appendix S2 S3). Biomass of the tree layer had no effect on richness for all species groups. Spatial autocorrelation in species richness Species richness showed strong positive spatial autocorrelation at the short distance classes, and showed negative or positive autocorrelation at intermediate distance (Fig. 4). Climatic variables (Supplementary material, Appendix S2) reduced most of the spatial autocorrelation for tree and total species richness, but significant autocorrelation still remained for herb at some distance classes. Adding the local factors into the models significantly reduced the remaining spatial autocorrelation, especially for herb species (Fig. 4).

Discussion

Total Tree Shrub Herb

14.8 12.7 37.5 35.5

Species richness in relation to climate Region

Successional stage Canopy seasonality Local factors pooled

0.8 1.4 0.0 2.6

Local Local

Region

Influence of local factors on species richness

Region

Topography (aspect and POS)

0.0 4.0 0.0 0.0 0.0 0.0 0.0 0.0 7.2 0.0 0.0 33.8 35.4 43.9 17.3 0.0 65.5 71.5 34.0 0.0 35.4 43.9 17.3 0.0 72.7 75.5 34.0 33.8 Total Tree Shrub Herb

87.5 88.2 71.5 69.3

51.8 48.9 33.0 41.2

Region Local Region Local Region

Tree density

0.0 0.0 0.0 0.0

Local Local

Region

Local

Region

AP MTCM WI Climatic variables pooled Total %SS

Table 4. Summary for the effects of climate and local factors on species richness at two scales: the regional scale vs the local scale. This table was summarized from the models in Supplementary material, Appendix S2 S3. For each variable, percentage of sum of squares explained (%SS) was reported. Slope and biomass were excluded from all the models and thus not presented. Abbreviations: WI, warmth index; MTCM, mean temperature for the coldest month; AP, annual precipitation, POS, position on slope.

138

respectively. The within-site effects of all variables (except successional stage) were lower than that at the regional scale, especially for climatic variables (Table 4).

Many authors have pointed out that species richness is often strongly correlated with climate (Oâ&#x20AC;&#x2122;Brien 1993, Francis and Currie 2003, Hawkins et al. 2003, Kreft and Jetz 2007). However, most of the previous studies at a large scale were based on species richness within geographic grids (but see Gentry 1988, Austin et al. 1996, Leathwick et al. 1998,

Figure 2. Relationships between species richness and climatic indices. WI, warmth index; MTCM, mean temperature for the coldest month; Adj. p, significance corrected for spatial autocorrelation.

Harrison et al. 2006). In this analysis, we used data at plot scale to examine geographic diversity patterns in relation to climate. We found that most of the variation in total and tree species richness was explained by climatic variables at the regional scale (73 76%; Table 4). Of the climatic variables, heat sum was the major correlate of total and woody species richness both at the regional and the local scales, while herb richness was more related to winter coldness than heat sum. Contrary to other studies (Hawkins et al. 2003, Kreft and Jetz 2007), water availability is not a limiting factor for forest diversity in the cold humid forest region of NE China (Table 3, 4). These results support the hypothesis that species richness is more closely correlated with energy availability in regions without water deficit

(Pausas and Austin 2001, Hawkins et al. 2003, Kreft and Jetz 2007). There are two versions of the energy hypothesis: the productivity hypothesis and the ambient energy hypothesis (Hawkins et al. 2003). The former proposes that plant richness is primarily constrained by the limitation of solar energy and water availability on productivity, while the later suggests that higher biological activity and lower winter mortality at higher temperatures are major mechanisms for more species (see reviews in Hawkins et al. 2003). In this study, total and woody species richness were the most closely associated with WI (Table 3, 4). WI is a direct measure of ambient temperature and thus the result supports the ambient energy hypothesis (Hawkins et al. 2003). At the same time, WI is highly correlated with forest productivity in humid East Asia (Ohsawa 1995) and the result can also support the productivity hypothesis. However, herb richness was more related to winter coldness that heat sum (Table 3, 4), and thus our results provide more support to the ambient energy hypothesis. Differences among growth forms

Figure 3. Effects of (a) canopy seasonality and (b) successional stage on understory species richness. Residuals of richness fitted to climatic variables (Supplementary material, Appendix S2) were used for analyses to examine the difference among forest groups when the effect of climate on richness had been accounted (Schmid et al. 2002). Grey dots indicate mean values. Forest groups that shared a same letter were not different at pB0.05 (p value adjustment method: Bonferroni). Abbreviations: DB, deciduous broadleaf forests; DN, deciduous needle-leaf forests; NBM: needle-leaf and broadleaf mixed forests; EN: evergreen needle-leaf forests.

Pattern of species richness is different for different functional groups (Pausas 1994, Pausas and Austin 2001). Differences among growth forms in this study can be summarized as the following three aspects. 1) Difference in geographical patterns. While total and tree species richness changed with both latitude and altitude, shrub richness did not change with latitude and herb richness did not vary with altitude significantly (Table 3). With higher latitudes in NE China, heat sum decreases, while climatic continentality increases (Zhou 1997). The later leads to an increasing seasonal canopy opening, which in turn results in higher shrub diversity (Fig. 3) (Specht and 139

Figure 4. Correlograms of Moranâ&#x20AC;&#x2122;s I showing patterns of spatial autocorrelation of species richness and residual autocorrelation after sequentially adding climatic variables and local factors (local) into the models. For variables used, see Supplementary material, Appendix S2. Filled circles (squares or triangles) indicate significant Moranâ&#x20AC;&#x2122;s I values (pB0.05), while open circles (squares or triangles) denote non-significant values.

Specht 1993, Quigley and Platt 2003). On the other hand, decreasing heat leads to lower species richness (Fig. 2). The opposite effects of heat sum and continentality may explain why shrub richness does not show a significant change along the latitudinal gradient. 2) Difference in the relative effects of climate vs local factors. From our results, it is clear that total and tree species richness are mainly limited by climate, while herb richness is more related to non-climatic factors (especially at the local scale; Table 4). The influence of climate on understory richness is mediated by the tree layer. Local scale studies have revealed that abiotic factors and disturbance affected understory richness via tree layer through various pathways, with the relative effects of different mechanisms changed with forest structure and tree species (Weiher 2003, Laughlin et al. 2005, Laughlin and Grace 2006). Thus it is natural that understory richness is much less related to climate compared with tree richness. Coincide with this, our results also showed that, climate was the main cause of spatial autocorrelation in tree richness, while the 140

spatial structures in shrub and herb richness were shaped by both climate and local factors (Fig. 4). 3) Difference in the relative effects of different local factors. An interesting finding of this study is that, while shrub richness was sensitive to seasonal canopy openness, herb richness was sensitive to forest successional stage (Fig. 3). Canopy seasonality is an important mechanism affecting plants under canopy in addition to gap dynamics (Quigley and Platt 2003), and we also observed significant difference between deciduous and evergreen forests for shrub species richness. The lack of sensitivity of herb richness to canopy seasonality may be caused by the fact that deciduous forests generally possess a denser shrub layer (unpubl.). Consequently, herb species can not benefit from seasonal canopy opening. Herb and total species richness was the highest in the middle-successional forests, consistent with other studies along the successional gradients (Pausas and Austin 2001, Saha 2003) and the hypothesis that community richness is higher under a medium disturbance (Connell 1978).

However, it remains unclear why shrub and tree richness were not different among sucessional stages in our study, which deserves further investigation. Some authors suggested that the remarkable increases of herb abundance caused by disturbance is an important mechanism affecting understory richness, and found that annual herb richness are more sensitive than richness of other species groups to habitat heterogeneity created by fire (Laughlin et al. 2005, Laughlin and Grace 2006). Pausas et al. (1999) also showed that herb species had a higher beta diversity than woody species in post-fire shrublands. It seems that different functional groups differ in their sensitivity to disturbance and sucessional stage, and these differences may be scale dependent. Tree density showed a positive effect on tree richness (Supplementary material, Appendix S2 S3), reflecting remarkable influence of sampling effect on species richness. However, tree density had a negative effect on herb richness, which reflect the effect of canopy tree shading and perhaps also other factors associated with tree density, e.g. water and mineral resources and litter depth (Laughlin and Abella 2007). However, the influence of trees on understory plants is very complex. For an example, Laughlin et al. (2005) found that understory richness was negatively related to subalpine fir basal area, but positively related to Engelmann spruce basal area. These complex effects may be a reason why tree layer biomass had no explanatory power for species richness in this study. The importance of local factors on geographic diversity gradients Climate has long been recognized as major correlate for large scale diversity patterns, and it is also well known that community richness is affected by a variety of local factors at the local scale. However, the relative importance of climate vs local factors has rarely been quantified systematically at a large scale (but see Austin et al. 1996, Harrison et al. 2006). Our results showed that local factors played an important role in explaining geographic diversity gradients, especially for understory species (Table 4). At the local scale, no climatic variable entered the model for herb species (Supplementary material, Appendix S3). Though we can not conclude that climate has no effect on herb richness along the altitudinal gradient, it is clear that the influence of local factors is far more powerful. This result is consistent with our previous findings that herb richness generally did not change significantly with altitude, and was more related to local factors (e.g. tree density) rather than estimated temperature (Zhao et al. 2004, Feng et al. 2006). In New Zealand, OhlemuÂ¨ ller and Wilson (2000) also found that both latitude and altitude had no explanatory power for herb richness, while the difference in dominant canopy trees showed a significant effect. Even for tree species, Austin et al. (1996) had demonstrated that local environment should not be ignored in any analysis of geographic richness patterns (see also Table 4). Our results, together with these studies, suggest the importance of examining diversity patterns using multivariate gradients to compare the differences among functional groups (Pausas and Austin 2001). We also suggest more studies on

community richness to be conducted across both latitude and altitude to examine the relative roles of different mechanisms at different scales (Austin et al. 1996, Grytnes 2003, Harrison et al. 2006). Acknowledgements Many members in the Terrestrial Ecology Group of Peking Univ. have participated in the field work. We specially thank B. Zhu, S. L. Piao, S. Q. Zhao, X. P. Wu and H. H. Shen for assistance in data collection, and Z. J. Zong for his assistance in species identification. We also thank the anonymous referees and B. Schmid for their valuable suggestions and comments. This study was supported by the National Natural Science Foundation of China (40638039, 40228001), and the China Postdoctoral Science Foundation (20070410021).

References Austin, M. P. et al. 1996. Patterns of tree species richness in relation to environment in south-eastern New South Wales. Aust. J. Ecol. 21: 154 164. Chen, C. D. (ed.) 1998. Chinaâ&#x20AC;&#x2122;s biodiversity: a country study organized by State Environmental Protection Administration. China Environmental Science Press. Connell, J. H. 1978. Diversity in tropical rain forests and coral reefs. Science 199: 1302 1310. Diniz-Filho, J. A. F. et al. 2003. Spatial autocorrelation and red herrings in geographical ecology. Global Ecol. Biogeogr. 12: 53 64. Dutilleul, P. et al. 1993. Modifying the t test for assessing the correlation between two spatial processes. Biometrics 49: 305 314. Feng, J. M. et al. 2006. Altitudinal patterns of plant species diversity and community structure on Yulong Mountains, Yunnan, China. J. Mountain Sci. 24: 110 116. Francis, A. P. and Currie, D. J. 2003. A globally consistent richness-climate relationship for angiosperms. Am. Nat. 161: 523 536. Fu, P. Y. (ed.) 1995. Clavis Platarum Chinae Boreali-Orientalis. Science Press. Gaston, K. J. 2000. Global patterns in biodiversity. Nature 405: 220 227. Gentry, A. H. 1988. Changes in plant community diversity and floristic composition on environmental and geographical gradients. Ann. Mo. Bot. Gard. 75: 1 34. Grytnes, J. A. 2003. Species-richness patterns of vascular plants along seven altitudinal transects in Norway. Ecography 26: 291 300. Harrison, S. et al. 2006. Regional and local species richness in an insular environment: serpentine plants in California. Ecol. Monogr. 76: 41 56. Hawkins, B. A. and Porter, E. E. 2003. Relative influences of current and historical factors on mammal and bird diversity patterns in deglaciated North America. Global Ecol. Biogeogr. 12: 475 481. Hawkins, B. A. et al. 2003. Energy, water, and broad-scale geographic patterns of species richness. Ecology 84: 3105 3117. Hawkins, B. A. et al. 2007. Red herrings revisited: spatial autocorrelation and parameter estimation in geographical ecology. Ecography 30: 375 384. Hillebrand, H. 2004. On the generality of the latitudinal diversity gradient. Am. Nat. 163: 192 211. Kira, T. 1945. A new classification of climate in Eastern Asia as the basis for agricultural geography. Horticultural Inst. Kyoto Univ.

141

Kreft, H. and Jetz, W. 2007. Global patterns and determinants of vascular plant diversity. Proc. Nat. Acad. Sci. USA 104: 5925 5930. Latham, R. E. and Ricklefs, R. E. 1993. Global patterns of tree species richness in moist forests: energy diversity theory does not account for variation in species richness. Oikos 67: 325 333. Laughlin, D. C. and Grace, J. B. 2006. A multivariate model of plant species richness in forested systems: old-growth montane forests with a long history of fire. Oikos 114: 60 70. Laughlin, D. C. and Abella, S. R. 2007. Abiotic and biotic factors explain independent gradients of plant community composition in ponderosa pine forests. Ecol. Model. 205: 231 240. Laughlin, D. C. et al. 2005. Understorey plant community structure in lower montane and subalpine forests, Grand Canyon National Park, USA. J. Biogeogr. 32: 2083 2102. Leathwick, J. R. et al. 1998. Environmental correlates of tree alpha-diversity in New Zealand primary forests. Ecography 21: 235 246. Legendre, P. 2000. Program Mod_t_test. Dept de Sciences Biologiques, Univ. de Montreal, /<http://www.fas.umontreal. ca/BIOL/legendre//>. Lennon, J. J. 2000. Red-shifts and red herrings in geographical ecology. Ecography 23: 101 113. Lomolino, M. V. 2001. Elevation gradients of species-density: historical and prospective views. Global Ecol. Biogeogr. 10: 3 13. O’Brien, E. M. 1993. Climatic gradients in woody plant species richness: towards an explanation based on an analysis of Southern Africa’s woody flora. J. Biogeogr. 20: 181 198. Ohlemu¨ ller, R. and Wilson, J. B. 2000. Vascular plant species richness along latitudinal and altitudinal gradients: a contribution from New Zealand temperate rainforests. Ecol. Lett. 3: 262 266. Ohsawa, M. 1995. Latitudinal comparison of altitudinal changes in forest structure, leaf-type, and species richness in humid monsoon Asia. Vegetatio 121: 3 10. Pausas, J. G. 1994. Species richness patterns in the understorey of Pyrenean Pinus sylvestris forest. J. Veg. Sci. 5: 517 524. Pausas, J. G. and Austin, M. P. 2001. Patterns of plant species richness in relation to different environments: an appraisal. J. Veg. Sci. 12: 153 166. Pausas, J. G. et al. 1999. Post-fire regeneration patterns in the eastern Iberian Peninsula. Acta Oecol. 20: 499 508. Prentice, I. C. et al. 1992. A global biome model based on plant physiology and dominance, soil properties and climate. J. Biogeogr. 19: 117 134. Qian, H. and Ricklefs, R. E. 2000. Large-scale processes and the Asian bias in species diversity of temperate plants. Nature 407: 180 182. Qian, H. et al. 2005. Beta diversity of angiosperms in temperate floras of eastern Asia and eastern North America. Ecol. Lett. 8: 15 22.

Download the Supplementary material as file E5507 from / <www.oikos.ekol.lu.se/appendix>.

142

Quigley, M. F. and Platt, W. L. 2003. Composition and structure of seasonally deciduous forests in the Americas. Ecol. Monogr. 73: 87 106. R Development Core Team 2007. R: a language and environment for statistical computing. R Foundation for Statistical Computing. Rahbek, C. 2005. The role of spatial scale and the perception of large-scale species-richness patterns. Ecol. Lett. 8: 224 239. Rangel, T. F. L. V. B. et al. 2006. Towards an integrated computational tool for spatial analysis in macroecology and biogeography. Global Ecol. Biogeogr. 15: 321 327. Ricklefs, R. E. et al. 2004. The region effect on mesoscale plant species richness between eastern Asia and eastern North America. Ecography 27: 129 136. Saha, S. 2003. Patterns in woody species diversity, richness and partitioning of diversity in forest communities of tropical deciduous forest biome. Ecography 26: 80 86. Schmid, B. et al. 2002. The design and analysis of biodiversity experiments. In: Loreau, M. et al. (eds), Biodiversity and ecosystem functioning: synthesis and perspectives. Oxford Univ. Press, pp. 66 75. Specht, A. and Specht, R. L. 1993. Species richness and canopy productivity of Australian plant communities. Biodiv. Conserv. 2: 152 167. Svenning, J.-C. and Skov, F. 2002. Mesoscale distribution of understorey plants in temperate forest (Kalø Denmark): the importance of environment and dispersal. Plant Ecol. 160: 169 185. Tang, Z. and Fang, J. 2006. Temperature variation along the northern and southern slopes of Mt. Taibai, China. Agric. For. Meteorol. 139: 200 207. Thornthwaite, C. W. 1948. An approach toward a rational classification of climate. Geogr. Rev. 38: 57 94. Wang, X. P. et al. 2006a. Climatic control of primary forest structure and DBH-height allometry in northeast China. For. Ecol. Manage. 234: 264 274. Wang, X. P. et al. 2006b. Climatic control on forests and tree species distribution in the forest region of northeast China. J. Integrative Plant Biol. 48: 778 789. Wang, X. P. et al. 2008. Forest biomass and root-shoot allocation in northeast China. For. Ecol. Manage. 255: 4007 4020. Weiher, E. 2003. Species richness along multiple gradients: testing a general multivariate model in oak savannas. Oikos 101: 311 316. Woodward, F. I. and Rochefort, L. 1991. Sensitivity analysis of vegetation diversity to environmental change. Global Ecol. Biogeogr. Lett. 1: 7 23. Zhao, S. Q. et al. 2004. Composition, structure and species diversity of plant communities along an altitudinal gradient on the northern slope of Mt. Changbai, northeast China. Biodiv. Sci. 12: 164 173. Zhou, Y. L. (ed.) 1997. Geography of the vegetation in northeast China. Science Press.