CROJFE - Volume 33, Issue 2 by Studio Bernardic

Croatia 2012

Original scietific paper – Izvorni znanstveni rad

Pre-Harvest Assessment based on LiDAR Data Hans Rudolf Heinimann, Jochen Breschan Abstract – Nacrtak Sourcing is the first line of competitiveness of timber supply networks, identifying and locat ing stands to be harvested that best fit to market demands. The sourcing process is difficult because information is either only available on an aggregated level or is even unavailable, e.g. on non-industrial forest owner’s land. Our study aimed to explore a LiDAR-data-based ap proach to improve the sourcing of stands to be harvested. We developed a spatially explicit approach, consisting of three steps: 1) harvest screening at the management unit scale or even larger, 2) location and delineation of cutting units, 3) characterization of tree attributes that are required for stand (cutting unit)-level bucking optimization. The study resulted in the following major findings. First, a tree map represented with Voronoi cells is a useful tool to support the harvest screening process, because it is easily readable and understandable, even by operations personnel. Second, cutting unit location and delineation can easily be done on the tree map, too. Third, the estimation of stem distribution over DBH of a cutting unit may easily be extracted from the spatial tree map database, assuming that there is a deterministic relationship between tree height and DBH. However, there are still issues to be improved, such as comparison of LiDAR-based results with ground-truth, the improvement of LiDAR-based tree delineation methods, the improvement of the estimation of stems over both DBH and tree height, or a mathematical formulation and solution of cutting unit layout. Keywords: LiDAR, tree delineation, Voronoi tessellation, harvest layout planning, pre-harvest assessment

1. Introduction – Uvod Supply chain management has been the dominating concept to coordinate all activities within a supply chain network, from »the sourcing« of stands to be harvested to the production of forest products, aiming to concurrently improve value recovery and to reduce supply cost. Recent work reported how supply chain activities may be optimized from the portfolio of stands to be harvested to a set of mils to be supplied with logs with mathematical techniques (Chauhan et al. 2011). However, this type of sophisticated approach is only feasible if reliable, accurate information on the stands to be harvested is available. In many cases there is only aggregated information for the stand level, whereas information about individual trees has been missing. The situation for nonindustrial forest owners is even worse; in many cases no useful information is available, and even government programs could not motivate those owners to increase the level of harvesting (Beach et al. 2005; Hyytiäinen and Penttinen 2008). Croat. j. for. eng. 33(2012)2

From a systems perspective there are three macro processes that operationally perform the supply of goods and services, sourcing, making, and delivering (Scor 2010). The sourcing process aims to 1) identify, locate, select and characterize supply sources, 2) to manage supplier networks and supplier agreements, and 3) to manage supply inventory (Scor 2010). The first step, the location, selection and characterization of supply points (stands to be harvested) depends on the availability of accurate, useful information. In terms of transaction cost economics the effort of information acquisition is called search cost (Pereira 2005), consisting of the cost of acquiring the information, and the opportunity cost for the searching time. We hypothesize that light detection and ranging (LiDAR) technology provides opportunities to improve the efficiency of sourcing stands to be harvested, primarily by increasing value recovery through improved matching of supply and demand. The present paper aims to explore the potential of LiDAR data,

169

H. R. Heinimann and J. Breschan

Pre-Harvest Assessment based on LiDAR Data (169–180)

particularly 1) to improve screening effectiveness for cutting units at the management unit level, 2) to evaluate alternative cutting unit layouts, and 3) to acquire information for cut-block-level bucking optimization, following ideas of (Chauhan et al. 2011; Laroze and Greber 1997). The paper first presents some background information on LiDAR technology and approaches to delineate trees and stands from LiDAR data, then develops a methodology to extract and present relevant information and finally presents a case study from a region in the Eastern Swiss Alps.

the state-of-the-art of full waveform LiDAR technology. It provides more accurate range detection, particularly over complex surfaces, but also offers an opportunity to extract additional parameters, such as dead branch crown base of trees, or the general characterization of the vertical structure of forest stands. However, we do not yet make use of the range of information of full-waveform data, particularly the profiles of intensity and amplitude.

2. Background – Pozadina istraživanja

Advances in processing techniques for LiDAR data offer the possibility to identify the location and the height of individual trees and of their crown geometry. The first processing step usually consists of surface smoothing, using filtering techniques, such as Gaussian Kernel filtering (Hyyppa et al. 2001; Morsdorf et al. 2004), followed by surface modeling techniques. Nine research groups compared the competitiveness of their tree extracting approaches on a joint data set (Kaartinen et al. 2008). Identification accuracy compared with ground truth was between 25% and 90%, and the so-called TopHat algorithm, a method of mathematical morphology, performed best. The TopHat algorithm is implemented in mathematical software tools, such as Matlab and Mathematica, and one can assume that other methods of mathematical morphology perform comparably or even better. Former TopHat studies came to the conclusion that the algorithm works well if tree crowns are relatively large and widely dispersed, but that its performance depends on the size of the morphological element (a circle) and the predominant tree crown size (Anderson et al. 2001). Another finding of the (Kaartinen et al. 2008) study was that an increase of laser pulse density from two points per square meter to eight points per square meter improved identification accuracy significantly. A recent study (Vauhkonen et al. 2012) compared tree extraction algorithms, however without investigating the performance of the TopHat algorithm, concluded that forest structure, particularly tree density and tree clustering affects algorithmic performance considerably. Traditionally, stand delineation is the process of experts interpreting aerial photographs, following best practice rules. However, those rules are not logical at all, but strongly depend on the skills of the analyst, the local nature of site conditions, and traditions of the local forest service (Sullivan 2008). There has been considerable attempt to automate the stand delineation process by segmenting digital imagery. Image segmentation is a process that divides an image into spa-

2.1 LiDAR Technology – Tehnologija LiDAR Light detection and ranging LiDAR is a remote sensing technology that creates a geo-referenced, 3D point cloud, representing a surface »sensed« by a laser sensor (Flood 2001). It is a »sister technology« of RADAR, radio wave detection and ranging, and was therefore also called LADAR, laser detection and ranging. Three functions have to be performed to yield geo-referenced coordinates of a surface, 1) ranging, providing the distance between the sensor and points of an object; 2) positioning, providing the coordinates of the sensor in space and time; 3) orientation, capturing the direction of the laser sensor at the time of laser emission. The first LiDAR systems were ground borne and used to detect scattering layers in the upper atmosphere (Fiocco and Smullin 1963). It was only during the 1980s when NASA developed experimental airborne LiDAR systems (Krabill et al. 1984), triggered by the availability of highly accurate global positioning system GPS, but it took about 10 years more until LiDAR technology started to becoming widely used, when commercial systems became available (Flood 2001). The application of airborne laser systems for forestry started by the end of the 1990s with the determination of terrain elevations, the estimation of stand height and volume, and the location and segmentation of individual trees (Hyyppä et al. 2004). At the same time papers providing a comprehensive overview on the technology and on the physical principles (Baltsavias 1999; Wehr and Lohr 1999) appeared. A review done in 2004 (Hyyppä et al. 2004) came to the conclusion that there is a relatively good understanding of extraction of digital terrain models DTMs and Crown Heights models CHMs, but that there are open issues such as the use of full waveform data, the complementary use of airborne and terrestrial laser scanning, and improved techniques to process 3D- point clouds. A more recent review (Mallet and Bretar 2009) presents

170

2.2 Extraction of Tree and Stand Attributes Izdvajanje stabala i opis sastojine

Croat. j. for. eng. 33(2012)2

Pre-Harvest Assessment based on LiDAR Data (169–180)

tially disjoint, homogenous areas (Mustonen et al. 2008). The development of automatic or semi-automatic stand segmentation methods is still an emerging field that has its roots – to our knowledge – in the work of the remote-sensing Institute of the University of Freiburg, Germany, in 2003 (Diedershagen et al. 2004; Diedershagen et al. 2003; Weinacker et al. 2004). A recent study from the same Institute reviews the stateof-the-art on stand segmentation (Koch et al. 2009). Available approaches are based on a normalized digital surface model that is converted into an image, which is analyzed with image segmentation techniques to delineate homogeneous areas. A similar procedure was used in Finland, confirming that image segmentation techniques based on crown height models perform better than the same techniques applied to spectral image data (Mustonen et al. 2008). However, there is still a strong need to improve stand segmentation methods, because the question how the abstraction process of aerial photograph interpreters is working, has not been conceded so far, and we still lack a consistent approach that produces comprehensible stand delineation results. Mountain forests, which are expected to provide ecosystem services, such as protection, have to be maintained in artificial steady-state equilibrium. Silvicultural interventions are aimed at triggering stand regeneration by imitating the so-called »gap dynamics«, which is a small-scale disturbance pattern of the forest canopy (Mccarthy 2001). As a consequence, there is a need to characterize forest gaps spatially explicit. To our knowledge, there are only few studies on gap identification (Vepakomma et al. 2008; Zhang 2008). A study on gap detection and mangrove forests resulted in the finding that morphological filtering clearly outperformed the so-called »height method« (Zhang 2008). Methods of mathematical morphology provide a high potential for the identification and characterization of regeneration gaps for uneven-aged management regimes and could be used for silvicultural priority assessment and protection service forests, as described by (Frehner et al. 2005).

3. Study Object and Methods – Objekt i metode istraživanja 3.1 Study Object – Objekt istraživanja The study object is located in eastern Switzerland and bounded by the following UTM (zone 32T) coordinates: west/south (563’500, 5’191’545); north/east (564’655, 5’192’455). The forests are located on a Northfaced steep slope at an altitude of 1300 to 1700 m above Croat. j. for. eng. 33(2012)2

H. R. Heinimann and J. Breschan

sea level. They belong to the subalpine vegetation zone that mainly consists of Norway spruce (Picea abies L. / Karst/). The rationale for the choice of the study object was threefold. First, it only consists of one tree species. Second, there is considerable amount of structural variability, both horizontally and vertically. And third, high-resolution LiDAR data were available.

3.2 LiDAR data – Podaci dobiveni iz LiDAR-ovih snimaka Airborne laser scanning was performed between September 11 and 15, 2010, with a Trimble Harrier 68 scanner (TRIMBLE, online). The Harrier 68 scanner is an advanced, pulsed laser corridor mapping system that generates extremely dense point clouds in combination with geo-referenced ortho images. It has a builtin full waveform digitization device, which enables the extraction of comprehensive vertical information from the acquired signals. The average flight height was about 700 m above ground, and the emitted laser point density was about four points per square meter. The laser scanner service provider did the data preprocessing, particularly the transformation from WGS 84 into the Swiss coordinate system and the extraction of the following data models: Þ DSM-FE (digital surface model first echo), providing the surface model including canopy, buildings, etc., Þ DSM-LE (digital surface model last echo), providing the terrain surface, however with »holes« for which no last echo response is available, Þ DTM (digital terrain model), providing the terrain surface without canopy, buildings, etc., Þ FDTM (filled digital terrain model), providing terrain surface, including »holes« filled by interpolation.

3.3 Workflow of and Tools for LiDAR processing Tijek rada i korišteni alati za analizu LiDARovih snimaka Post-processing of LiDAR data aims at extracting information that is useful for a specific type of problem. However, post-processing of LiDAR data to extract stand and tree information is a relatively new research topic. As a consequence, there is not »the one best way« for post processing. Our approach consists of the following analysis steps: 1) extraction of a crown height model CHM, 2) identification of canopy gaps, 3) extraction of tree attributes, and 4) the assessment of the harvesting corridor. Table 1 illustrates the workflow along the four main steps by allocating procedures and software tools. We used ArcGIS 10 and

171

H. R. Heinimann and J. Breschan

Pre-Harvest Assessment based on LiDAR Data (169–180)

Table 1 Pre-harvest assessment workflow. Toolboxes refer to ArcGIS 10 and to Matlab R2010a software Tablica 1. Tijek izrade plana sječe s pripadajućim opisom pojedinog postupka i korištenim alatima Step – Korak Extract crown height model CHM Izrada digitalnoga modela visine krošanja

Identification of canopy gaps Obilježavanje progala

Extract tree attributes Prikupljanje podataka o stablima

Procedure – Postupak

Toolbox – Alat

Import digital terrain model DTM and digital surface model DSM

ArcGIS

Unos digitalnoga modela terena i digitalnoga modela površine Subtract DTM from DSM to yield canopy height model CHM

ArcGIS Raster Calculator

Preklapanje digitalnih modela kako bi se dobio digitalni model visine krošanja Smooth CHM

ArcGIS Focal Statistics

Korekcija digitalnoga modela krošanja

(Gaussian Kernel Filter)

With CHM identify canopy gaps Pomoću digitalnoga modela krošanja prepoznati površinu progala

Matlab image toolbox, series of morphological opening and closing operations

With CHM, identify tree locations (local maxima) and assign tree heights ht

ArcGIS Focal Statistics,

Odrediti položaj stabla u sastojini pomoću digitalnoga modela krošanja te svakomu stablu dodijeliti odgovarajuću visinu

Raster Calculator

With tree height ht, estimate DBH, see equation (2)

ArcGIS Raster Calculator

Pomoću visine stabla izračunati prsni promjer (jednadžba 2) With ht, estimate tree volume Vt, see equation (1)

ArcGIS Raster Calculator

Pomoću visine stabla izračunati obujam stabla (jednadžba 1) With ht, DBH and Vt create GIS tree point layer

ArcGIS MultiValue to Point

Pomoću visine, prsnoga promjera i obujma napraviti bazu stabala u GIS-u With tree point layer create tree Voronoi cells

MatLab, Voronoi function

Pomoću GIS-ove baze napraviti Voronoiev dijagram Define cable road – Odrediti žičnu liniju Establish buffer for cable road with maximum lateral yarding distance Evaluate cable corridor Ocjena žične linije

Postaviti buffer s najvećom postranom udaljenosti privlačenja With tree point layer, select trees that are located within the buffer zone U GIS-ovoj bazi označiti stabla koja se nalaze unutar površine žične linije With selected trees, calculate DBH distribution Izračunati prsni promjer odabranih stabala

Matlab software to implement the series of procedures that are required to perform the four main steps. The first step, the extraction of the canopy height model CHM, is rather well understood (Hyyppä et al. 2004), it is the result of subtracting the digital terrain model DTM from the digital surface model, the DSM, and of interpolating empty grid nodes (Solberg et al. 2006). The second step, identification of canopy gaps, is essential for Norway spruce stands in higher elevations because trees are not distributed uniformly, but are clustered. Natural regeneration initiates in canopy gaps, assuming that there is enough light and heat getting on the ground to enable the growth of seedlings. We implemented an algorithm proposed by (Zhang 2008), which is based on methods of mathe-

172

ArcGIS line object Arc GIS buffer function Arc GIS spatial join function Export attribute file to EXCEL

matical morphology (alternating sequential filter), and which proved to be more flexible than other methods. Subtracting the gap area from the forest area results in the area in which trees may occur. The third step consists of identification of tree location, estimation of tree height and tree volume. Former studies reported that the crown surface model needs to be smoothed (Hyyppä et al. 2004; Solberg et al. 2006) to get close-to-reality results. Whereas mild smoothing produces a high share of »true trees«, but also many »false trees«, tough smoothing results in underestimating of »true trees« (Solberg et al. 2006). Therefore, the best setting for smoothing is usually identified by trial and error. We used a 3x3 Gaussian Kernel filter, the smoothness effect of which is slightly larger than the one of the filter proposed by (Hyyppa et al. 2001; Croat. j. for. eng. 33(2012)2

Pre-Harvest Assessment based on LiDAR Data (169–180)

Morsdorf et al. 2004). Locations of trees correspond to local maxima on the smoothed canopy height model CHM. A local maximum is a node that has larger height values than its eight grid neighbors (Solberg et al. 2006), and the cell centroid is an estimate for the location coordinates of that tree. Tree segmentation aims at defining a segment around each tree location to represent the crown footprint. A region growing algorithm (Solberg et al. 2006) proved to yield crown footprints with reasonable accuracy. For pre-harvest assessment we are interested in the base area that a single tree occupies. Computational geometry provides a method called Voronoi partitioning, a procedure to »partition the plane with points into convex polygons such that each polygon contains exactly one generating point and every point in a given polygon is closer to its generating point than to any other« (Weisstein online). Voronoi segmentation was used to investigate the influence of available plant space on yield (Mead 1966), and for spatially explicit stand modeling (Courne et al. 2009). Based on tree location points, and excluding canopy gaps, we created Voronoi cells for individual trees, which results in a tree map for a specific area. The advantage of such a map is that trees may easily be allocated to harvesting units or may be clustered into stands. The fourth step of the analysis consists of the preharvest assessment of a specific cutting unit. A cable corridor is the simplest geometric shape of a cutting unit, consisting of a rectangle with the length of the cable road and the width equaling two times the maximum lateral yarding distance. Once this shape is defined, a spatial query such as »select tree locations that are located within the cable corridor« yields a tree map with Voronoi cells, the total area of which equals the harvest area. The estimation of the harvest volume and the volume distribution over DBH may be done on the selected tree subset.

3.4 Tree attribute estimation – Procjena značajki stabala The traditional procedure measures DBH in the field, and then estimates both tree volume and height with empirical relationships, characterized with volume or height functions, respectively. Whereas DBH is the parameter that can easily be measured by ground surveying, height is the tree parameter that can be estimated most accurately from LiDAR data. Therefore, there is a need to have a volume estimation procedure with tree height as the main parameter. The Swiss National Forest Inventory SNFI (Brändli 2010) is based on permanent plots where information on DBH and tree height is gathered for a subset of all sample trees. We used SNFI tree data of the plots loCroat. j. for. eng. 33(2012)2

H. R. Heinimann and J. Breschan

cated in the valley of the test area, consisting of 756 tree data records, 135 of which with DBH and tree height measured in the field, and with tree volume estimated with SNFI volume functions. Our volume modeling approach followed the philosophy described by (Hoffmann 1982), using an exponential function with the natural logarithm of height and its forth power as independent variables [1]. (1) Vt represents tree volume in cubic meters, ht tree height in meters, extracted from the LiDAR data, whereas a, b, c and σ are model parameters. The second term of (1), , is s factor to correct for the bias resulting from logarithmic transformation, see (Beauchamp and Olson 1973). Regression analysis yielded the following parameter estimates: a = –9.85; b = 3.51; c = –0.0085; σ2 = 0.29. DBH estimation as function of tree height is based on a height prediction formula (Ye 1995), which was solved for tree height, resulting in (2). (2) The estimation of tree volume and DBH is deterministic, neglecting the residual variation of about 30% for (1).

4. Results and Discussion – Rezultati i rasprava Our case study aims at exploring a pre-harvest assessment methodology based on LiDAR data and at assessing its feasibility and utility. The first step of harvest planning consists of screening forest areas for possible cutting units that fulfill two basic requirements: 1) there is a silvicultural requirement for an intervention, and 2) the possible log mix fits as close as possible to market demand. The second step consists of laying out cutting units that are operationally feasible, economically efficient, and environmentally sound. The third and final step is to estimate the total harvest volume, the distribution of volume over DBH, and the possible distribution of log classes over small length diameter and length that could be produced from the standing trees. Below, we will illustrate those three steps for our study area.

4.1 Tree Location and Segmentation – Položaj i razvrstavanje stabala Traditional harvest screening is based on stand maps, aerial photographs and local knowledge of the forest guards and officers. Stand maps provide aggre-

173

H. R. Heinimann and J. Breschan

Pre-Harvest Assessment based on LiDAR Data (169–180)

Fig. 1 Tree Map characterized with Voronoi cells. Each cell represents a single tree classified into thicket, pole, timber and mature timber development stages. White polygons are gaps. Harvest screening yields areas for wood harvesting (timber 1, timber 2, mature timber stages) Slika 1. Karta stabala prikazana pomoću Voronoievih ćelija. Svaka ćelija predstavlja pojedinačno stablo razvrstano u koljik, stupovlje, pilansko drvo i prezrelo drvo. Bijeli poligoni označuju progale. Za sječu je predviđeno pilansko i prezrelo drvo gated information that is in many cases far too general for pre-harvest assessment. We hypothesize that the tree map (Fig. 1) is the type of information that is best supporting the screening for cutting units process. Whereas age class of stands is an important planning variable for even-aged forest management, it is usually not known for uneven-aged forests. Therefore, stand development stages are used, which may be defined in terms of the dominant tree height, e.g. the average height of the 100 most dominant trees per hectare. Fig. 1 illustrates a tree map extracted from LiDAR data with Voronoi cells for each tree and each tree classified according to the development stage. The development of the thicket stage may be manipulated by silvicultural tending operations, whereas development at the pole stage can be controlled by pre-commercial thinning. Harvesting operations for wood production occur at the timber development stages that are presented in Fig. 2 with three classes: timber 1, timber 2 and mature timber. A visual assessment of

174

Table 2 Characteristic Tree Attributes for the management unit. The management unit, part of which is shown in Fig. 1, consists of 22.1 ha stocked forest area. Results for DBH > 12 cm Tablica 2. Značajke stabala u istraživanoj gospodarskoj jedinici obrasle površine 22,1 ha (slika 1). Prikazani su rezultati za stabla prsnoga promjera > 12 cm Attribute

Mean

Min

Max

Std. Dev.

Značajke stabla

Arit. sred.

Min.

Maks.

St. dev.

28.3

11.4

51.9

7.0

2.7

0.2

5.5

1.3

140

Tree height, m Visina stabla, m DBH, cm Prsni promjer, cm Tree volume, m3 Obujam stabla, m3 Voronoi cell area, m2 Površina Voronoieve ćelije, m2

Croat. j. for. eng. 33(2012)2

Pre-Harvest Assessment based on LiDAR Data (169–180)

the tree map yields areas in which the timber development stage is predominant, those areas are candidates areas for cut block layout. It is not possible to identify all trees in a specific area (Vauhkonen et al. 2012), but dominant trees are located in the top canopy layer, in which tree identification seems to work quite accurately. Compared to previous approaches based on stand maps, the increase in information quality and accuracy is considerable and the error has to be accepted. Based on individual tree data it is possible to produce additional maps, providing information on volume density distribution or on stand density.

4.2 Tree Attribute Extraction – Izdvajanje značajki stabala We applied our screening approach to a management unit, consisting of 22.1 ha stocked forest area and a total standing volume of about 12,500 m³, corresponding to 565 m3.ha-1. Table 2 presents characteristic values for tree height, DBH, tree volume and Voronoi cell area. The figures illustrate that the management

H. R. Heinimann and J. Breschan

unit is characterized by considerable structural variability, covering tree height ranges between 11 m and 52 m, or tree volume ranges between 0.2 and 5.5 m³. About 80% of the standing volume consists of trees with a volume higher than 2.5 m³. Assuming that the volume will be harvested with a cable yarding system, the dominance of large-diameter trees will require a yarder with a load capacity higher than 25 kN. Tabular information (Table 2) has been used to characterize management units and forest stands. However, harvest unit layout is a spatial decision, resulting in the delineation of a cutting unit. This location decision has a big impact on operational efficiency, because the delineation fixes cutting unit characteristics such as 1) the total harvest volume, 2) distribution of trees over tree volume, and 3) harvesting intensity. Harvesting intensity is an important characteristic for cable yarding with considerable amount of set up and dismantling cost, which is usually measured in cubic meters per unit length of the cable road. The higher the volume per set up, the lower the harvesting cost is.

Fig. 2 Location of mature trees with a volume higher than 2.5 m3. The map is supporting harvest screening and localizing possible cutting units Slika 2. Karta položaja prezrelih stabala, obujma > 2,5 m3. Karta pomaže pri planiranju položaja mogućih sječina Croat. j. for. eng. 33(2012)2

175

H. R. Heinimann and J. Breschan

Spatial information (Fig. 2) on the location of mature trees is useful to identify the location of harvest units that yield high timber volume. It has been welldocumented that mountain forests in the Swiss Alps are over aged and should be regenerated to continuously provide protection services (Frehner et al. 2005). »Mature tree maps« are a useful tool to identify the hotspots for a regeneration requirement for regional areas.

4.3 Pre-Harvest Assessment of a Cutting Unit Procjena sječnih jedinica We located and delineated a cutting unit – often called harvest layout, consisting of a cable corridor between two forest roads (Fig. 3), aiming to access two mature timber clusters, one located at the northern, and the other located at the southern road. The length of the cable road is about 260 m between head spar and tail spar. Assuming a maximum lateral yarding

Pre-Harvest Assessment based on LiDAR Data (169–180)

distance of 25 m, the cable corridor has a total area of about 1.3 ha, of which about 1.0 ha consists of stocked forest. The cable corridor is the simplest geometric shape of the cutting unit, created by defining the cable road as a line object in a GIS system and by defining a buffer of 25 m around the line object (Fig. 3). Trees located within the cutting unit are selected with a spatial query, such as »select tree locations that are located within the cutting unit area«, resulting in a cutting unit tree map (Fig. 3, right). Our cable corridor accesses about 350 trees with the total volume of 985 m³, out of which 865 m³ (88%) are from trees with tree volume higher than 2.5 m³. The average standing volume of more than 800 m³ per hectare is extremely high, and even seems to be wrong for people that are not familiar with the local conditions. However, it is well documented that Switzerland has the highest average standing volume per hectare in Europe (Brändli 2010)

Fig. 3 Cutting Unit Layout (cable corridor). A cable corridor is the simplest geometric shape of a cutting unit, consisting of a rectangle with the length of the cable road and the width equaling twice the maximum lateral yarding distance (left). The potential harvesting volume can be estimated with a spatial query such as »select tree locations that are located within the cable corridor«, yielding a map with selected trees (right). Grid spacing: 100 m Slika 3. Odabrana žična linija (sječna linija). Žična je linija predstavljena najjednostavnijim geometrijskim oblikom, s dužinom užeta i dvostrukom udaljenošću postranoga privlačenja (slika lijevo). Moguće užito drvo može se procijeniti vrlo jednostavnom naredbom »označiti stabla koja se nalaze na području žične linije« (slika desno)

176

Croat. j. for. eng. 33(2012)2

Pre-Harvest Assessment based on LiDAR Data (169–180)

H. R. Heinimann and J. Breschan

Table 3 Characteristic Tree Attributes for a cable corridor. The cutting unit, shown in Fig. 3, consists of 1.0 ha of stocked forest area. Results for DBH > 12 cm Tablica 3. Značajke stabala na žičnoj liniji obrasle površine 1 ha (slika 3). Prikazani su rezultati za stabla prsnoga promjera > 12 cm Attribute

Mean

Min

Max

Std. Dev.

Značajke stabla

Arit. sred.

Min.

Maks.

St. dev.

31.5

11.6

46.2

8.6

3.4

0.2

5.5

1.6

Tree height, m Visina stabla, m DBH, cm Prsni promjer stabla, cm Tree volume, m3 Obujam stabla, m3 Voronoi cell area, m2 Površina Voronoieve ćelije, m2

with local maxima up to 1000 m³ per hectare. This is a result of a systematic underuse during the last five decades, resulting in over mature, over aged stands, particularly in mountain areas (Ott 1973). The main purpose of pre-harvest assessment is to locate possible cutting units and to assess volume, quality and value of the standing resource. With the possible layout of Fig. 3, the standard resource is characterized with attributes such as tree height, DBH, and tree volume (Table 3). The average tree volume within the cutting unit is 3.4 m³, covering a range between 0.2 m³ and 5.5 m³. The value of the timber depends on how stems are converted into logs, which is realized by so-called cutting unit or stand level bucking decisions. The supply chain management paradigm aims to further improve the match between supply and demand and to optimize bucking operations before harvesting is carried out in the field (Chauhan et al. 2011). However, related optimization procedures only produce useful results if stand characteristics are captured at the minimum level of accuracy. The available optimization approaches of cutting unit level bucking have usually been using the following parameters: number of stems per diameter class, a set of bucking patterns for each diameter class, and market revenue for each bucking pattern per diameter class (Laroze and Greber 1997). Fig. 4 presents the distribution of the number of stems over DBH for the cable corridor, assuming that there is a deterministic relationship between DBH and tree height. The stem distribution over DBH shows a large variability in tree size. Whereas only about 1% of Croat. j. for. eng. 33(2012)2

Fig. 4 Stem distribution over DBH for a cable corridor. The cutting unit, shown in Fig. 3, consists of 1.0 ha of stocked forest area. Results for DBH > 12 cm. DBH classes < 0.12 m refer to small diameter trees, >0.50 cm to large diameter trees Slika 4. Raspodjela stabala po prsnom promjeru na žičnoj liniji obrasle površine 1 ha (slika 3). Prikazani su rezultati za stabla prsnoga promjera > 12 cm the total volume are classified as small diameter timber (DBH < 20 cm), close to 60% of the total volume belong to large diameter timber (DBH > 50 cm), which is especially challenging for harvesting and processing. Up to now, similar results could only be obtained by doing a full-scale field survey, often in parallel with tree marking, measuring the DBH of each tree to be harvested and calculating tree volume with a tariff function. Fullscale field surveys, also called »100% timber cruising« (Bell and Dilworth 1997), are labor intensive and extremely costly, so that they are rarely applied anymore. Instead, sampling schemes have been a dominant approach, having the limitation that only an estimate of the real stem distribution over DBH is available.

5. Conclusions – Zaključci Our study aimed to explore a LiDAR-data-based approach to improve the sourcing of stands to be harvested as a set of supply points for timber supply networks. We developed a spatially explicit approach, consisting of three steps: 1) harvest screening at the management unit scale or even larger, 2) location and delineation of cutting units, and 3) characterization of tree attributes required for the optimization of stand (cutting unit)-level bucking.

177

H. R. Heinimann and J. Breschan

Our study resulted in the following major findings. First, a tree map represented with Voronoi cells is a useful tool to support the harvest screening process, because it is easily readable and understandable, even by operations personnel. Second, cutting unit location and delineation can easily be done on the tree map, too. This is particularly useful for cable yarding operations, where the layout of a cable road is always determining the delineation of a cutting unit, and where a cutting unit is never equivalent to a single stand. Third, the estimation of stem distribution over DBH of the cutting unit may easily be extracted from the spatial tree map database, assuming that there is a deterministic relationship between tree height and DBH. A tree map with Voronoi delineation improves the spatial information on timber resources tremendously, compared to best practices, which are based on stand maps, for which only aggregated information is available. Additionally, it presents the spatial variability pattern of tree size and volume, which is useful for harvest planning, as well as for silvicultural assessment. Large scale application of our approach would have significant impacts on different supply chain actors. Processing industry would have information where timber resources are located, even on the land of non-industrial forest owners or on public land, where the forest service owns useful information but does not make it accessible to interested supply chain actors. Harvesting managers and harvesting contractors would be able to locate and delineate cutting units so that both operational efficiency and value recovery could be optimized. However, LIDAR data acquisition is still quite expensive, and cost is a major constraint to the introduction and use of the present approach. Whereas the cost of DGPS and of LIDAR sensors have been dropping (Flood 2001), operation of the carrier aircraft (helicopter, fixed-wing airplane) is still expensive. Unmanned aerial vehicles, UAVs, are a next step of development, which will reduce the cost significantly. The carrier platforms – e.g. octocopter (Wallace et al. 2012), helicopter (Yi et al. 2011) – are still in a pre-commercial state of development. Costs are expected to drop from about 45 $.km-2 to about 5 $.km-2 (Johnson 2006). There are still some issues that have to be improved. The smoothing mechanism of the canopy height model CHM is the critical first link for all type of consecutive analysis. Therefore, further research on tree delineation is still required. Methods of mathematical morphology, eventually combined with well-documented approaches (Vauhkonen et al. 2012), such as Gaussian Kernel filtering, provide potential for improvement. Another important issue is to compare LiDAR-based results with ground-truth, which however

178

Pre-Harvest Assessment based on LiDAR Data (169–180)

requires considerable efforts and costs, see e.g. (Solberg et al. 2006). A further line of future research should try to improve the characterization of stem distribution over both DBH and tree height. Whereas tree height may be extracted with considerable accuracy from LiDAR data, DBH has to be estimated indirectly. Our approach used a deterministic relationship between tree height and DBH, whereas DBH varies for equal tree heights due to crown surface variation. And finally, mathematical identification of optimal or at least nearoptimal cutting unit location and delineation could further improve operational efficiency and value recovery.

6. References – Literatura Anderson, H. E., Reutebuch, S. E., Schreuder, G. F., 2001: Use of Automated Individual Tree Crown Recognition and Measurement Algorithms in Forest Inventories. In Precision Forestry, eds. D. Briggs et al., 11–22. Seattle, WA, June 17–20. University of Washington, College of Forest Resources, College of Engineering. USDA Forest Service. Baltsavias, E. P., 1999: Airborne laser scanning: basic relations and formulas. ISPRS Journal of Photogrammetry and Remote Sensing, 54(2): 199–214. Beach, R. H., Pattanayak, S. K., Yang, J. C., Murray, B. C., Abt, R. C., 2005: Econometric studies of non-industrial private forest management: a review and synthesis. Forest Policy and Economics, 7(3): 261–281. Beauchamp, J. J., Olson, J. S., 1973: Corrections for bias in regression estimates after logarithmic transformation. Ecology, 54(6): 1403–1407. Bell, J. F., Dilworth, J. R., 1997: Log scaling and timber cruising. Corvallis OR: Cascade Printing Company. 444 p. Brändli, U. B., 2010: Schweizerisches Landesforstinventar Ergebnisse der dritten Erhebung 2004–2006. Birmensdorf: Eidg. Forschungsanstalt für Wald, Schnee und Landschaft WSL. 312 p. Chauhan, S. S., Frayret, J. M., Lebel, L., 2011: Supply network planning in the forest supply chain with bucking decisions anticipation. Annals of Operations Research, 190(1): 93–115. Courne, X., De, P. H., Guyard, T., Bayol, B., Griffon, S., De Coligny, F., Borianne, P., Jaeger, M., De Reffye, P., 2009: A Forest Growth Simulator Based on Functional-Structural Modelling of Individual Trees. In Plant Growth Modeling, Simulation, Visualization and Applications (PMA), 2009 Third International Symposium on, 34–41. 9–13 Nov. 2009. Diedershagen, O., Koch, B., Weinacker, H., 2004: Automatic segmentation and characterisation of forest stand parameters using airborne LiDAR data, multispectral and fogis data. In Laser-Scanners for Forest and Landscape Assessment. Institute for Forest Growth Institute for Remote Sensing and Landscape Information Systems, 208–212, Freiburg, Albert Ludwigs University. Croat. j. for. eng. 33(2012)2

Pre-Harvest Assessment based on LiDAR Data (169–180) Diedershagen, O., Koch, B., Weinacer, H., Schütt, C., 2003: Combining LiDAR-and GIS data for the extraction of forest inventory parameters. In Proceedings of the ScandLaser Scientific Workshop on Airborne Laser Scanning of Forests, Umeå, Schweden, 157–165. Fiocco, G., Smullin, L., 1963: Detection of scattering layers in the upper atmosphere (60–140 km) by optical radar. Nature, 199: 1275–1276. Flood, M., 2001: Laser altimetry: from science to commercial LiDAR mapping. Photogrammetric engineering and remote sensing, 67(11). Frehner, M., Wasser, B., Schwitter, R., 2005: Nachhaltigkeit und Erfolgskontrolle im Schutzwald. Wegleitung für Pflegemassnahmen in Wäldern mit Schutzfunktionen (Sustainability and effectivity assessment in protection forests. Guidelines for tending measures in forests managed for protection services). Bundesamt für Umwelt, Wald und Landschaft (BUWAL), Bern, 564 p. Hoffmann, C., 1982: Die Berechnung von Tarifen für die Waldinventur (The computation of tarifs for forest inventories). Forstwissenschaftliches Centralblatt, 101(1): 24–36. Hyyppä, J., Hyyppä, H., Litkey, P., Yu, X., Haggren, H., Rönnholm, P., Pyysalo, U., Pitkänen, J., Maltamo, M., 2004: Algorithms and methods of airborne laser-scanning for forest measurements. International Archives of Photogrammetry, Remote Sensing and Spatial Information Sciences, 36(8/ W2): 82–89. Hyyppä, J., Kelle, O., Lehikoinen, M., Inkinen, M., 2001: A segmentation-based method to retrieve stem volume estimates from 3D tree height models produced by laser scanners. Geoscience and Remote Sensing, IEEE Transactions on, 39(5): 969–975. Hyytiäinen, K., Penttinen, M., 2008: Applying portfolio optimisation to the harvesting decisions of non-industrial private forest owners. Forest Policy and Economics, 10(3): 151–160. Johnson, P., 2006: Unmanned aerial vehicle as the platform for lightweight laser sensing to produce su-meter accuracy terrain maps for less than $ 5 per km2. online. Accessed [Sep15-2012]. [http://s122642433.websitehome.co.uk/uavLiDAR/ report.pdf] Kaartinen, H., Hyyppä, J., Liang, X., Litkey, P., Kukko, A., Yo, X., Hyyppä, H., Holopaiinen, M., 2008: Accuracy of automatic tree extraction using airborne laser scanner data. In SilviLaser 2008, Edinburgh, UK. Koch, B., Straub, C., Dees, M., Wang, Y., Weinacker, H., 2009: Airborne laser data for stand delineation and information extraction. International Journal of Remote Sensing, 30(4): 935–963. Krabill, W., Collins, J., Link, L., Swift, R., Butler, M., 1984: Airborne laser topographic mapping results. Photogrammetric engineering and remote sensing, 50: 685–694. Laroze, A. J., Greber, B. J., 1997: Using tabu search to generate stand-level, rule-based bucking patterns. Forest Science, 43(2): 157–169. Croat. j. for. eng. 33(2012)2

H. R. Heinimann and J. Breschan Mallet, C., Bretar, F., 2009: Full-waveform topographic LiDAR: State-of-the-art. ISPRS Journal of Photogrammetry and Remote Sensing, 64(1): 1–16. McCarthy, J., 2001: Gap dynamics of forest trees: a review with particular attention to boreal forests. Environmental Reviews, 9(1): 1–59. Mead, R., 1966: A relationship between individual plantspacing and yield. Annals of Botany, 30(2): 301–309. Morsdorf, F., Meier, E., Kötz, B., Iitten, K. I., Dobbertin, M., Allgöwer, B., 2004: LiDAR-based geometric reconstruction of boreal type forest stands at single tree level for forest and wildland fire management. Remote Sensing of Environment, 92(3): 353–362. Mustonen, J., Packalen, P., Kangas, A., 2008: Automatic segmentation of forest stands using a canopy height model and aerial photography. Scandinavian Journal of Forest Research, 23(6): 534–545. Ott, E., 1973: Erhebungen über den gegenwärtigen Zustand des Schweizer Waldes als Grundlage waldbaulicher Zielsetzungen, Schweizerische Anstalt für das Forstliche Versuchswesen. Pereira, P., 2005: Do lower search costs reduce prices and price dispersion? Information Economics and Policy, 17(1): 61–72. SCOR, 2010: Supply Chain Operations Reference (SCOR) model. Overview – Version 10.0. Accessed [Aug-14-2012]. [http://supply-chain.org/f/SCOR-Overview-Web.pdf] Solberg, S., Naesset, E., Bollandsas, O. M., 2006: Single tree segmentation using airborne laser scanner data in a structurally heterogeneous spruce forest. Photogrammetric engineering and remote sensing, 72(12): 1369. Sullivan, A. A., 2008: LiDAR Based Delineation in Forest Stands. College of Forest Resources, University of Washington, Seattle. 83 p TRIMBLE, 2012: Trimble Harrier 68i Corridor Mapping System – Datasheet. [Accessed Aug-14-2012 2012]. Available: [http://trl.trimble.com/docushare/dsweb/Get/Document-560731/] Vauhkonen, J., Ene, L., Gupta, S., Heinzel, J., Holmgren, J., Pitkänen, J., Solberg, S., Wang, Y., Weinacker, H., Hauglin, K. M., 2012: Comparative testing of single-tree detection algorithms under different types of forest. Forestry, 85(1): 27–40. Vepakomma, U., St-Onge, B., Kneeshaw, D., 2008: Spatially explicit characterization of boreal forest gap dynamics using multi-temporal LiDAR data. Remote Sensing of Environment, 112(5): 2326–2340. Wallace, L., Lucieer, A., Watson C., Turner, D., 2012: Development of a UAV-LiDAR System with Application to Forest Inventory. Remote Sensing, 4(6): 1519–1543. Wehr, A., Lohr, U., 1999: Airborne laser scanning – an introduction and overview. ISPRS Journal of Photogrammetry and Remote Sensing, 54(2): 68–82. Weinacker, H., Koch, B., Heyder, U., Weinacker, R., 2004: Development of filtering, segmentation and modelling mod-

179

H. R. Heinimann and J. Breschan

Pre-Harvest Assessment based on LiDAR Data (169–180)

ules for LiDAR and multispectral data as a fundament of an automatic forest inventory system. International Archives of Photogrammetry, Remote Sensing and Spatial Information Sciences, 36(8/W2): 50–55.

Forstliche Biometrie, Albert-Ludwigs-Universitaet, Freiburg i.Br., Germany Yi, L., Hyyppä, X. J., Jaakkola, A., 2011: MiniUAV-Borne LiDAR for Fine-Scale Mapping. Geoscience and Remote Sensing Letters, IEEE, 8(3): 426–430.

Weisstein, E. W., 2012: Voronoi Diagram. MathWorld - A Wolfram Web Resource. Accessed [Aug-14-2012]. [http:// mathworld.wolfram.com/VoronoiDiagram.html]

Zhang, K., 2008: Identification of gaps in mangrove forests with airborne LiDAR. Remote Sensing of Environment, 112(5): 2309–2325.

Ye, R., 1995: Waldsimulation auf der Basis automatischer Luftbildmessung und unter Kontrolle von GIS. Abteilung

Sažetak

Planiranje pridobivanja drva na osnovi LiDAR-ovih snimaka Prikupljanje je podataka prvi korak razvoja konkurentnosti u procesu pridobivanja drva te određivanje sastojina za sječu koje najbolje odgovaraju potrebama tržišta. Ono je težak postupak jer su informacije ili vrlo općenite ili čak nedostupne, kao što je to u slučaju šuma privatnih šumovlasnika. Cilj je ovoga rada bio istražiti mogućnost primjene LiDAR-ovih snimaka za unapređenje prikupljanja podataka o sastojinama koje će se sjeći. Razvijen je izrazito pros torni pristup, koji se sastoji od triju koraka: Þ planiranje radova pridobivanja drva na razini odjela/odsjeka ili čak na većim površinama (revir ili gospodarska jedinica), Þ određivanje položaja i prepoznavanje sječnih jedinica (odjel/odsjek), Þ opis značajki stabala koje su potrebne za optimiziranje sortimentne strukture sastojine. Na temelju rezultata istraživanja doneseni su sljedeći zaključci. Karta stabala, predstavljena Voronoievim dija gramom, koristan je alat za planiranje pridobivanja drva jer je lako čitljiva i razumljiva u operativnoj primjeni. Nadalje, na istoj je karti lako odrediti i označiti sječnu liniju. Procjenu raspodjele prsnih promjera stabala u sječini moguće je lako izlučiti iz baze podataka prostornoga položaja stabala, uz pretpostavku postojanja povezanosti visine i prsnoga promjera stabala. Međutim, još uvijek ima prostora za poboljšanja, kao što je usporedba rezultata dobivenih s LiDAR-ovih snima ka i terenskih izmjera podataka, unapređenje metoda prepoznavanja stabala temeljenih na LiDAR-ovim snimcima, unapređenje procjene sortimentne strukture na temelju prsnoga promjera i visine stabala ili matematičko formulira nje i definiranje rasporeda sječnih jedinica. Ključne riječi: LiDAR, prepoznavanje parametara stabla, Voronoiev mozaik, planiranje sjekoreda, planiranje pridobivanja drva

Authors’ address – Adresa autorâ:

Received (Primljeno): July 20, 2012 Accepted (Prihvaćeno): August 31, 2012

180

Prof. Hans Rudolf Heinimann, PhD. e-mail: hans.heinimann@env.ethz.ch Jochen Breschan, PhD. e-mail: jochen.breschan@env.ethz.ch Institute of Terrestrial Ecosystems Department of Environmental Systems ETH Zürich Universitätsstrasse 22 8092 Zürich SWITZERLAND Croat. j. for. eng. 33(2012)2

Original scietific paper – Izvorni znanstveni rad

Development of a Multi-Criteria Decision Support Tool for Energy Wood Supply Management Martin Kühmaier, Karl Stampfer Abstract – Nacrtak The use of wood for heat and energy production has increased over the last two decades and there is still increasing demand. Harvesting energy wood can provide an extra source of income for the forest owners and generally affects the economics of forest operation positively, but impacts on environmental and social aspects, like the ecological balance of the forest, working conditions and rate of employment, need to be accounted as well. Individual stakeholders are faced with a number of options regarding energy wood harvest and supply systems though they often only have limited knowledge about their impacts. It is also important to understand that individual stakeholders often have different preferences regarding the balance between economic, ecological and social impacts. A computer-based decision support tool was developed in Microsoft Excel® using Visual Basic for Applications® to assist stakeholders in identifying the most suitable energy wood supply chain to meet their needs. The tool considers a number of criteria such as energy efficiency, nutrient balance, stability and vitality of the remaining stand and soil, contribution margin, supply guarantee, employment rate and working safety. Users can specify site, stand and environment data and technology parameters, and set their individual preferences for balance between the criteria. The overall utility of various treatment alternatives are calculated by an additive utility model and presented via on-screen graphs and tables. The tool is designed to convey knowledge gained in research through practical and understand able tool. In addition to presenting the decision support tool, this paper explores the influence of criteria weighting, terrain conditions, transport distance and moisture content on selecting the most suitable supply chain. Keywords: Decision support, utility model, supply management, energy wood, bioenergy harvesting, forest management planning, multiple-purpose forestry

1. Introduction – Uvod The current contribution of wood fuels to the total energy production is already relatively high in some European countries. In the last two decades the demand for energy wood has been increasing and there is still potential to intensify utilization across Europe (Hall 1997; Moiseyev et al. 2011). The new directive (EU 2009) on renewable energy sources sets ambitious targets for all Member States. The EU should reach a 20% share of energy from renewable sources by 2020. Croat. j. for. eng. 33(2012)2

The development of renewable energy systems does not have a long history, which is the reason for a general lack of experience and understanding of renewable energy sources and technology – within the public sector as well as industry and among private users. The potential of energy wood in European Union countries is significant (Hall 1997; Berndes et al. 2003), but more has to be done to improve harvesting, transport and combustion technology. In addition to technology development, further improvements are needed regarding knowledge about

181

M. Kühmaier and K. Stampfer

Development of a Multi-Criteria Decision Support Tool for Energy Wood .....(181–198)

bioenergy in general and forest fuels in particular (Kärhä 2011). The conversion of forestry woody biomass into fuel and moving this resource from the forests to the plant is the challenge of any energy wood supply chain (EWSC), which consists of various processes, e.g. felling, extracting, chipping and transporting (Stam pfer et al. 2011). There is also an ongoing dis cu ssion about ecological and social consequences of forest biomass harvesting, such as nutrient depletion of the soils, loss of biodiversity or working safety (Makeschin 1994; Bohlin and Roos 2002; Röser et al. 2006; Bright et al. 2010). The different players in the energy wood supply management (EWSM), e.g. forest managers have to take environmental, economic and social factors into account and aggregate them to insure a sustainable utilization. The directive on renewable energy (EU 2009) also requires national action plans for the development of renewable energy sources, and it establishes sustainability criteria for biofuels. Due to the multitude of alternatives and objectives, choosing a supply chain approach is a complex task in EWSM. Given the challenge to choose the most suitable supply chain considering economic, environmental and social aspects, a decision support tool could be helpful. Multi Criteria Analysis (MCA) is one of many tools available to assist decision makers in the decision process. In recent years MCA has been extensively used for problem solving, primarily within the area of natural resource management. Most of these applications have been successful in developing applied management options leading to improved environmental and social management. A MCA approach enables investigators and decision makers to integrate the components of sustainable (social, economic and environmental) development (Hwang and Yoon 1981; Fath et al. 1999; Solomon and Hughey 2007). Howard (1991) generally questioned the applicability of mathematical programming techniques for ill-structured problems, which are quite common in multiple-purpose forestry. Though well accepted in industry and business applications, MCA techniques as decision support tools have rarely been applied for the evaluation in forestry (e.g. Mendoza 1989; Canham 1990; Næsset 1997; Sheppard and Meitner 2005; Wolfslehner et al. 2005; Wolfslehner and Vacik 2008; Kangas et al. 2008). For European conditions, Kangas (1993); Kangas et al. (2001); Kangas and Kuusipalo (1993); Pukkala and Kangas (1993); Vacik and Lexer (2001); Lexer et al. (2005); Kühmaier and Stampfer (2010) provide examples of MC-solutions for multiobjective and multi-criteria decision problems including biodiversity and amenity values.

182

MCA tools for supporting decision making in EWSM are rare but, due to the increasing attention to sustainability in this area, they are becoming more important in recent years (Windisch et al. 2010). There have been several studies about estimating local energy wood fuel resources (e.g. Vainio et al. 2009; López-Rodríguez et al. 2009; Padari et al. 2009; Fernandes and Costa 2010; Gómez et al. 2010; Emer et al. 2011; Ranta and Korpinen 2011), cost analyses (e.g. Laitila 2006; Laitila et al. 2010; Tahvanainen and Anttila 2011) or optimizing supply networks (e.g. Ranta 2005; Kanzian et al. 2009; Kim et al. 2011). Nevertheless, there is still a lack of tools for comparing and evaluating EWSCs on a multi-criteria level and giving recommendations for the selection of the most suitable systems for a given condition and set of objectives. The aim of this study is to develop a decision support tool for EWSM based on MCA that assesses sustainability including economic, environmental, and social criteria. A model was designed to evaluate alternatives and scenarios based on user defined site, stand, environment and machine data. The tool covers the current state-of-the-art energy wood supply chains and is implemented in a common spreadsheet platform.

2. Material and Methods – Materijal i metode An essential step in developing a decision support system is the design of the framework, which includes defining the system to be modeled. The framework combines different aspects of DSS-development. The process model represents the flow of data and information throughout the decision-making process, and describes the exchange of information among various DSS components. The formal model includes the algorithms, rules, and mathematical equations needed to formally describe the modeled system. Finally, the implementation model comprises software architecture and technical solutions to implement the master model (Lexer et al. 2005).

2.1 Framework – Okosnica To ensure the quality of the consultation process, the user is guided through a standardized decisionmaking process (Fig. 1). To begin, the user is provided with a default scenario (based on average values from previous studies) or with an existing scenario from a previously saved session. The user is advised to follow the recommended menu starting with worksheet 1 (S1) and finishing with sheet 5 (S5). After identifying his/her individual preferences, the user initializes the Croat. j. for. eng. 33(2012)2

Fig. 1 Framework of the decision support tool for energy wood supply management Slika 1. Okvir alata za pomoć pri odlučivanju za upravljanje dobavom energijskoga drva

Development of a Multi-Criteria Decision Support Tool for Energy Wood .....(181–198)

Croat. j. for. eng. 33(2012)2 M. Kühmaier and K. Stampfer

183

Development of a Multi-Criteria Decision Support Tool for Energy Wood .....(181â&#x20AC;&#x201C;198)

Fig. 2 State-of-the-art energy wood supply chains Slika 2. Najsuvremeniji lanci dobave energijskoga drva

M. KĂźhmaier and K. Stampfer

184

Croat. j. for. eng. 33(2012)2

Development of a Multi-Criteria Decision Support Tool for Energy Wood .....(181–198)

M. Kühmaier and K. Stampfer

Fig. 3 Objectives, criteria and indicators in energy wood supply management Slika 3. Ciljevi, kriteriji i pokazatelji pri upravljanju lancem dobave energijskoga drva

planning process for a particular forest by classifying the current stand, site and environmental conditions. Harvesting and transportation data is the last input made by the user. All the input data (grey colored in Fig. 1) will be transferred into utility values after executing several intermediate calculations (utility analysis). By aggregating these values and ranking the results the most suitable EWSC will be determined. The model provides the decision maker with tables and figures that summarize the suggestions of the model.

2.2 Process and formal model – Postupak i zadani model Energy wood supply chains (alternatives) In estimating the most suitable EWSC, the DS tool consists of 48 alternatives within 6 supply groups (Fig. 2). The alternatives vary in terms of harvesting and transporting machines used, harvesting methods (whole tree, tree length, cut-to-length), chipping location (stand, forest road, terminal, plant), and the type of woody biomass harvested (whole trees, harvesting residues, or both). Bundling is an optional process for Croat. j. for. eng. 33(2012)2

compressing harvesting residues. Biomass can be transported as forest residues, round-wood, pressed bundles or chips (Hakkila 2004; Stampfer and Kanzian 2006; Kühmaier et al. 2007; Kärhä 2011; Stampfer et al. 2011). Selection and weighting of criteria and indicators For the evaluation process, independent criteria and indicators (C&I) have been chosen. C&I-approaches appear to be effective in measuring aspects of sustainable forest management (Prabhu et al. 1999; Wolfslehner et al. 2005). For this tool, 6 indicators were defined with quantitative values – nutrient loss, bearing pressure, energy-efficiency, damage on remaining stand, contribution margin and operating time – and 2 index values – supply guarantee and working safety (Fig. 3). In a database (target system matrix), indicator values were assigned to each alternative. The data matrix describes how good the particular criterion fulfills the respective alternative. The user indicates his/her preferences by weighting (S1 in Fig. 1) pre-defined criteria. If one of these criteria does not fit within the scope of

185

M. Kühmaier and K. Stampfer

Development of a Multi-Criteria Decision Support Tool for Energy Wood .....(181–198)

Table 1 Limits for the technological evaluation of energy wood supply chains Tablica 1. Ograničenja pri tehnološkom vrednovanju lanaca dobave energijskoga drva Criteria – Kriteriji

Classes – Razredi

Morphology – Reljef

uniform – ujednačen, non-uniform – neujednačen

Soil strength – Nosivost podloge

high – zadovoljavajuća, low – nezadovoljavajuća

Slope – Nagib

<35%, 35–60%, >60%

DBH – Prsni promjer stabala

<25 cm, <50 cm, >50 cm

Extraction distance – Udaljenost privlačenja

<80 m, 80–800 m, >800 m

Soil type – Vrsta tla Chipping place – Mjesto iveranja

non-sensitive, sensitive (soil depth <30 cm, virgin soil, podzol, rendzina, terra fusca, moor) neosjetljiva, osjetljiva (dubina tla <30 cm, podzol, renzina, smeđa tla, treset) stand – sastojina, forest road – šumska cesta, terminal – stovarište, plant – energana

objectives, the user may give a weight of zero. The proportions of the weights are calculated in percentage values to generate a total of 1. Specification For the technological evaluation, an algorithm, similar to a decision tree, filters feasible alternatives. A decision tree (Magee 1964) is a decision support tool that uses a tree-like graph or model of decisions and their possible consequences. 7 criteria with predefined limits have been used to exclude not applicable alternatives within the evaluation process (Table 1). Site and stand specification (S2 in Fig. 1) acts as input data for the decision tree. The slope is a limiting factor for wheeled (30%) and tracked (60%) machines. The given limits are average values; they can vary depending on relief and soil bearing capacity. The extraction distance is a limiting factor for cable-operated machines, e. g. tower yarders (800 m) and skidders (80 m). The limiting diameter at breast height (DBH) for fellerbuncher, harvester and processor depends on the type of harvesting head. A strongly varying morphology and low soil strength are restricting factors for groundbased systems as a result of reduced trafficability (Kühmaier and Stampfer 2010).The utilization of forest residues can result in ecological risks (Krapfenbauer 1983) as well as in growth reduction (Sterba 2003), as valuable nutrients are removed from the forest. Therefore, the utilization of residues is only recommended for nutrient-rich soil conditions. Hence the whole-tree (WT) method should be excluded on sensitive soils. The performances of indicator values vary not only according to site and stand data, but also to according to environment and machine specifications (S3 and S4 in Fig. 1). Chipping place, transport distance and moisture content, for example, play a decisive role for the

186

efficiency of the whole EWSC. The main characteristics of the harvesting processes in the supply chain are specified with productivity, system costs and fuel consumption. Default Input values were taken from energy wood harvesting/supply studies (e.g. Kühmaier et al. 2007; Stampfer et al. 2011). Furthermore, productivity models are implemented in the DSS to help the user estimate the productivity of the various processes. The specification data is also used to calculate indicator values for energy efficiency, contribution to margin and employment rate. The indicator values for all other criteria are fixed for each alternative and therefore unchangeable. All the indicator values (target system matrix) act as input for the utility analysis. Transformation of target values into utility values To evaluate the overall utility of alternatives, for cases where more than one solution is possible, an approach borrowed from multiple-attribute utility theory (MAUT) was adopted (Goicoechea et al. 1982). It is assumed that there are a certain number of criteria and a one-dimensional utility function for each of these criteria (Kühmaier and Stampfer 2010). In case of quantitative indicators, such as with the amount of € per MWh as indicator for contribution to margin, quantitative utility functions were estimated and used to transfer an indicator from its measurement scale to the dimensionless »utility« scale on the interval [0–1]. In this study, for all quantitative data, the utility functions have been scaled with score range procedure on an interval scale (Kangas et al. 2008). Interval scale can be interpreted as local scale; the length of the interval depends on specific planning situation (Kainulainen et al. 2007). In cases where no quantitative values for the indicators can be provided to measure the progress toCroat. j. for. eng. 33(2012)2

Development of a Multi-Criteria Decision Support Tool for Energy Wood .....(181–198)

wards objectives, alternatives were directly compared in pairwise comparisons employing Saaty’s ratio-scale approach and expert judgment (Saaty 1977). All pairwise comparisons were performed by experts. Similar approaches to synthesize preference information across multiple attributes were described in Vacik and Lexer (2001) and Lexer et al. (2005). The resulting utility values were normalized and stored in a database (utility matrix). Aggregation Aggregation of utility values is necessary to describe the overall utility of alternatives. This aggregation is done by criteria weighting within the utility function with respect to their importance. The relations between the weights of different criteria describe the tradeoffs between the criteria (Kangas et al. 2008). The most suitable alternative is the one with the highest overall utility (Kühmaier and Stampfer 2010).The most applied multi-attribute utility function is the linear additive utility function written as U = a × (a × UEE + b × UNB + g × USS + d × USV) + + b × (e × UCM + z × USG) + c × (h × UFM + q × UWS) (1) under the constraints that weighting coefficients a + b + c = 1, α + β + γ + δ = 1, ε + ζ = 1 and η + θ = 1. The objectives and criteria corresponding to the weighting coefficients a, b, c and α, β, γ, δ, ε, ζ, η, θ are presented in Fig. 3. UEE, UNB, USS, USV, UCM, USG, UEM and UWS are the utility values for the criteria energy efficiency, nutrient balance, soil stability, stand vitality, contribution margin, supply guarantee, employment and working safety. Report The results of the calculation are available in anon screen report. The report displays the most suitable EWSC within each EWSC group and a ranking of these groups. The overall utility is visible both as a value and as a graph. In this way it is possible to see differences at a glance. As additional information, the absolute values of the indicators are displayed. This helps the user to get a better understanding of the impacts when using certain EWSCs. Another sheet shows the supply costs subdivided into several processes available as €/loose m³, €/MWh, €/solid m³, and €/dry tons. This information is also available as a chart. Finally the supply costs of energy wood is compared with other fuels, like natural gas, heating oil, black coal, wood briquettes, pellets, and firewood.

2.3 Implementation – Primjena The program was constructed in Microsoft Excel® using Visual Basic for Applications® technology Croat. j. for. eng. 33(2012)2

M. Kühmaier and K. Stampfer

(VBA). VBA offers a sophisticated programming tool, which facilitates the creation of applications with user interfaces and custom dialog boxes. By generating buttons, the user is guided more easily through the system. This user guidance is organized in a starting page with general information about the tool and links to several worksheets and sub-sheets. To execute the tool, a first time user is instructed to follow the intended sequence (S1–S5 in Fig. 1). There is always one active sheet visible. If the user clicks on a button. the inactive sheet will be hidden and the new active sheet will be visible. To assist the user, some simple calculating sheets can be opened and closed for specific processes, e.g. for calculating productivity or showing details within the EWSC groups. Additional directions on entering specification data are available through comments if the user clicks on particular terms. The weights of each criterion are defined by the user. With the interactive interface, the user sees immediately how the current set of weights affects his preferences. The user can enter or change specification data (grey-covered boxes in Fig. 1) and save input data and results. However, in the input fields, default values are given, compiled from previous studies, so that entering data is not mandatory to make the tool work. The calculation of intermediate values, the target system and the utility matrix as well as the ranking of the alternatives are in the background and are not visible to the user. The applicant may store the scenario (including all parameter settings) at any time, print the report, exit the scenario session or return to the beginning.

3. System demonstration – Prikaz sustava Since the model has a wide range of scenarios with 115 potential input fields, for the purpose of this demonstration, variations were limited to changes in weighting as well as in slope, transport distance and moisture content. The remaining specification data (Annex: Table A1 and A2) is constant for all scenarios. For all other input fields no technological restrictions were used.

3.1 Weighting scenarios – Scenariji težina In the first example the weighting of the economic criteria was changed and the performance of supply groups 2, 3, 5 and 6 was analyzed (Fig. 4). Group 2 (harvester-forwarder) was always the most suitable EWSC. Group 3 (chain saw-tower yarder) is most similar to group 6 (feller buncher-forwarder), but if the importance of economics is increasing the group 3 has an advantage because the supply costs are lower, especially for felling and extracting (by-production).

187

M. Kühmaier and K. Stampfer

Development of a Multi-Criteria Decision Support Tool for Energy Wood .....(181–198)

Fig. 4 Most suitable EWSCs according to the importance of economic criteria Slika 4. Najpogodniji lanci dobave energijskoga drva s obzirom na privrednu važnost Furthermore chipping and storage of logs and residues (group 6) is more expensive than processing logs (group 3). Group 5 (feller buncher-chipper-shuttle) is not recommended for energy wood supply, as the supply costs, especially for chipping in the stand, are too high. Group 5 and 6 have a negative contribution margin (–7.89 € and –7.03 €, respectively). If no economic criteria are considered, there is little difference (spread of

0.12) between the supply chains, but this will change dramatically as the economic criteria weighting increases to 50% (0.50) or 100% (0.88). This clearly shows that economic considerations most highly affect the difference between supply chains and have a greater influence on EWSC choice than ecological and social criteria that are very similar between the system options. From an ecological point of view all EWSCs are similar with the exception of group 5, which is the only poor one. The harvester and forwarder system is evaluated as one of the best alternatives from an economic, ecological and social point of view. Example 2 shows an analysis of EWSCs in flat (<30%) versus steep terrain (>60%) with the weightings described in Table 2. Eight scenarios have been compared in this example. In flat terrain, all EWSCs are technically applicable so 48 alternatives are available to the decision support tool. Under these circumstances, where the decision process is complex, the tool provides the greatest value. In the balanced scenario, group 2 has been selected as the most suitable supply chain with a 3.94 €/m³ loose contribution margin and a primary energy consumption of 1.63%. Group 3 is recommended in the ecological weighted scenario mainly due to the absence of bearing pressure with tower yarder extraction. The economic weighted scenario delivers the same result as the balanced scenario. From a social aspect, group 2 is also the most suitable but only where forest residues, extracted separately by a forwarder, are used as energy wood. Therefore, the productivity of the EWSC is very low but offers a high rate of employment (0.24 hours/m³ loose). Emissions and energy consumption are also very high for this scenario and there is a negative contribution margin. In this case it

Table 2 Weighting scenarios Tablica 2. Scenariji (težinskoga) vrednovanja Criteria

Balanced

Economic

Ecological

Social

Kriterij

Uravnotežen

Privredni

Okolišni

Socijalni

Energy efficiency – Energetska djelotvornost

15%

Nutrient balance – Ravnoteža hraniva

15%

Soil stability – Stabilnost tla

15%

Stand vitality – Vitalnost sastojina

15%

Contribution margin – Kontribucijska marža

17%

30%

10%

Supply guarantee – Jamstvo dobave

17%

30%

10%

Employment – Zaposlenost

17%

10%

30%

Working safety – Sigurnost pri radu

17%

10%

30%

188

Croat. j. for. eng. 33(2012)2

Development of a Multi-Criteria Decision Support Tool for Energy Wood .....(181–198)

M. Kühmaier and K. Stampfer

Table 3 Most suitable energy wood supply chains on flat terrain Tablica 3. Najpogodniji lanci dobave drva na ravnome terenu Privredni

Socijalni

Most suitable supply group

Group 2

Group 3

Group 2

Najpogodnije grupe dobave

Grupa 2

Grupa 3

Grupa 2

Most suitable supply chain Najpogodniji lanac dobave

Sječa i izradba drva harvesterom, izvoženje forvarderom, prijevoz šumskoga ostatka, iveranje kod energane

Okolišni

Felling and processing with harvester, extracting with forwarder, transportation of residues, chipping at plant

Uravnotežen

Sječa i izradba drva harvesterom, izvoženje forvarderom (trupci), prijevoz trupaca, iveranje kod energane

Ravan teren

Felling and processing with harvester, extracting with forwarder (logs), transportation of logs, chipping at plant

Social

Sječa motornom pilom, iznošenje stabala stupnom žičarom, izradba drva procesorom, prijevoz šumskoga ostatka, iveranje kod energane

Economic

Felling with chain saw, extracting with tower yarder (tree), processing with processor, transportation of residues, chipping at plant

Ecological

Sječa i izradba drva harvesterom, izvoženje forvarderom (trupci), prijevoz trupaca, iveranje kod energane

Balanced

Felling and processing with harvester, extracting with forwarder (logs), transportation of logs, chipping at plant

Flat terrain

Nutrient loss – Gubitak hraniva, kg/ha

640

960

640

960

Bearing pressure – Dodirni tlak, kPa

300

3.15

3.51

3.15

7.31

1.63%

1.82%

1.63%

3.78%

11%

3.94

2.67

3.94

–11.42

0.10

0.12

0.10

0.24

Emissions – Polucije, kg CO2/m loose 3

Energy consumption – Potrošnja energije, kWh/m loose Energy efficiency – Energetska učinkovitost, % Damage on stand – Oštećenost sastojine, % Contribution margin – Kontribucijska marža, €/m loose 3

Working hours – Norma vremena, h/m loose

is readily identifiable that too strong a focus on social criteria can have negative impact on ecology and economics. The trade-offs of a multi-criteria evaluation are clearly distinguishable (Table 3). Harvesting will be carried out as a by-production in the ecological scenario. In this case only residues will be used as fuel wood. The logs are delivered to saw and paper mills. The impacts (indicators) for all by-production processes are valued at 10%. This quota correlates approximately to the ratio of revenues for energy wood versus logs. Based on this ratio, the impacts of some indicators are reduced, e.g. damage to remaining stand is denoted with only 3% in contrast to real damage of 29%. In steep terrain, the option of wheeled or tracked vehicles is eliminated, so only harvesting operations with a tower yarder (group 3) are possible. In the balanced scenario, the processing is done by a processor Croat. j. for. eng. 33(2012)2

at the roadside and only the logs are transported to the plant, where they are stored and chipped. This scenario has an energy efficiency of 1.97% and a contribution margin of –4.57 €/m³ loose. From an ecological point of view, the same EWSC as in flat terrain should be used. Harvesting will be carried out as by-production for the ecological and economic scenario. The social scenario suggests using logs and residues for energy production. This result has once again a negative contribution margin but the rate of employment is boosted to 0.37 hours/m³ loose (Table 4).

3.2 Sensitivity analysis – Analiza osjetljivosti The following sensitivity analysis shows the impacts on energy demand and the contribution margin as it relates to transport distance and moisture content. A longer transport distance increases transport time and therefore costs, which results in a lower con-

189

M. Kühmaier and K. Stampfer

Development of a Multi-Criteria Decision Support Tool for Energy Wood .....(181–198)

Table 4 Most suitable energy wood supply chains on steep terrain Tablica 4. Najpogodniji lanci dobave drva na strmom terenu Privredni

Socijalni

Most suitable supply group

Group 3

Najpogodnije grupe dobave

Grupa 3

Most suitable supply chain Najpogodniji lanac dobave

Sječa motornom pilom, iznošenje stabala stupnom žičarom, izradba drva procesorom, prijevoz trupaca i šumskog ostatka, iveranje kod energane

Okolišni

Felling with chain saw, extracting with tower yarder (tree), processing with processor, transportation of logs & residues, chipping at plant

Uravnotežen

Sječa motornom pilom, iznošenje stabala stupnom žičarom, izradba drva procesorom, prijevoz šumskoga ostatka, iveranje kod energane

Strmi teren

Felling with chain saw, extracting with tower yarder (tree), processing with processor, transportation of residues, chipping at plant

Social

Sječa motornom pilom, iznošenje stabala stupnom žičarom, izradba drva procesorom, prijevoz šumskoga ostatka, iveranje kod energane

Economic

Felling with chain saw, extracting with tower yarder (tree), processing with processor, transportation of residues, chipping at plant

Ecological

Sječa motornom pilom, iznošenje stabala stupnom žičarom, izradba drva procesorom, prijevoz trupaca, iveranje kod energane

Balanced

Felling with chain saw, extracting with tower yarder (tree), processing with processor, transportation of logs, chipping at plant

Steep terrain

Nutrient loss – Gubitak hraniva, kg/ha

640

960

Bearing pressure – Dodirni tlak, kPa

3.72

3.51

5.10

1.97%

1.82%

2.67%

Damage on stand – Oštećenost sastojine, %

29%

Contribution margin – Kontribucijska marža, €/m3 loose

–4.57

2.67

–7.98

0.34

0.12

0.37

Emissions – Polucije, kg CO2/m3 loose 3

Energy consumption – Potrošnja energije, kWh/m loose Energy efficiency – Energetska učinkovitost, %

Working hours – Norma vremena, h/m loose

tribution margin (Fig. 5). Group 3, 5 and 6 have a negative contribution margin even at short distances of approximately –5.8 €/m³ loose due to high harvesting costs. The contribution margin will further decrease to –6.8 €/m³ loose and –8.1 €/m³ loose for distances of 50 and 100 km, respectively. Only group 2 has positive results for short, medium and long distances of 4.8, 3.6 and 2.1 €/m³ loose because the use of logs as energy wood results in lower felling, extracting and chipping costs. Only distances of more than 170 km will generate negative results.

and 3.3% in relation to the energy content of the wood. Increasing the transport distance to 100 km causes an increase in energy demand of about 0.8 percentage points (Fig. 5). For group 2 and 5, the energy demand is almost uniformly distributed among individual processes. The extraction with tower yarder in group 3 requires 50% of the entire energy demand due to low productivity (8.5 m³/h) and high fuel consumption (16 l/h). The low productivity (17 m³ loose/h) of the mobile chipper explains the huge energy demand for chipping in group 5 (Table 5).

Longer transport distances require more fuel and therefore a higher energy demand. For a distance of 10 km, EWSCs have an energy demand between 1.4

For a transport distance of 38 km, the reduction of the moisture content from 50 to 30% results in an increase in contribution margin ranging between 0.8 and

190

Croat. j. for. eng. 33(2012)2

Development of a Multi-Criteria Decision Support Tool for Energy Wood .....(181–198)

M. Kühmaier and K. Stampfer

Fig. 5 Performance of EWSC as a function of transport distance Slika 5. Značajke lanaca dobave energijskoga drva u ovisnosti o udaljenosti prijevoza Table 5 Energy demand for energy wood supply processes Tablica 5. Energetske potrebe postupaka dobave energijskoga drva Process – Postupak

Group 2 – Grupa 2

Group 3 – Grupa 3

Group 5 – Grupa 5

Group 6 – Grupa 6

Felling, delimbing – Sječa, kresanje grana

27%

17%

23%

Extracting – Privlačenje drva

22%

50%

30%

Chipping – Iveranje

24%

20%

63%

25%

Transportation – Prijevoz

27%

22%

11%

22%

kWh/m loose

1.6 €/m³ loose and a reduction in energy demand ranging between 0.13 and 0.19 percent points (Fig. 6). Lower moisture content reduces the unit weight, which decreases fuel consumption and allows the transport of more wood (energy) per load. The impact is greater with logs than residues since residues have a much lower density.

4. Discussion – Rasprava This paper presents the development of a Multiple Criteria Decision Support Tool for EWSM, its implementation as an easy-to-use software and the demonstration of its use through scenario analysis. The tool focuses on state-of-the-art energy wood supply chains covering felling through transportation to the plant and evaluates alternative systems based on ecological, economic and social criteria. Croat. j. for. eng. 33(2012)2

Several papers (e.g. Spinelli et al. 2007; Cremer and Velazquez-Marti 2007; Jylhä and Laitila 2008; Ovaskaenen et al. 2008; Rottensteiner et al. 2008; Lehtimäki and Nurmi 2011) explore case studies of energy wood harvesting systems with focus on a specific machine mix, terrain condition, stand type and environmental requirements. These single studies allow for an effective efficiency assessment of that particular system but, do not allow for easy comparisons with other supply chains. Other studies explore more than one EWSC but tend to focus only on certain processes within the supply chain, e.g. harvesting or transportation (Johansson et al. 2006; Kanzian et al. 2009; Tahvanainen and Anttila 2011). Further projects consider the type of woody biomass (Röser et al. 2006), or aim to optimize the supply network and allocation of material flows (Kanzian et al. 2009; Vainio et al. 2009). Criteria used in the evaluation of supply systems are gener-

191

M. Kühmaier and K. Stampfer

Development of a Multi-Criteria Decision Support Tool for Energy Wood .....(181–198)

Fig. 6 Performance of EWSC as a function of moisture content Slika 6. Značajke lanaca dobave energijskoga drva u ovisnosti o sadržaju vlage ally limited to costs (Laitila 2006; Heikkilä et al. 2006; Laitila 2008; Röser et al. 2011) and in some cases CO2 emissions (Alam et al. 2011). Even with the wealth of knowledge generated in recent years on EWSCs, there still has not been an application developed to meet the need for an approach to easily and effectively evaluate EWSCs based on all the important sustainability criteria. By using a multi-criteria approach, this tool represents an innovative approach to evaluate state-of-theart EWSCs based on economic, environmental and social criteria. A multi criteria decision support approach creates a good starting point for the practical application of sustainable EWSCs, where various dimensions of sustainability are integrated (Kangas et al. 2008). Economic targets are important for forest owners, but with increased emphasis on environmental, ecological and social outcomes in forestry, it has become increasingly important to optimize these outcomes in balance with economics. The proposed method attempts to include all the priorities of various parties (i.e. forest managers, harvesting companies, funding agencies) to ensure an integrated and holistic evaluation. The tool is flexible in allowing users to define their priorities through individual criteria weighting. Where stakeholders do not wish to, or cannot, assign weights, the system uses a default assumption of equal weighting for ecological, economic and social criteria. The default allocation of weights prevents decision makers from exhibiting a preference for a policy op-

192

tion, as the allocated weights are common to all the policy options that are to be considered in later process steps. Despite this implication it is clear that the weighting exercise is sensitive to the expressed differences of those involved in the policy making process. Nevertheless, the explicit process through which this weighting occurs is expected to limit the strategic and politically motivated behavior that might occur at this and later stages, where similar issues might arise (Solomon and Hughey 2007). Some users, especially if they are inexperienced, may also have problems to set their preferences. If so, a fixed weighting matrix, developed by experts, individually or in group decision (Laukkanen et al. 2005), would be an alternative. In this case the worksheet S1 would not be active. Additionally, to include several stakeholder groups (e.g. supply chain managers, biodiversity protection managers, hunters, recreationists), distinct weighting matrices could be created for each group according to their objectives. The tool provides a consistent approach to filter alternatives that are ecologically or technically feasible. Like all deterministic models, it cannot account for the effects of the randomness, but provides the possibility to assess the most common scenarios. The model represents the current situation and assessments of future scenarios require a forecast by the user. An implementation of methods to predict future conditions, e.g. a growth simulator, would be possible (Lexer et al. 2005) but was seen to add too much complexity in this case. Croat. j. for. eng. 33(2012)2

Development of a Multi-Criteria Decision Support Tool for Energy Wood .....(181–198)

The use of indicators present a potential issue with accuracy so it is important (i) to select indicators that are appropriate for the purpose in question, and (ii) to establish a common understanding regarding the meaning of an indicator between the system developer and end-user, who finally must communicate the results to their client (Lexer et al. 2005). Indicators can only be used if they are available for each alternative. The quality and data availability depends often on the experience of the user (for input fields) and on information from previous studies and research projects. In our case, the data quality was sufficient for most of the indicators. For some indicators, additional differentiation would enhance the quality of the technological evaluation and the calculation of overall utility. The impact of nutrient loss was evaluated according to harvesting method and type of biomass used but could be improved by including tree species and utilization type (e.g. thinning, final cut). Similarly for the bearing pressure, including slope conditions, geology, soil type, and morphology with respect to damage related to harvesting method, machine type and experience of the operator, would improve the tool assessment. Finally, safety assessment could be improved by including machine type and the operator experience. With these opportunities in mind, increasing the amount and quality of data to run the tool, will also increase the effort to run the tool as well as the complexity of calculations and analysis. As already mentioned in the weighting process, estimating specification data can be a challenge for the user. To reduce the effort, some specification data was fixed and is not visible to the user. For all other input fields, default values were used from previous research projects and studies. These default values are primarily from Austria, i.e. mountainous conditions. The demonstration example shows that the tool is capable of generating consistent and feasible results in a complex multi-dimensional decision space. The final proof of acceptance of the tool will be provided by end-user adoption. Early feedback has been encouraging, indicating that most steps in the process are intuitive. Editing of customized parameter data sets, as well as printing of reports, are easy and do not require specialized computer knowledge since the tool is implemented within well-known software. To address tool specific issues, each worksheet and input field have user help notes. Based on the results of the system demonstration and of running additional scenarios not included in this report, the following conclusions could be derived for balanced scenarios: Þ Fully mechanized systems should be preferred over highly, partly or non-mechanized systems. Croat. j. for. eng. 33(2012)2

M. Kühmaier and K. Stampfer

Þ Felling and processing with harvester achieves a higher utility value than doing this work with a chainsaw. Þ The method of extracting energy wood showed no significant differences. Only the value for skidding was slightly lower. Þ The cut-to-length method and the tree-lengthmethod were more suitable than the whole tree method. Þ The utilization of whole trees or logs should be preferred over the supply of forest residues. Þ The supply and chipping of energy wood at the terminal or at the plant gained better results than chipping on the forest road. Þ Chipping in the stand and bundling are not recommended because of low productivity. This was the first time a multi-criteria approach was used in decision support for complex EWSM decisions. The implemented supply chains should be continuously checked and updated, and if some new systems were established they should be added into the tool. The significance of C&I will be examined and if necessary completed with further C&I. Nevertheless, the goal should be to get a higher differentiation and more detailed results without making it more complex for the user. Kühmaier and Stampfer (2010) developed a SDSS for timber harvesting systems based on GIS. To include spatial level in EWSM, SDSS should be combined with the algorithms and calculations from the present tool as a goal for future research.

Acknowledgments – Zahvala The authors would like to thank Austrian Research Promotion Agency (FFG) and Climate and Energy Fund for funding this research project. We would also like to thank Mark Brown of the University of the Sunshine Coast for editing and feedback on the manuscript and the tool. For testing purpose the DS tool can be requested under office915@boku.ac.at.

5. References – Literatura Alam, A., Kilpeläinen, A., Kellomäki, S., 2011: Impacts of initial stand density and thinning regimes on energy wood production and management-related CO2 emissions in boreal ecosystems. European Journal of Forest Research, Article in press. Berndes, G., Hoogwijk, M., Van den Broek, R., 2003: The contribution of biomass in the future global energy supply: a review of 17 studies. Biomass and Bioenergy 25(1): 1–28.

193

M. Kühmaier and K. Stampfer

Development of a Multi-Criteria Decision Support Tool for Energy Wood .....(181–198)

Bohlin, F., Roos, A., 2002: Wood fuel supply as a function of forest owner preferences and management styles. Biomass and Bioenergy 22(4): 237–249.

Jylhä, P., Laitila, J., 2007: Energy wood and pulpwood harvesting from young stands using a prototype whole-tree bundler. Silva Fennica 41(4): 763–779.

Bright, R. M., Strømman, A. H., Hawkins, T.R., 2010: Environmental assessment of wood-based biofuel production and consumption scenarios in Norway. Journal of Industrial Ecology 14(3): 422–439.

Kainulainen, T., Leskinen, P., Korhonen, P., Haara, A., Hujala, T., 2007: A statistical approach to assess interval scale preferences in discrete choice problems. Journal of the Operational Research Society 60: 252–258.

Canham, H. O., 1990: Decision matrices and weighting summation valuation in forest planning. North. J. Appl. For. 7: 77–79.

Kangas, J., 1993. A multi-attribute preference model for evaluating the reforestation chain alternatives of a forest stand. For. Ecol. Manage. 59: 271–288.

Cremer, T., Velazquez-Marti, B., 2007. Evaluation of two harvesting systems for the supply of wood-chips in Norway spruce forests affected by bark beetles. Croatian Journal of Forest Engineering 28(2): 145–155.

Kangas, A., Kangas, J., Pykäläinen, J., 2001: Outranking methods as tools in strategic natural resources planning. Silva Fennica 35(2): 215–227.

Emer, B., Grigolato, S., Lubello, D., Cavalli, R., 2011. Comparison of biomass feedstock supply and demand in Northeast Italy. Biomass and Bioenergy 35(8): 3309–3317. EU, 2009: Directive 2009/28/EC of the European Parliament and of the Council of 23 April 2009 on the promotion of the use of energy from renewable sources and amending and subsequently repealing Directives 2001/77/EC and 2003/30/ EC. Official Journal L 140, 05/06/2009 P. 0016–0062. Fath, B. D., Patten, B. C., 1999: Review of the foundations of network environ analysis. Ecosystems 2: 167–179. Fernandes, U., Costa, M., 2010: Potential of biomass residues for energy production and utilization in a region of Portugal. Biomass and Bioenergy 34(5): 661–666. Goicoechea, A., Hansen, D. R., Duckstein, L., 1982: Multiobjective Decision Analysis with Engineering and Business Applications. Wiley, New York, 519 pp. Gómez, A., Rodrigues, M., Montañés, C., Dopazo, C., Fueyo, N., 2010: The potential for electricity generation from crop and forestry residues in Spain. Biomass and Bioenergy 34(5): 703–719. Hakkila, P., 2004: Developing technology for large-scale production of forest chips – Wood Energy Technology Programme 1999-2003. Research report, VTT Processes, 1–99. Hall, D. O., 1997: Biomass energy in industrialised countries – A view of the future. For. Ecol. Manage. 91: 17–45. Heikkilä, J., Laitila, J., Tanttu, V., Lindblad, J., Sirén, M., Asikainen, A., 2006: Harvesting alternatives and cost factors of delimbed energy wood. Forestry Studies 45: 49–56. Howard, A. F., 1991: A critical look at multiple criteria decision making techniques with reference to forestry applications. J. For. Res. 21: 1649–1659. Hwang, C. L., Yoon, K., 1981: Multiple attribute decision making – methods and applications. A state-of-the-art survey. Lecture Notes in Economics and Mathematical Systems, Vol. 186. Springer, New York. Johansson, J., Liss, J.-E., Gullberg, T., Björheden, R., 2006: Transport and handling of forest energy bundles – Advantages and problems. Biomass and Bioenergy 30(4): 334–341.

194

Kangas, J., Kuusipalo, J., 1993: Integrating biodiversity into forest management planning and decision making. For. Ecol. Manage. 61, 1–15. Kangas, A., Kangas, J., Kurttila, M., 2008: Decision Support for Forest Management. Series: Managing Forest Ecosystems Vol. 16, 224 p. Kanzian, C., Holzleitner, F., Stampfer, K., Ashton, S., 2009: Regional energy wood logistics – Optimizing local fuel supply. Silva Fennica 43(1): 113–128. Kärhä, K., 2011: Industrial supply chains and production machinery of forest chips in Finland. Biomass and Bioenergy 35(8): 3404–3413. Kim, J., Realff, M. J., Lee, J. H., Whittaker, C., Furtner, L., 2011: Design of biomass processing network for biofuel production using an MILP model. Biomass and Bioenergy 35(2): 853–871. Krapfenbauer, A., 1983: Von der Streunutzung zur Ganzbaumnutzung. Centralblatt für das Gesamte Forstwesen 100(2–3): 143–174. Kühmaier, M., Kanzian, C., Holzleitner, F., Stampfer, K., 2007: Wertschöpfungskette Waldhackgut. Optimierung von Ernte, Transport und Logistik. Im Auftrag vom BM für Land- und Forstwirtschaft, Land NÖ, Stadt Wien und ÖBf AG, 282 S., Eigenverlag, Wien. Kühmaier, M., Stampfer, K., 2010: Development of a MultiAttribute Spatial Decision Support System in Selecting Timber Harvesting Systems. Croatian Journal of Forest Engineering 31(2): 75–88. Laitila, J., 2006: Cost and sensitive analysis tools for forest energy procurement chains. Forestry Studies 45: 5–10. Laitila, J., 2008: Harvesting technology and the cost of fuel chips from early thinnings. Silva Fennica 42(2): 267–283. Laitila, J., Heikkilä, J., Anttila, P., 2010: Harvesting alternatives, accumulation and procurement cost of small-diameter thinning wood for fuel in Central Finland. Silva Fennica 44(3): 465–480. Laukkanen, S., Palander, T., Kangas, J., Kangas, A. 2005: Evaluation of the multicriteria approval method for timberharvesting group decision support. Silva Fennica 39(2): 249–264. Croat. j. for. eng. 33(2012)2

Development of a Multi-Criteria Decision Support Tool for Energy Wood .....(181–198)

M. Kühmaier and K. Stampfer

Lehtimäki, J., Nurmi, J., 2011: Energy wood harvesting productivity of three harvesting methods in first thinning of scots pine (Pinus sylvestris L.). Biomass and Bioenergy 35(8): 3383–3388.

Rottensteiner, C., Affenzeller, G., Stampfer, K., 2008: Evaluation of the feller-buncher Moipu 400E for energy wood harvesting. Croatian Journal of Forest Engineering 29(2): 117–128.

Lexer, M. J., Vacik, H., Palmetzhofer, D., Oitzinger, G., 2005: A decision support tool to improve forestry extension services for small private landowners in southern Austria. Computers and Electronics in Agriculture 49: 81–102.

Saaty, T. L., 1977: Ascaling method for priorities in hierarchicalstructures. Journal of Mathematical Psychology 15(3): 234–281.

López-Rodríguez, F., Atanet, C.P., Blázquez, F.C., Celma, A.R., 2009: Spatial assessment of the bioenergy potential of forest residues in the western province of Spain, Caceres. Biomass and Bioenergy 33(10): 1358–1366. Magee, J. F., 1964: Decision Trees for Decision Making. Harvard Business Review 42: 126–138. Makeschin, F., 1994: Effects of energy forestry on soils. Biomass and Bioenergy 6(1–2): 63–79. Mendoza, G. A., Sprouse, W., 1989: Forest planning and decision making under fuzzy environments: an overview and illustrations. For. Sci. 35: 481–502. Moiseyev, A., Solberg, B., Kallio, A. M. I., Lindner, M., 2011: An economic analysis of the potential contribution of forest biomass to the EU RES target and its implications for the EU forest industries. Journal of Forest Economics 17(2): 197–213. Næsset, E., 1997: A spatial decision support system for longterm forest management plan by means of linear programming and a geographical information system. Scand. J. For. Res. 12: 77–88. Ovaskaenen, H., Palander, T., Jauhlaenen, M., Lehtemäki, J., Tekkanen, L., Nurmi, J., 2008: Productivity of energywood harvesting chain in different stand conditions of early thinnings. Baltic Forestry 14(2): 149–154+225. Padari, A., Muiste, P., Mitt, R., Pärn, L., 2009: Estimation of Estonian wood fuel resources. Baltic Forestry 15(1): 77–85. Prabhu, R., Colfer, C. J. P., Dudley, R. G., 1999: Guidelines for Developing, Testing and Selecting Criteria and Indicators for Sustainable Forest Management. Center for International Forestry Research Jakarta. Pukkala, T., Kangas, J., 1993: A heuristic optimization method for forest planning and decision making. Scand. J. For. Res. 8: 560–570. Ranta, T., 2005: Logging residues from regeneration fellings for biofuel production-a GIS-based availability analysis in Finland. Biomass and Bioenergy 28(2): 171–182. Ranta, T., Korpinen, O.-J., 2011: How to analyse and maximise the forest fuel supply availability to power plants in Eastern Finland. Biomass and Bioenergy 35(5): 1841–1850.

Sheppard, S. R. J., Meitner, M., 2005: Using multi-criteria analysis and visualisation for sustainable forest management planning with stakeholder groups. For. Ecol. Manage. 207: 171–187. Solomon, D. S., Hughey, K. F. D., 2007: A proposed Multi Criteria Analysis decision support tool for international environmental policy issues: a pilot application to emissions control in the international aviation sector. Environmental Science & Policy 10: 645–653. Spinelli, R., Cuchet, E., Roux, P., 2007: A new feller-buncher for harvesting energy wood: Results from a European test programme. Biomass and Bioenergy 31(5): 255–266. Stampfer, K., Kanzian, C., 2006: Current state and development possibilities of wood chip supply chains in Austria. Croatian Journal of Forest Engineering 27: 135–145. Stampfer, K., Gronalt, M., Holzleitner, F., Kanzian, C., Kühmaier, M., Leitner, T., Zazgornik, J., 2011: Modellentwicklung für ökoeffiziente Versorgungslogistik am Beispiel der Waldbiomasse. Forschungsbericht im Auftrag von FFG und Klima- und Energiefonds im Rahmen 1. Ausschreibung Programmlinie Neue Energien 2020, Wien, 218 S. Sterba, H., 2003: Growth after biomass removal during precommercial thinning. In: Limbeck-Lilineau, B., Steinmüller, Th., Stampfer, K. (Hrsg.), »Austro2003: High Tech Forest Operations for Mountainous Terrain«, Institute of Forest Engineering, 1–9. Tahvanainen, T., Anttila, P., 2011: Supply chain cost analysis of long-distance transportation of energy wood in Finland. Biomass and Bioenergy 35(8): 3360–3375. Vacik, H., Lexer, M. J., 2001: Application of a spatial decision support system in managing the protection forests of Vienna for sustained yield of water resources. For. Ecol. Manage. 143: 65–76. Vainio, P, Tokola, T., Palander, T., Kangas, A., 2009: A GISbased stand management system for estimating local energy wood supplies. Biomass and Bioenergy 33(9): 1278–1288. Windisch, J., Sikanen, L., Röser, D., Gritten, D., 2010: Supply chain management applications for forest fuel procurement – cost or benefit? Silva Fennica 44(5): 845–858.

Röser, D., Pasanen, K., Asikainen, A. 2006: Decision-support program »EnerTree« for analyzing forest residue recovery options. Biomass and Bioenergy 30: 326–333.

Wolfslehner, B., Vacik, H., Lexer, M. J., 2005: Application of the analytic network process in multi-criteria analysis of sustainable forest management. Forest Ecology and Management 207(1–2): 157–170.

Röser, D., Sikanen, L., Asikainen, A., Parikka, H., Väätäinen, K., 2011: Productivity and cost of mechanized energy wood harvesting in Northern Scotland. Biomass and Bioenergy 35(11): 4570–4580.

Wolfslehner, B., Vacik, H., 2008: Evaluating sustainableforest management strategies with the Analytic NetworkProcess in a Pressure-State-Response framework. Journal of Environmental Management 88(1): 1–10.

Croat. j. for. eng. 33(2012)2

195

M. Kühmaier and K. Stampfer

Development of a Multi-Criteria Decision Support Tool for Energy Wood .....(181–198)

6. Annex – Dodatak Specification data for system demonstration: Table A1 Site, stand and environment specification Tablica A1. Stanišne, sastojinske i okolišne pojedinosti Parameter – Parametar Tree species – Vrsta drva

Specification – Specifikacija 75% Spruce, 20% Beech, 5% Ash 75 % smreka, 20 % bukva, 5 % jasen

Average DBH – Srednji prsni promjer

15 cm

Amount of utilization – Iskorištenost pri sječi

280 m3

Transport distance – Udaljenost prijevoza

38 km

Moisture content (fresh) – Sadržaj vlage (svježe)

50%

Moisture content (dry) – Sadržaj vlage (suho)

30%

Storage costs at terminal – Trošak skladištenja na stovarištu

3 €/m3

Storage costs at plant – Trošak skladištenja kod energane

1 €/m3

Relevance in case of joint production – Važnost u slučaju vezane proizvodnje Purchase price – Nabavna cijena Overhead costs – Opći troškovi Revenues – Prihodi

196

10% 1.5 €/m3 4% 20 €/MWh

Croat. j. for. eng. 33(2012)2

Development of a Multi-Criteria Decision Support Tool for Energy Wood .....(181–198)

M. Kühmaier and K. Stampfer

Table A2 Machine specification Tablica A2. Podaci o strojevima Fuel consumption

System costs

Postupci

Potrošnja goriva

Trošak sustava

Proizvodnost

Felling with chainsaw – Sječa stabla motornom pilom

0.53 l/h

33.5 €/h

3.0 m3/h

Felling and delimbing with chainsaw – Sječa stabla i kresanje grana motornom pilom

0.53 l/h

33.5 €/h

2.0 m3/h

Felling and processing with chainsaw – Sječa i izradba drva

0.53 l/h

33.5 €/h

1.5 m3/h

Felling with feller-buncher – Sječa feler bančerom

6.5 l/h

130.0 €/h

5.8 m3/h

Felling and processing with harvester – Sječa i izradba drva harvesterom

15.6 l/h

135.0 €/h

18.0 m3/h

Extracting with forwarder (logs) – Izvoženje forvarderom (oblo drvo)

11.1 l/h

90.0 €/h

15.5 m3/h

Extracting with forwarder (residues) – Izvoženje forvarderom (šumski ostatak)

11.1 l/h

90.0 €/h

4.0 m3/h

Extracting with forwarder (tree) – Izvoženje forvarderom (stabla)

11.1 l/h

90.0 €/h

7.8 m3/h

Extracting with skidder (stem) – Privlačenje skiderom (debla)

7.3 l/h

60.0 €/h

9.0 m3/h

Extracting with skidder (tree) – Privlačenje skiderom (stabla)

7.3 l/h

60.0 €/h

10.0 m3/h

Extracting with tower yarder (stem) – Iznošenje stupnom kamionskom žičarom (debla)

16.0 l/h

130.0 €/h

9.3 m3/h

Extracting with tower yarder (tree) – Iznošenje stupnom kamionskom žičarom (stabla)

16.0 l/h

130.0 €/h

8.5 m3/h

Extracting with shuttle (chips) – Privlačenje šatlom (iver)

10.0 l/h

88.0 €/h

17.1 m3/h

Processing with processor – Izradba procesorom

8.0 l/h

110.0 €/h

10.0 m3/h

Bundling – Izradba svežnjeva

11.5 l/h

120.0 €/h

4.9 m3/h

Chipping with tractor & chipper (logs) – Iveranje obloga drva – traktorom pokretani iverač

34.8 l/h

140.0 €/h

67.0 m3 loose/h

Chipping with tractor & chipper (residues) – Iveranje šumskoga ostatka – traktorom pokretani iverač

34.8 l/h

140.0 €/h

26.0 m3 loose/h

Chipping with tractor & chipper (bundles) – Iveranje svežnjeva – traktorom pokretani iverač

34.8 l/h

140.0 €/h

54.0 m3 loose/h

Chipping with tractor & chipper (trees) – Iveranje stabala – traktorom pokretani iverač

34.8 l/h

140.0 €/h

29.0 m3 loose/h

Chipping with chipper on truck (logs) – Iveranje obloga drva – iverač na kamionu

40.5 l/h

220.0 €/h

134.0 m3 loose/h

Chipping with chipper on truck (residues) – Iveranje šumskoga ostatka – iverač na kamionu

40.5 l/h

220.0 €/h

52.0 m3 loose/h

Chipping with chipper on truck (bundles) – Iveranje svežnjeva – iverač na kamionu

40.5 l/h

220.0 €/h

108.0 m3 loose/h

Chipping with chipper on truck (trees) – Iveranje stabala – iverač na kamionu

40.5 l/h

220.0 €/h

85.0 m3 loose/h

Chipping with mobile chipper in stand (trees) – Iveranje stabala pokretnim iveračem u sastojini

28.0 l/h

155.0 €/h

17.0 m3 loose/h

Transportation (logs) – Prijevoz obloga drva

40 l/100 km

80.0 €/h

Transportation (residues) – Prijevoz šumskoga ostatka

25 l/100 km

65.0 €/h

Transportation (bundles) – Prijevoz svežnjeva

40 l/100 km

82.0 €/h

Transportation (trees) – Prijevoz stabala

25 l/100 km

65.0 €/h

Transportation (chips) – Prijevoz ivera

50 l/100 km

67.0 €/h

Process

Croat. j. for. eng. 33(2012)2

Productivity

197

M. Kühmaier and K. Stampfer

Development of a Multi-Criteria Decision Support Tool for Energy Wood .....(181–198)

Sažetak

Razvoj višekriterijskoga alata za pomoć pri odlučivanju kod upravljanja dobavom energijskoga drva Uporaba drva za ogrjev i proizvodnju energije pokazuje rast u posljednja dva desetljeća uz daljnji porast potražnje. Pridobivanjem se energijskoga drva može ostvariti dodatan izvor prihoda za šumovlasnike, koje općenito pozitivno djeluje na ekonomičnost šumskih operacija, ali i utječe na okolišne i socijalne čimbenike, poput ravnoteže u šumskom okolišu, uvjete rada, razinu zaposlenosti, što također treba vrednovanjem obuhvatiti. Pojedini se izvođači radova suočavaju s brojnim mogućnostima izbora sustava pridobivanja drva za energiju, uz često ograničeno znanje o nji hovim značajkama. Također je važno shvatiti da pojedini izvođači radova često imaju različite predrasude u odnosu na ravnotežu između ekonomskih, okolišnih te socijalnih utjecaja pojedinih sustava pridobivanja drva za energiju. U Microsoft Excelu, uz primjenu Visual Basic for Applications, razvijen je računalni program za potporu pri odlučivanju kao pomoć izvođačima radova u prepoznavanju najpogodnijega lanca dobave energijskoga drva s obzirom na njihove potrebe. Aplikacija uzima u obzir brojne kriterije, kao što su energetska učinkovitost, ravnoteža hraniva, stabilnost i vitalnost sastojine i tla nakon izvođenja radova, kontribucijska marža, jamstvo dobave, razina zaposlen osti te sigurnost pri radu. Korisnicima je omogućen unos stanišnih, sastojinskih te okolišnih podataka i tehnoloških parametara, kao i promjena postavki njihovih osobnih prioriteta za ravnotežu između pojedinih kriterija. Cjelokupna korisnost različitih mogućih postupaka izračunava se pomoću zbirnoga modela korisnosti, koja se grafički i tablično prikazuje na zaslonu računala. Aplikacija je oblikovana radi prijenosa znanja stečenih tijekom istraživanja te predstavlja praktičan i razumljiv alat. Dodatno u predstavljanju alata za pomoć pri odlučivanju ovaj rad istražuje utjecaj ponderiranja kriterija, te renskih prilika, udaljenosti prijevoza te sadržaja vlage na odabir najpogodnijega lanca dobave energijskoga drva. Ključne riječi: potpora odlučivanju, model korisnosti, upravljanje dobavom, energijsko drvo, pridobivanje bio mase, planiranje gospodarenja šumom, višenamjensko šumarstvo

Authors’ addresses – Adrese autorâ:

Received (Primljeno): May 31, 2012 Accepted (Prihvaćeno): July 3, 2012

198

Martin Kühmaier, PhD. e-mail: martin.kuehmaier@boku.ac.at Prof. Karl Stampfer, PhD. e-mail: karl.stampfer@boku.ac.at University of Natural Resources and Applied Life Sciences Vienna Department of Forest and Soil Sciences Institute of Forest Engineering Peter Jordan Straße 82 A-1190 Vienna AUSTRIA Croat. j. for. eng. 33(2012)2

Original scietific paper – Izvorni znanstveni rad

Truck Transportation and Chipping Productivity of Whole Trees and Delimbed Energy Wood in Finland Juha Laitila, Kari Väätäinen Abstract – Nacrtak The present study aims to evaluate the competitiveness of various supply systems of smalldiameter wood harvested from young stands for fuel. Trees were harvested for the cost com parison either as (i) multi-stem delimbed shortwood or (ii) as whole trees or (iii) the harvesting was based on bundle-harvesting using the Fixteri II bundle harvester. The cutting of whole trees and multi-stem delimbed shortwood was carried out using a conventional harvester head equipped with multi-tree-handling accessories. Forwarding was carried out using a standard medium-sized forwarder. The comparison of procurement costs was done at stand level as a function of breast height diameter (5–13 cm) and on-road transportation distance (5–160 km). The harvested wood was chipped either at a roadside landing or at a terminal using a trailermounted drum chipper. The comparison of the supply systems was done using recently pub lished productivity parameters and data obtained from complementary field studies reported herein. According to the cost assessment, whole-tree harvesting and chipping at the roadside landing was almost invariably the most cost-efficient supply system. The cost of whole-tree and multistem delimbed shortwood chips was at the same level when the breast height diameter of the harvested trees was 11 cm (pine) or more. The cutting of whole trees is cheaper, but the cost difference diminished as a function of tree size. The productivity of forwarding, transportation and chipping of multi-stem delimbed shortwood was significantly higher compared to that of whole trees. When applying roadside chipping in Finnish conditions with small and sparsely located forest holdings, acquiring large enough concentrations of wood for profitable production is a great challenge. Machine relocations can be reduced by transporting raw material to terminals or the end-use facility to be chipped. However, the low bulk density of the initial material restricts the operation radius unless the wood biomass is pre-processed. According to the results of our study, harvesting of multi-stem delimbed shortwood is a promising way to simplify operations and to reduce transportation and chipping costs. In the case of whole-tree bundling, savings in transportation and chipping costs did not offset the high felling and compaction costs, and the bundling system was the least competitive alternative. Keywords: whole trees, multi-stem delimbed shortwood, chipping, transportation, small tree harvesting, young stands, terminals

1. Introduction – Uvod The number of heating and power plants using forest chips has increased from 250 units close to 1000 units during the last ten-year period in Finland (Asikainen and Anttila 2009). Furthermore several new Croat. j. for. eng. 33(2012)2

biomass plants are planned or under construction (Laitila et al. 2010a). In 2007 the Council of Europe accepted the proposal of the European Commission that the EU member countries should produce 20% of their energy using renewable sources by the year 2020. Each member country has its own target. The EU obligates

199

J. Laitila and K. Väätäinen

Truck Transportation and Chipping Productivity of Whole Trees ... (199–210)

Finland to increase the share of renewable energy sources in energy consumption from 28.5% to 38% by the year 2020. The Finnish long-term climate and energy strategy assigns wood-based energy an important role in achieving this goal. Currently, processing residues from the forest industry are the most important source of wood-based fuels, but these by-products can be considered to be currently fully utilised (Ylitalo 2011). In addition, the availability of processing residues has decreased during the last few years as a consequence of several closures of pulp and paper mills and the decreased production capacity of sawmills and plywood mills (Kallio et al. 2011; Ylitalo 2011). Thus, the most important means of increasing the consumption of wood for energy in the future is the utilisation of forest chip resources. Accordingly, the Ministerial Working Group of the Finnish Government for climate and energy policy has targeted the use of forest chips for energy, i.e. logging residues and stumps from final fellings and small trees from early thinnings, to 13.5 million solid cubic metres in 2020. In the year 2010, Finnish heating and power plants consumed 16.0 million m³ of solid wood fuels, of which 6.2 million m³ comprised forest chips (Ylitalo 2011). This amount included 2.5 million m3 of whole trees and delimbed trees originating from early thinnings, while logging residues and stumps from clearcutting areas were the raw material for more than half of the forest chips consumed by heating and power plants in 2010 (Ylitalo 2011). The technical harvesting potential of forest chips in Finland has been estimated to be 14–20 million m3 per year (Laitila et al. 2008; Kärhä et al. 2008). About 45% of this amount could be obtained from small-dimension wood from early thinnings (Laitila et al. 2008). According to the most recent estimates (Salminen et al. 2012), some 75% of the sustainable cutting potential of industrial wood is being made use of, and this is assumed to mean an increasing proportion of the energy wood accrued will be composed of stemwood of industrial wood dimensions or dimensions very close to industrial wood. The harvesting of industrial roundwood and energy wood from early thinnings is costly, owing to the small stem size and low removals. Furthermore, dense undergrowth often weakens the profitability of early thinnings (Oikari et al. 2010). In Finland the procurement of thinning wood chips is mainly based on chipping at the roadside storage (73%) or at the terminal (24%) (Kärhä 2011). Chipping at the end-use facility is not so common in thinning wood harvesting compared to logging residue or stump wood chip production. Chipping at the landing

200

is a suitable and quite cost-competitive procurement system for power and heating plants of all size categories. A terminal operates as a buffer storage, which enables a more secure supply of fuel chips, and also serves as a process management tool for the whole supply chain. The terminal is also a compromise between chipping at the landing and at the plant. The raw material is transported in an unprocessed form to the terminal and delivered to the plant as chips. However, the low bulk density of the initial material restricts the operation radius unless the wood biomass is pre-processed. The whole-tree bundler (Fixteri) was developed in order to rationalise the integrated harvesting of smalldiameter energy wood and pulpwood and to reduce transportation costs through load compaction (Jylhä and Laitila 2007; Nuutinen et al. 2011; Jylhä 2011). The transportation can be arranged using standard forwarding equipment and long-distance transportation equipment (Laitila et al. 2009). The bundles are, on average, 2.7 m in length, 0.65 m in diameter, and 0.5 m³ (solid) in volume. Conventional single-grip harvester heads equipped with multi-tree handling equipment are also suitable for cutting whole trees and multi-stem delimbed shortwood (Laitila et al. 2010b; Belbo 2011). The working technique of whole-tree cutting and multi-stem delimbed shortwood cutting is basically the same apart from the delimbing of the tops of tree bunches (e.g. Laitila et al. 2010b). In energy wood cutting with single-grip harvesters, the trees are cut and accumulated to the chamber of the multi-tree handling harvester head. Subsequently the tree bunch is moved to an upright position alongside the strip road for the processing of the trees to forwarding length and piling. In whole-tree harvesting the tree bunch is fed through the feed rollers and delimbing knives of the harvester head up to the crosscutting point of forwarding length (5–7 metres). After cross-cutting, the harvester crane moves the undelimbed top bunch onto the base bunch alongside the strip road. In multi-stem delimbed shortwood harvesting, both the base and top bunch of the trees are fed through the feed rollers and delimbing knives of the harvester head and piled at the side of the strip road. The top bunch is topped at the top diameter of 3–5 cm and the target length of the bolts is usually about 5 metres. After multi-tree delimbing only short bits of branches are left in the tree bunches.

1.1 Aim of the study – Cilj rada This study aims to evaluate the competitiveness of various supply systems of small-diameter wood harCroat. j. for. eng. 33(2012)2

Truck Transportation and Chipping Productivity of Whole Trees ... (199–210)

J. Laitila and K. Väätäinen

Fig. 1 A conventional harvester head equipped with multi-tree-handling accessories (up) and Fixteri II bundle harvester. Chipping of energy wood at a roadside landing or at a terminal using a trailer-mounted drum chipper (down) Slika 1. Harvesterska glava (gore lijevo) i bandler harvester Fixeri II – stroj za izradu svežnjeva (gore desno). Iveranje energijskoga drva na pomoćnom stovarištu ili u energani pomoću priključenoga bubnjastoga iverača (dolje) vested from young stands for fuel. Trees were harvested for the cost comparison either as (i) multi-stem delimbed shortwood or (ii) as whole trees or (iii) the harvesting was based on bundle-harvesting using the newly-developed Fixteri II bundle harvester (Fig. 1). The cutting of whole trees and multi-stem delimbed shortwood was carried out using a conventional harvester head equipped with multi-tree-handling (MTH) accessories. Forwarding was carried out using a standard medium-sized forwarder. The comparison of procurement costs was done at stand level as a function of breast height diameter (5–13 cm) and on-road transportation distance (5–160 km). The harvested wood was chipped either at a roadside landing or at a terminal using a trailer-mounted drum chipper (Fig. 1). Multi-stem delimbed shortCroat. j. for. eng. 33(2012)2

wood and whole-tree bundles were transported to the terminal using a conventional timber truck and whole trees were transported using a biomass truck equipped with solid side panels and bottom (Fig. 2). The chips from the roadside landing and from the terminal were transported using a standard chip truck (Fig. 2). The comparison of the supply systems was done using recently published productivity parameters and data obtained from complementary chipping and transportation studies reported herein. The strategic goal of the study was to improve the cost-efficiency of production of forest chips from wood harvested from young stands, especially by testing technologies for chipping and on-road transporting of small-diameter thinning wood. Updated data on costs and productivity are useful when selecting the appropriate supply systems for wood harvested from various young stands.

201

J. Laitila and K. Väätäinen

Truck Transportation and Chipping Productivity of Whole Trees ... (199–210)

Fig. 2 A biomass truck equipped with solid side panels and bottom (left up), a standard chip truck (right up), and a standard timber truck (down) Slika 2. Kamion za prijevoz biomase s obloženim tovarnim prostorom (gore lijevo), uobičajeni kamion za prijevoz iverja (gore desno) te kamion za prijevoz drva (dolje)

2. Material and Methods – Materijal i metode 2.1 The time studies of truck transportation Studije vremena kamionskoga prijevoza In the truck transportation study, the payload and the loading and unloading times of whole trees were compared to those of multi-stem delimbed shortwood when wood material was transported from a roadside landing to be chipped at the terminal. Both loading and unloading were done using the crane of the timber truck. The studies were carried out during daylight hours in May 2010. The empirical data comprised fourteen truckloads, of which five were multi-stem delimbed shortwood (146 tonnes/184 m³) and nine whole trees (165 tonnes/269 m³). The volumes and payloads of the timber lots were based on measure-

202

ments at the terminal. The dry green densities of multi-stem delimbed shortwood and whole trees were defined according to the SCAN-CM 43:95 standard and their moisture contents according to CEN/TS 1477-2. Sampling of wood was integrated into the terminal chipping experiment, which was carried out immediately after the transporting experiment. The average dry green density of the multi-stem delimbed shortwood was 398 kg/m³ and moisture 49.6%. The corresponding values for the whole trees were 357 kg/ m³ and 41.8%. Whole trees consisted of birch (Betula pendula and Betula pubescens), pine (Pinus sylvestris), spruce (Picea abies) and aspen (Populus tremula). Multistem delimbed shortwood consisted of pine and birch. The average length of the whole trees was 8.7 m and the length of the multi-stem delimbed shortwood was either 3 m or 5 m. The breast height diameters (d1.3) of Croat. j. for. eng. 33(2012)2

Truck Transportation and Chipping Productivity of Whole Trees ... (199–210)

the whole trees and multi-stem delimbed shortwood were 7–9 cm and 10–12 cm, respectively. The timber was transported by a Volvo FH 460 timber truck comprised of a three-axle tractor and a fouraxle trailer equipped with solid side panels and bottom and inside moving carriage (Fig. 2). The truck was equipped with a Kesla 2011ZT hydraulic timber crane. The idea was to load the back end of the carriage first and move it backwards to load the front part last. Thanks to the inside moving carriage, even the rearmost part of the load space could be loaded by using the crane of the timber truck without disconnecting the trailer. One bunch of the whole trees and multistem delimbed shortwood was loaded on the tractor and two or three bunches on the trailer. The total length of the truck-trailer combination was 25.25 m and the bulk volume was 150 m³ (45 m³ + 105 m³). The tare weight of the truck-trailer combination was 29 760 kg, which is about 8 000–10 000 kg heavier compared to a conventional timber truck (Peltola 2004). The time study was carried out manually using the Rufco-900 field computer (Nuutinen et al. 2008). The working time was recorded by applying a continuous timing method, where a clock runs continuously and the times for different elements are separated from each other by numeric codes (e.g. Harstela 1991). The accuracy of the Rufco-900 field computer was 0.6 s (Nuutinen et al. 2008). The harvester’s working time was divided into effective working time (E0h) and delay time (Harstela 1991), which is a common method employed in Nordic work studies. Auxiliary times (e.g. planning of work and preparations) were included in the work phases in which they were observed. The skilful truck driver had extensive working experience. He had fifteen years of working experience in transporting roundwood and two years of working experience in transporting biomass.

2.2 The time studies of chipping – Studije vremena iverača The aim of the time study was to compare the chipping productivity of whole trees and multi-stem delimbed shortwood in terminal conditions. The studies were carried out during daylight hours in May 2010. The mobile industrial chipper used for the experiment was a trailer-mounted Rudnick&Enners MTH 900 X 1000/13 drum chipper. The weight of the tandem axle trailer chipper was 19 000 kg and it was 9 650 mm long, 2 550 mm wide and 4 000 mm high. The height of the feeding tube was 900 mm and its width was 1 000 mm. The chain feed ran longitudinally and the discharge of the chipped material was realised by a pivotable blower with throw control (Rudnick&Enners Maschinen Croat. j. for. eng. 33(2012)2

J. Laitila and K. Väätäinen

und Anlagenbau GmbH 2003). The drum carried four blades and the blades were sharpened at the beginning of the shift. The chipper year model was 2004 and it had been renovated due to a fire in 2007. The chipper was driven independently and it was powered by a 368 kW Scania V8 diesel engine. The towing vehicle was a three-axle Scania 144 G 530 and the tractor was equipped with a heavy-duty Kesla 2010T timber loader used for bringing the wood to the hydraulic in-feed system. The skilful chipper operator had three years of working experience in chipping. The time study was carried out manually using the continuous timing method and field computer (Nuutinen et al. 2008). The dry green density of the chipped material was defined according to the SCANCM 43:95 standard, moisture content according to CEN/TS 1477-2 and particle size distribution according to the CEN/TS 15149-1 standard. The chipped timber was free of impurities (stones, mineral soil, etc.), which enabled uninterrupted chipping work at the asphalt field of the terminal. In the chipping experiment the chips were blown to the ground.

2.3 The cost comparison of the supply systems Usporedba troškova s obzirom na duljinu daljinskoga transporta The logging cost calculations were made for an arti ficial stand, where the cutting removal was 1500 pine (Pinus sylvestris) stems per hectare and the breast height diameter (d1.3) of the removed trees varied between 5–13 cm (Laitila et al. 2010b). The volume of the multi-stem delimbed shortwood as a function of 5–13 cm breast height diameter was 4–73 dm3 and the volume of whole trees was 8–99 dm3, respectively. The harvesting intensity amounted to 12–149 m3/ha when trees were harvested with branches and 7–110 m3/ha when trees were delimbed. The total length of the strip road network at the stand was assumed to be 600 m/ha, based on an average strip road spacing of 20 m (Niemistö 1992). The forwarding distance was 300 m and the on-road transportation distance varied between 5–160 km. The distance between the terminal and enduse-facility was constant, 15 km, except when the terminal was located at the end-use-facility. The productivity of cutting whole trees and delimbed stemwood using the multi-tree processing technique was based on the study of Heikkilä et al. (2005) and the productivity model was published in the Excel-based »Cost calculator for delimbed energy wood”« cost calculation program (Laitila 2006). In the case of bundle harvesting, the productivity model of Nuutinen et al. (2011) was used. The forwarding productivity of multi-stem delimbed shortwood was cal-

203

J. Laitila and K. Väätäinen

Truck Transportation and Chipping Productivity of Whole Trees ... (199–210)

culated using the model of Kuitto et al. (1994). When forwarding loose and bundled whole trees, the functions of Laitila et al. (2007 and 2009) were applied. The payload of the medium-sized forwarder was set at 9 m³ for multi-stem delimbed shortwood, 6 m³ for whole trees, and 24 pieces for bundles (Laitila 2008; Laitila et al. 2009; Laitila et al. 2010). The effective time (E0) productivities of cutting, bundling and forwarding were converted into gross effective time (E15) productivities with the coefficients applied in the study of Jylhä et al. (2010). The hourly costs were as follows: mediumsized thinning harvester equipped with multi-treehandling accessories, 81 €/h (VAT 0%), forwarder, 61€/h (VAT 0%), and Fixteri II bundle-harvester, 107 €/h (VAT 0%) (Laitila and Väätäinen 2011). Multi-stem delimbed shortwood and whole-tree bundles were transported using a conventional timber truck with a trailer, assuming a payload of 48 m³ for multi-stem delimbed shortwood (Laitila and Väätäinen 2011). In the case of bundle trucking, a maximum payload of 100 pieces was used (Laitila and Väätäinen 2011). The payload of loose whole trees was set at 30 m³ on the basis of the study reported herein. The payload of the chip truck was 44 m³ (Laitila 2012). The road transportation times were composed of driving with an empty load, driving loaded and terminal times (incl. loading, unloading, waiting and auxiliary time). Time consumption of driving, with full load and empty load, was calculated as a function of transportation distance according to the speed functions of Nurminen and Heinonen (2007). One bunch of the delimbed 5 m longwood was loaded on the tractor and two bunches on the trailer. The bundle load consisted of two bunches on the tractor and three bunches on the trailer. According to the time study, the loading time of a truck with 5 m multitree delimbed shortwood was 19.0 minutes and the unloading time was 14.2 minutes. For whole-tree bundles, the loading time was 25.1 minutes and unloading time 17.1 minutes (Laitila et al. 2009). In the case of loose whole trees, a loading time of 30.0 minutes and an unloading time of 11.6 minutes were recorded. The preparation time of the crane was 2.6 minutes at the roadside landing and 2.3 minutes at the delivery point for all timber assortments. Cleaning of residues from the roadside landing and delivery point took 4.2 minutes per loose whole-tree or bundle truckload (Laitila et al. 2009). Handling of the inside moving carriage took 1.2 minutes per load when transporting loose whole trees. With multi-stem delimbed shortwood and bundles, the handling times of the bunks and trailer were 1.4 and 2.2 minutes, respectively (Laitila et al. 2009). Binding and opening the load of multi-stem delimbed

204

shortwood took 9.9 minutes, while it took 12.5 minutes with bundles (Laitila et al. 2009). 25 minutes were added to the terminal times, corresponding to the time the trucks spent on items other than actual loading or unloading, e.g. driving at storage point, turning, waiting, weighing, arrangements at the delivery point and sampling. The loading time of the chip truck-trailer unit at the roadside landing was equal to the direct and indirect chipping times. The chipper productivity at the roadside storage was estimated to be 34 m³/E15h for whole trees and at the terminal the chipping productivity was 44 m³/E15h (Laitila 2008). According to the comparative time study reported herein, the chipping productivity of multi-stem delimbed shortwood was 1.22 higher compared to whole trees. This denotes that the chipper productivity at the roadside storage was estimated to be 41 m³/E15h for multi-stem delimbed shortwood and 54 m³/E15h at the terminal. The chipping productivity of whole-tree bundles was estimated to be equal to that of multi-stem delimbed shortwood. 35 minutes were added to the direct and indirect chipping times, corresponding to the time the chip trucks spent on items other than actual loading, e.g. driving at storage point, turning, detaching the trailer, waiting, weighing, unloading and sampling (Laitila 2012). At the terminal, the loading time of chips was estimated to be 15 minutes and the work was done using a wheel loader. The unit cost of chip loading was 1 €/m3 (Laitila 2008). Chips were transported from the terminal to the end use facility with truck-trailer units that are similar to the ones used when transporting chips from the roadside storages. The unloading time of the chips at the end use facility was estimated to be 25 minutes. The hourly cost of the mobile drum chipper was 154.2 €/E15h in the roadside chipping operations and 145.1 €/E15h when operating at the terminal (Laitila and Väätäinen 2011). The timber truck hourly cost was 86.4 €/E15h for driving and 60.7 €/E15h for terminal time (Laitila and Väätäinen 2011). The corresponding values for the chip trucktrailer unit and biomass truck were 88.2 €/E15h & 63.3 €/E15h and 89.0 €/E15h & 63.2 €/E15h, respectively (Laitila and Väätäinen 2011).

3. Results – Rezultati 3.1 The time studies – Studije vremena The average payload of loose whole trees was 30 m3 in the transportation study. The loading productivity of whole trees was 1.0 m3 per minute, and unloading productivity was 2.58 m3 per minute. The loading and Croat. j. for. eng. 33(2012)2

Truck Transportation and Chipping Productivity of Whole Trees ... (199–210)

J. Laitila and K. Väätäinen

tribution of whole-tree and multi-stem delimbed shortwood chips was comparable, excluding the proportion of small particles (<8 mm). This is due to the high percentage of branches and needles in the wholetree material.

3.2 The cost comparison – Usporedba troškova

Fig. 3 Chipping productivity of whole trees & multi-stem delimbed shortwood at the terminal (m3/E0h) and average grapple load in chipper feeding Slika 3. Proizvodnost iveranja cijelih stabala i debala kratkoga drva na glavnom stovarištu (m3/E0h) te prosječan zahvat hvatala prilikom iveranja

The supply system based on whole-tree harvesting resulted in the lowest logging cost in the cost comparison (Fig. 4). The logging costs of whole-tree and multi-stem delimbed shortwood were at the same level when the breast height diameter of the harvested trees was 11 cm (pine) or more. The cutting of whole trees is cheaper, but the cost difference diminished as a function of tree size. The forwarding costs of multistem delimbed shortwood and whole-tree bundles were lower due to higher bulk density and therefore bigger payloads. Also the loading and unloading were more efficient compared to whole trees. In the case of whole-tree bundling. the logging costs exceeded the harvesting of multi-stem delimbed shortwood when the breast height diameter of the removed trees was 8 cm or more (Fig. 4). The transporting costs of loose whole trees were the highest (Fig. 5) and these costs were also the most sensitive to changes in transporting distance. Trans-

unloading productivity of 5 m and 3 m long multistem delimbed shortwood amounted to 2.53 & 3.39 m3 per minute and 1.7 & 2.56 m3 per minute, respectively. When loading or unloading 5 m long multi-stem delimbed shortwood, the size of handling unit (grapple load volume) is significantly bigger compared to 3 m long multi-stem delimbed shortwood, which improves productivity. The average reposition time of the chipper at the terminal – that is, when the chipper is moved within the work site as work proceeds – was 4.5 minutes. Preparing time for chipping, including all preparatory measures from starting the machine to raising the rotation speed of the machine in readiness for chipping, was on average 2.4 minutes at the terminal. The chipping productivity was 55 m3/E0h for whole trees and 67 m3/E0h for multi-stem delimbed shortwood (Fig. 3). The average grapple load volume in chipper feeding was 0.33 m3 (211 kg) for whole trees and 0.49 m3 (317 kg) for multi-stem delimbed shortwood. Feeding of the chipper was effective. The chipper was only running idle during 1.5% of the effective chipping time when chipping whole trees and 0.9% when chipping multi-stem delimbed shortwood. The particle size disCroat. j. for. eng. 33(2012)2

Fig. 4 Logging costs of whole trees, multi-stem delimbed shortwood and whole-tree bundles at the roadside landing. Logging cost, €/m3 = Cutting + Forwarding Slika 4. Troškovi pridobivanja biomase (sječa i privlačenje) iz cijelih stabala, okresanoga kratkoga drva i cijelih stabala u svežnjevima do pomoćnoga stovarišta uz šumsku cestu

205

J. Laitila and K. Väätäinen

Truck Transportation and Chipping Productivity of Whole Trees ... (199–210)

Fig. 5 Transporting costs of chips, loose whole trees, whole-tree bundles and multi-stem delimbed shortwood as a function of transporting distance Slika 5. Usporedba troškova prijevoza iverja, cijelih stabala, svežnjeva cijelih stabala i okresanoga kratkoga drva kao funkcija udaljenosti prijevoza

Fig. 6 Cost structure of whole-tree and multi-stem delimbed shortwood chips at the plant, €/m3. Forwarding distance is 300 m and transporting distance to the plant or the terminal is 50 km Slika 6. Sustav troškova pri iveranju na glavnom stovarišru (energani), €/m3. Udaljenost je privlačenja 300 m, a udaljenost je prijevoza do energane ili glavnoga stovarišta 50 km porting costs of whole-tree bundles and multi-stem delimbed shortwood were about 2 €/m3 lower compared to whole-tree and multi-stem delimbed shortwood chips mainly due to shorter loading and unloading times and somewhat bigger payloads. Also the

206

standard timber truck hourly cost was lower compared to the chip truck-trailer unit. The harvesting cost of whole-tree and multi-stem delimbed shortwood chips at the plant ranged between 42.0 and 51.0 €/m3 when the breast height diamCroat. j. for. eng. 33(2012)2

Truck Transportation and Chipping Productivity of Whole Trees ... (199–210)

eter of the harvested trees was 8 cm (Fig. 6). Whole-tree harvesting and chipping at the roadside landing was the most cost-efficient supply system in the cost assessment, because the cutting of whole trees was significantly cheaper compared to whole-tree bundling or harvesting trees that have been delimbed. The lower comminution cost at the terminal was not enough to cover the extra handling cost of chips at the terminal or the delivery cost to the plant. Fig. 7 shows the harvesting cost of chips at the plant calculated as a function of breast height diameter when the terminal was located in direct connection to the end-use facility. According to the results, the harvesting system based on terminal chipping and multi-stem delimbed shortwood undercut the harvesting cost of the harvesting system based on whole-tree harvesting and roadside chipping when the breast height diameter of the harvested trees was 11 cm (pine) or more. In the case of whole-tree bundling, savings in transportation and chipping costs did not offset the high felling and compaction costs, and the bundling system was the least competitive alternative (Fig. 7).

4. Discussion – Rasprava Whole-tree harvesting and chipping at the roadside landing was the most cost-efficient supply system in the cost comparison. The cost of whole-tree and multi-stem delimbed shortwood chips was at the same level when the breast height diameter of the harvested

J. Laitila and K. Väätäinen

trees was 11 cm (pine) or more. The cutting of whole trees is cheaper, but the cost difference diminished as a function of tree size. The productivity of forwarding, transportation and chipping of multi-stem delimbed shortwood was significantly higher compared to that of whole trees. When applying roadside chipping in Finnish conditions with small and sparsely located forest holdings, acquiring large enough concentrations of wood for profitable production is a great challenge. Machine relocations can be reduced by transporting raw material to terminals or the end-use facility to be chipped. However, the low bulk density of the initial material restricts the operation radius unless the wood biomass is pre-processed. According to the results of our study, harvesting of delimbed energy wood is a promising way to simplify operations and to reduce transportation and chipping costs. In the case of whole-tree bundling, savings in transportation and chipping costs did not offset the high felling and compaction costs, and the bundling system was the least competitive alternative. Eriksson and Björheden (1989) figured out that supply via terminals does not pay off, and our results do not refute this in general. The GIS study of Kanzian et al. (2009) indicate that supply costs will increase by just 10% if half the fuel and 26% if all the fuel goes through terminals in Austria. On the other hand, the additional costs can be designated as expenses necessary to ensure constant supply throughout the year (e.g. Gronalt and Rauch 2007; Kanzian et al. 2009; Laitila et al. 2010a).

Fig. 7 Harvesting cost of chips at the plant as a function of breast height diameter, €/m3 Slika 7. Troškovi pridobivanja drvnoga iverja kao funkcija prsnoga promjera stabla, €/m3 Croat. j. for. eng. 33(2012)2

207

J. Laitila and K. Väätäinen

Truck Transportation and Chipping Productivity of Whole Trees ... (199–210)

Delimbed material produces uniform fuel stock devoid of needles and branches, which may be a benefit especially at some power plants with a restricted capability to handle high levels of chlorine and alkali metals contained in the branch material. Sufficient quantities of alkali metals and chlorine cause agglomeration of bed sand as well as corrosion in fluidised and circulating fluidised bed boilers and heat exchangers (Nurmi and Hillebrand 2007). Since harvesting undelimbed assortments increases nutrient export from the site, which can affect soil productivity, the whole-tree harvesting alternatives included in the present study cannot be recommended on infertile soil stands (Äijälä et al. 2010). However, in Scandinavian studies, no significant growth reductions have been observed due to whole-tree harvesting from Scots pine first-thinning stands within the follow-up periods of at most 20 years (Jylhä 2011). In the simulation by Heikkilä et al. (2007) based on nitrogen removal from the sites, leaving 40% of logging residues on Finnish first-thinning sites, increased thte stem volume by 2.6% over a 10-year post-harvesting period compared to complete whole-tree harvesting.

5. References – Literatura Äijälä, O., Kuusinen, M., Koistinen, A., 2010: Hyvän metsänhoidon suositukset. Energiapuun korjuu ja kasvatus (Forest management recommendations. Energy wood harvesting and silviculture). Metsätalouden kehittämiskeskus Tapio. 56 p. (In Finnish) Asikainen, A., Anttila, P., 2009: Jatkuuko metsäenergian käytön kasvu? (Will the growth of forest energy continue in the future?). In publication: Hänninen, R., Sevola, Y. (Editors). Metsäsektorin suhdannekatsaus 2009-2010. p. 55-57. (In Finnish) Belbo, H., 2011: Efficiency of accumulating felling heads and harvesting heads in mechanized thinning of small diameter trees. Linnaeus University Dissertations No 66/2011. 42 p. Eriksson, L. O., Björheden, R., 1989: Optimal storing, transport and processing for a forest-fuel supplier. European Journal of Operational Research 43(1): 26–33. Harstela, P., 1991: Work studies in forestry. University of Joensuu. Silva Carelica 18. 41 p. Heikkilä, J., Laitila, J., Tanttu, V., Lindblad, J., Sirén, M., Asikainen, A., Pasanen, K., Korhonen, K. T., 2005: Karsitun energiapuun korjuuvaihtoehdot ja kustannustekijät (Harvesting alternatives and cost factors of delimbed energy wood). Working Papers of the Finnish Forest Research Institute 10. 56 p. (In Finnish.) Heikkilä, J., Sirén, M., Äijälä, O., 2007: Management alternatives of energy wood thinning stands. Biomass and Bioenergy (31): 255–266.

208

Jylhä, P., Laitila, J., 2007: Energy wood and pulpwood harvesting from young stands using a prototype whole-tree bundler. Silva Fennica 41(4): 763-779. Jylhä, P., Dahl, O., Laitila, J., Kärhä, K., 2010: The effect of supply system on the wood paying capability of a kraft pulp mill using Scots pine harvested from first thinnings. Silva Fennica, 44(4): 695-714. Jylhä, P., 2011: Harvesting undelimbed Scots pine (Pinus sylvestris L.) from first thinnings for integrated production of kraft pulp and energy. Dissertationes Forestales 133. 73 p. Kallio, M., Anttila, P., McCormick, M., Asikainen, A., 2011: Are the Finnish targets for the energy use of forest chips realistic – Assessment with a spatial market model. Journal of Forest Economics 17 (2011): 110-126 Kanzian, C., Holzleitner, F., Stampfer, K., Ashton, S., 2009: Regional energy wood logistics – optimizing local fuel supply. Silva Fennica 43(1): 113–128. Kuitto, P.-J., Keskinen, S., Lindroos, J., Oijala, T., Rajamäki, J., Räsänen, T., Terävä, J., 1994: Puutavaran koneellinen hakkuu ja metsäkuljetus. (Mechanized cutting and forest haulage). Metsäteho Report 410. 38 p. (In Finnish.) Kärhä, K., Elo, J., Lahtinen, P., Räsänen, T., Keskinen, S., Saijonmaa, P., Heiskanen, H., Strandström, M., Pajuoja, H., 2010: Kiinteiden puupolttoaineiden saatavuus ja käyttö Suomessa 2020 (Availability and use of solid wood fuels in Finland in 2020). Työ- ja elinkeinoministeriön julkaisuja. Energia ja ilmasto. 66/2010. 68 p. (In Finnish) Kärhä, K., 2011: Industrial supply chains and production machinery of forest chips in Finland. Biomass and Bioenergy 35(8): 3404-3413. Laitila, J., 2006: Cost and sensitive analysis tools for forest energy procurement chains. Forestry Studies 45: 5–10. Laitila, J., Asikainen, A., Nuutinen, Y., 2007: Forwarding of whole trees after manual and mechanized felling bunching in pre-commercial thinnings. International Journal of Forest Engineering 18(2): 29–39. Laitila, J., 2008: Harvesting technology and the cost of fuel chips from early thinnings. Silva Fennica 42(2): 267–283. Laitila, J., Asikainen, A., Anttila, P., 2008: Energiapuuvarat (Energy wood resources). In publication: Kuusinen, M. & Ilvesniemi, H. (Editors). Energiapuunkorjun ympäristövaikutukset, tutkimusraportti. Publications of Tapio and Metla. p. 6–12. (In Finnish) Laitila, J., Kärhä, K., Jylhä, P., 2009: Time Consumption Models and Parameters for Off- and On-road Transportation of Whole-tree Bundles. Baltic Forestry 15: 105–114. Laitila, J., Leinonen, A., Flyktman, M., Virkkunen, M., Asikainen, A., 2010a: Metsähakkeen hankinta- ja toimituslogistiikan haasteet ja kehittämistarpeet (Challenges and development needs of forest chip procurement and delivery logistics). VTT Tiedotteita Research – Notes 2564. 143 p. (In Finnish) Croat. j. for. eng. 33(2012)2

Truck Transportation and Chipping Productivity of Whole Trees ... (199–210) Laitila, J., Heikkilä, J., Anttila, P., 2010b: Harvesting alternatives, accumulation and procurement cost of small-diameter thinning wood for fuel in Central-Finland. Silva Fennica 44(3): 465–480. Laitila, J., Väätäinen, K., 2011: Kokopuun ja rangan autokuljetus ja haketustuottavuus (Truck transportation and chipping productivity of whole trees and multi-stem delimbed shortwood). Metsätieteen aikakauskirja 2/2011:107-126. (In Finnish) Laitila, J., 2012: Methodology for choice of harvesting system for energy wood from early thinning. Dissertationes Forestales 143. 68 p. Niemistö, P., 1992: Runkolukuun perustuvat harvennusmallit (Thinning models based on the number of stems). Finnish Forest Research Institute, Research Papers 432. 18 p. (In Finnish). Nurmi, J., Hillebrand, K., 2007: The characteristics of wholetree fuel stock from silvicultural cleanings and thinnings. Biomass and Bioenergy 31(6): 381–392. Nuutinen, Y., Väätäinen, K., Heinonen, J., Asikainen, A., Röser, D., 2008: The accuracy of manually recorded time study data for harvester operation shown via simulator screen. Silva Fennica 42(1): 63–72. Nuutinen, Y., Kärhä, K., Laitila, J., Jylhä, P., Keskinen, S., 2011: Productivity of whole-tree bundler in energy wood

J. Laitila and K. Väätäinen

and pulpwood harvesting from early thinnings. Scandinavian Journal of Forest Research 26: 329-338. Oikari, M., Kärhä, K., Palander, T., Pajuoja, H., Ovaskainen, H., 2010: Analyzing the views of wood harvesting professionals related to the approaches for increasing the cost-efficiency of wood harvesting from young stands. Silva Fennica 44 (3): 481–495. Peltola, J., 2004: Puutavara-autojen rakenteen vaikutus omamassaan (The effect of the structure of timber trucks on their mass). Metsäteho Report 176. 22 p. (In Finnish). Rudnick & Enners Maschinen- und Anlagenbau GmbH., 2003: Betriebsanleitung. Mobiler Trommelhacker REMTH 900x1000/13. Kom. Nr. 3246. 41 p. Salminen, O., 2012: Ainespuun kestävistä hakkuumahdollisuuksista hyödynnetään 75 prosenttia. Energiapuuhakkuut tehostavat puunkäyttöä (Cuttings of industrial roundwood are 75% of the sustainable supply potential. Energy wood harvesting will improve the utilization of wood). Metlan Tiedote 23.1.2012. Release available on the Internet: http:// www.metla.fi/tiedotteet/2012/2012-01-23-ainespuu.htm (In Finnish) Ylitalo, E., 2011: Puun energiakäyttö 2010 (Wood consumption in energy generation in the year 2010). Metsätilastotiedote 16/2011. 7 p. (In Finnish)

Sažetak

Kamionski prijevoz i proizvodnost iveranja cijelih stabala i okresanoga kratkoga drva u Finskoj Ovo je istraživanje usmjereno na ocjenu konkurentnosti nekoliko sustava pridobivanja energijskoga drva iz mladih sastojina. Za usporedbu troškova sustava rada korištena je biomasa iz triju izvora: (i) energijsko drvo iz okresanoga kratkoga drva, (ii) cijelih stabala i (iii) svežnjeva načinjenih pomoću novorazvijenoga stroja bandlera harvestera (stroja za izradu svežnjeva) Fixteri II (slika 1). Sječa stabala provedena je pomoću harvestera, a privlačenje je obav ljeno forvarderom. Usporedba troškova pridobivanja biomase učinjena je na razini sastojine kao funkcija prsnoga promjera stabala (5–13 cm) i udaljenosti prijevoza (5–160 km). Iveranje je biomase obavljeno na pomoćnom stovarištu uz šumsku cestu ili u energani (glavnom stovarištu) pomoću priključenoga bubnjastoga iverača. Okresano kratko drvo i svežnjevi cijelih stabala prevezeni su do energane pomoću kamionskoga skupa, dok su cijela stabla prevezena kamionima s obloženim tovarnim prostorom. Sustav je rada ocijenjen pomoću nedavno objavljenih podataka o proizvodnosti te usporedbom sličnih istraživanja. Cilj je istraživanja bio poboljšati troškovnu učinkovitost proizvodnje drvnoga iverja dobivenoga iz mladih sastojina, a posebno ispitati tehnologije iveranja tankoga drva iz proreda na pomoćnom stovarištu. Prema procjeni troškova iveranje cijelih stabala na pomoćnom stovarištu uz šumsku cestu bilo je gotovo uvijek najučinkovitiji sustav pridobivanja biomase. U slučaju izrade svežnjeva uštede u prijevozu i iveranju ipak nisu dostigle troškove nastale pri sječi i izradi svežnjeva, pa je taj sustav najmanje konkurentan. Trošak pridobivanja iverja iz cijelih stabala i iz okresanoga kratkoga drva bio je jednak pri prsnom promjeru stabala od 11 cm (bor) ili više. Privlačenje debala i svežnjeva biomase isplativije je zbog veće iskoristivosti tovarnoga prostora (veća gustoća tovara). Također, utovar i istovar bili su učinkovitiji u odnosu na utovar i istovar cijelih stabala. Croat. j. for. eng. 33(2012)2

209

J. Laitila and K. Väätäinen

Truck Transportation and Chipping Productivity of Whole Trees ... (199–210)

Prosječan tovar cijelih stabala pri prijevozu bio je 30 m3, utovar je bio 1,0 m3 po minuti, a istovar 2,58 m3 po minuti. Utovar i istovar debala (duljine od 3 i 5 m) iznosio je 2,53 i 3,39 m3 po minuti te 1,7 i 2,56 m3 po minuti. Pri utovaru debala duljine 5 m mnogo je bolja iskoristivost hvatala, pa tako i proizvodnost raste. Priprema za iveranje, uključujući sve pripremne radove od pokretanja stroja do povećanja brzine iverača, trajala je u prosjeku 2,4 minute. Proizvodnost iveranja bila je 55 m3/E0h za cijela stabla i 67 m3/E0h za iverje od okresanoga kratkoga drva. Prosječan zahvat hvatalom iznosio je 0,33 m3 (211 kg) za cijela stabla i 0,49 m3 (317 kg) za okresana kratka drva. Raspodjela veličina čestica sječke dobivenih od cijelih stabala i okresanoga kratkoga drva usporediva je, isključujući udio malih čestica (< 8 mm) koje nastaju zbog visokoga postotka grana i iglica u sustavu iveranja cijelih stabala. Pri primjeni iveranja na pomoćnom stovarištu uz šumsku cestu u finskim uvjetima s malim i prostorno udaljen im šumskim poduzećima velik je izazov jer je potrebna velika količina biomase da bi sustav njezina pridobivanja bio isplativ. Premještanje strojeva može se izbjeći prijevozom sirovine (biomase) do energana i glavnih stovarišta gdje se potom obavlja iveranje. Prema rezultatima ovoga istraživanja pridobivanje iverja iz okresanoga kratkoga drva može smanjiti troškove prijevoza i iveranja biomase. Deblovna metoda stvara jedinstvenu i homogenu zalihu goriva oslobođenu iglica i grana koje mogu biti prednost pogotovo u onim energanama s ograničenim mogućnostima za obradu visoke razine klora i alkalijskih metala sadržanih u granama stabala. Uporaba cijelih stabala prilikom pridobivanja biomase smanjuje proizvodnost tla jer se nutri jenti iznose iz šume, stoga se takvi sustavi pridobivanja ni ne preporučuju. Ključne riječi: iveranje cijelih stabala, iveranje okresanoga kratkoga drva, prijevoz, glavno stovarište

Authors’ address – Adresa autorâ:

Received (Primljeno): March 22, 2012 Accepted (Prihvaćeno): July 17, 2012

210

Juha Laitila, PhD. e-mail: juha.laitila@metla.fi Kari Väätäinen, MSc. e-mail: kari.vaatainen@metla.fi Finnish Forest Research Institute Joensuu Research Unit P.O. Box 68, FI-80101 Joensuu FINLAND Croat. j. for. eng. 33(2012)2

Original scietific paper – Izvorni znanstveni rad

Utilization, Cost, and Landowner Return from Whole-Tree Chipping Young Loblolly Pine Thinnings Mathew F. Smidt, John McDaniel Abstract – Nacrtak In the southern United States thinning loblolly pine to produce whole-tree and clean chips for energy is likely to compete with thinning for roundwood pulpwood as bioenergy markets ex pand. Since chip harvests have higher yields per hectare and can tolerate smaller tree size, comparisons between harvesting costs and associated landowner returns are difficult to make. We completed a gross and continuous timing study on a 67 hectare whole-tree chip harvest (skidder, 2 feller-bunchers, disk chipper, and tracked loader) in southern Alabama in order to compare production rates, harvest costs, and landowner returns. The stand was 12 year old loblolly pine established on retired crop land with an average total biomass volume of 195 green tons (gt) ha-1 and average stem volume 0.22 gt tree-1 and 0.19 gt tree-1 for harvested trees. The whole-tree pine chip harvest totaled 98 gt ha-1 compared to estimated roundwood stem volume of 48 gt ha-1 and clean chip volume of 73 gt ha-1. Machine production averaged 49.5 gt PMH-1 for felling and 61.7 for skidding and 72.8 for loading, but loader production was limited by transportation and market availability. On average the crew produced 332 gt each 10 hour shift and skidding would have limited production at 392 gt PMH-1. Simulations generated from continuous timing data were used to compare harvest costs and potential landowner revenue from the site for whole-tree chips, clean chips and roundwood pulpwood. Keywords: whole tree, chipping, biomass, logging, economics, thinning, loblolly pine

1. Introduction – Uvod Opportunities for early thinnings in loblolly pine are available due to interest by landowners in shortening rotation length and increasing early stand revenue. Lower density planting (South 2003) and early thinnings (Stiff and Stansfield 2003) may lead to earlier sawlog harvests. While landowner guides indicate that stands may be thinned once they average DBH greater than 15 cm, height greater than 12 m, and basal area more than 22 m2ha-1, lower harvesting revenue may result from early thinning due to reduction in harvesting efficiency (Traugott and Dicke 2006). Lower harvest efficiency of roundwood is produced primarily by the small volume per tree. Truck payloads are reduced and handling of more stems per unit volume increases harvesting cost. In-woods chipping may increase transportation efficiency by ensuring maximum payloads. In-woods whole-tree Croat. j. for. eng. 33(2012)2

chipping reduces harvest cost by effectively increasing net volume per stem by harvesting tops, limbs and some needles. In-woods clean chipping effectively increases the value of the material through improved yield of acceptable chips to the mill when compared to roundwood (Stokes and Watson 1991). While markets are available for both whole-tree chips for energy (cogeneration facilities at pulp and paper mills) and clean chips for pulp and paper, they are much smaller than the roundwood market. Many proponents of bioenergy believe that the market for whole-tree chips or clean chips might expand to include electric power generation and eventually production of transportation fuels. Currently pellet mills purchase roundwood pulpwood directly in competition with pulp and paper. Differences in end product value, thinning yield, and harvest costs of the merchandizing options make

211

M. F. Smidt and J. McDaniel

Utilization, Cost, and Landowner Return from Whole-Tree Chipping Young... (211–223)

it difficult for both procurement and landowners to evaluate the choices. The net yield of acceptable chips was slightly greater for in-woods chips vs roundwood because some of the forest residue from roundwood production would be delivered as clean chips (Stokes and Watson 1991). Logger’s ability to add value through chipping may also provide an incentive for in-woods clean chipping. In the region whole-tree chipping is generally limited to stands where the merchantable volumes are low due to tree size or species, but the total volume per hectare and per stand are high. Delivered prices for energy wood do not usually provide enough income for competitive stumpage prices, but energy wood harvest may provide a landowner service by increasing management alternatives and reducing site preparation costs. In addition harvesting roundwood residuals may provide a landowner service or increase utilization of the harvesting system (Bakeret et al. 2010). Integrated harvests of roundwood and biomass chips (Baker et al. 2010; Bolding and Lanford 2005; Han et al. 2004; Puttock 1995) and round-wood and clean chips (Greene and Carruth 1994; Shrestha and Lanford 2002; Spinelli et al. 2008) are possible when it is desirable to merchandize higher value material, but the low chipper utilization results in higher chipping costs. Since chipping capital costs are high, one of the largest factors in maintaining the competitiveness of in-woods chipping versus roundwood harvest will be the potential to increase and maintain chipper utilization. While the harvesting system plays some role in chipper utilization, the dominant effect is the availability of trucking resources and market. Often modeling of harvesting systems assumes that a truck will be available when a load is produced at the landing. Trucking and markets constraints are among the most important limits to system production (Greene et al. 2004), so modeling that does not address that scarcity in a complex way will overestimate system production. The objectives of this study were to describe the production of whole-tree chips from an early thinning of a loblolly pine plantation and use those productions and stand data to simulate the harvest of a similar stand by whole-tree chipping, in-woods clean chipping, and roundwood pulpwood.

2. Methods – Metode We monitored the thinning of 67 hectare, 12 year old loblolly pine (Pinus taeda) stand. The stand was established by planting 1600 trees per hectare on retired cropland that was grassland at the time of planting. The site is in Conecuh County, Alabama, USA, in

212

the lower coastal plain physiographic region. Soils had clayey surface and subsurface horizons from marine parent materials and were mapped as a Vertic Hapludalf. Slopes were gentle (< 5%) with occasional hills with chalk outcrops. Most of the rock outcrops were in hardwood or grass cover because they were not planted or the seedlings did not survive in the thinner soils. The thinning occurred in October and November 2011 with a plan to remove every 5th row and select small trees and trees with defects from the remaining rows to yield a post-harvest basal area of 13 m2ha-1. The stand was sampled prior to harvest with 46–0.02 hectare plots (0.05 acre) set on a rectangular grid. Measurements in the plots included DBH for each tree and a subsample of total height from three trees closest to the plot center. A model of height to diameter was developed from those subsampled height data. Regionally developed tree weight equations (Clark and Saucier 1990) were used to estimate total biomass and merchantable stem mass to a top diameter of 6.3 cm (2.5 in). Mass per stem expected to yield clean chips was estimated at 62% of total green weight (Watson and Stokes 1994). Following the harvest, we sampled the same plots to determine the distribution of both the thinned trees and the residual stand. Post-harvest sampling included an assessment of the woody debris in the stand that resulted from breakage and loss during the harvest. The volume of debris piles at the landings was also estimated. The logging system was equipped to produce whole-tree chips delivered to pulp and paper cogeneration facilities. The crew consisted of two 720 Tigercat feller-bunchers, a 620 Tigercat skidder and a T250B Tigercat tracked loader. There was an older model wheeled loader at the landing to clear debris. The chipper was a Precision Husky model 3086. The logger equipped the chipper with an in-feed conveyor manufactured on the frame of a log trailer. The conveyor increased chipper production since the in-feed conveyor was full consistently during chipping. Trucking was shared among the contractor’s two chipping crews and the principle markets for this harvest were 48 and 61 km from the harvest site. The markets limited production to about 1400 gt per week. The fellerbuncher and skidder were equipped with MultiDat data recorders with Garmin GPS 15 receivers that were set to collect a position every 30 seconds. The loader was equipped with a MultiDat without a GPS receiver. We placed a monitor on the chipper that subsequently malfunctioned so no direct chipper use data were available. On November 8 and 17 we visited the site to collect video of all the machines on the site. The chipCroat. j. for. eng. 33(2012)2

Utilization, Cost, and Landowner Return from Whole-Tree Chipping Young... (211–223)

M. F. Smidt and J. McDaniel

Table 1 Element and variable descriptions from continuous timing of the whole-tree chipping operation Tablica 1. Opis radnih zahvata i varijabli pri iveranju cijeloga stabla Machine Stroj

Element Zahvat

Description – Opis

Move and Cut

From approach of 1st tree cut to the cut of the last tree in the felling head

Kretanje i sječa

Od primicanja prvomu stablu, sječe do sječe zadnjega posječenoga stabla

Felling

Move and Dump

From the cut of the last tree in the felling head to the release of trees from the felling head

Obaranje

Kretanje i odlaganje

Od sječe zadnjega stabala do otpuštanja stabala iz sječne glave vozila

Cycle time

Begins and ends with release of trees from the felling head

Turnus vremena

Počinje i završava otpuštanjem stabala iz sječne glave vozila

Travel empty

Movement away from the landing to the forest until the machine stops near the first load

Prazno vozilo

Kretanje s pomoćoga stovarišta do sastojine sve dok se vozilo ne zaustavi radi utovara

Load

From the time the machine stops near the load to forward motion after the trees are in the grapple

Utovar

Od vremena zaustavljanja vozila blizu oborenih stabala radi utovara do ponovnoga kretanja vozila s punim hvatalom

Skidding

Travel loaded

From forward motion after the trees are grappled to include all forward motion until the grapple is opened at the landing

Privlačenje

Utovareno vozilo

Od kretanja vozila s punim hvatalom sve do pražnjenja hvatala na pomoćnom stovarištu

Pile

Activity on the landing after the grapple is opened until motion to the forest

Uhrpavanje

Radnje na pomoćnom stovarištu nakon pražnjenja hvatala pa sve do povratka vozila u sastojinu

Cycle time

Begins and ends with beginning of travel unloaded element

Loader Utovarivač

Turnus vremena

Počinje i završava kretanjem praznoga vozila

Load

Grapple swings from the pile to to in-feed deck and return

Utovar

Zamah hvatala od složaja do utovarnoga prostora i nazad

Delay

Loader stationary with engine running for less than 9 minutes

Kašnjenja

Stanka utovarivača s upaljenim motorom duže od 9 minuta

Pile

Grapple swings from the skidder unloading area to the tree pile

Uhrpavanje

Zamah hvatalom od mjesta istovara skidera do složaja drva

Clean-up

Loader activity to move limbs or broken tops from the in-feed area to the tree pile or slash (discard) pile

Čišćenje

Radnje micanja dijelova stabala s utovarnoga prostora do složaja otpada

Variable – Varijable

Description – Opis

Total stems

All trees cut in the cycle

Felling

Ukupno oboreno

Sva stabla posječena u radnom turnusu

Obaranje

Hardwood stems

All hardwood stems cut in the cycle

Oblovina

Sva pridobivena oblovina u jednom turnusu

Distance

Straight line distance (m) from active landing center to largest load in cycle

Skidding

Udaljenost

Pravocrtna udaljenost (m) od pomoćnoga stovarišta do mjesta utovara

Privlačenje

Load Number

Number of loading elements in the cycle

Broj utovara

Broj utovarnih elemenata u radnom turnusu

per and loader were observed using a video camera on a tripod just off the landing. We mounted VIO POV cameras on the feller-buncher and the skidder. The work captured on the video was analyzed using TimerPro software and cycle and element data were exported for further analysis. For the skidder we synchronized the time study data with the GPS positions Croat. j. for. eng. 33(2012)2

from the MultiDat and imported them into ARCMAP 10 to determine the straight line distance from the landing to the pile collected during each cycle. Element definitions and independent variable descriptions are presented in Table 1. For each of the time study periods we counted trees in sample of piles built by the feller-buncher to develop a distribution of pile

213

M. F. Smidt and J. McDaniel

Utilization, Cost, and Landowner Return from Whole-Tree Chipping Young... (211–223)

sizes. Tree count in piles varied due to tree size because the feller-buncher attempted to build piles of equivalent mass.

2.1 Simulations – Simulacije We used the regressions and means from the cycle time, production data, and stand parameters from this harvest to develop parameters for a harvest simulation of a whole-tree chipping operation. Simulations of a clean chipping operation and a roundwood pulpwood operation were developed using parameter estimates from published sources. Simulations were developed and run in Stella 9.03. Stella is a dynamic simulation program, which was used to incorporate regression equations and stochastic events to generate production flow on an hourly basis. Each of the simulations harvested 11 140 trees over 20 hectares. Felling production (trees pmh-1) was determined by the felling regression model and that used a bunch size from a normal distribution of bunch sizes from study data and pile size selected from a normal distribution from study data. Skidding productivity was based on a distribution of skid distances from a circular harvest area with the landing at the center. Tracts feasible for chipping are usually 50 ha or larger and would have access roads through the tract to enable landing densities often less than 20 ha per landing. Skidding productivity (trees pmh-1) was determined from regression equations using skid distance and load number. Skid distance was selected randomly from zones of equivalent area (4 hectare) in concentric circles around the landing. Pile size was selected from normal distribution of pile sizes from study data. If landing volume exceeded four truck loads of trees, skidding production ceased for that period. For landing processes production was stimulated by truck arrival based on Poisson distributions with means of 1.2, 1.5, and 1.75 trucks per hour. Production rate of the loader or chipper could be limited by truck arrival, wood available on the landing, or production capacity of chipping or delimbing. For chipper production rate for whole-tree chipping we used the estimate from this study. Clean chip production rates have been estimated for softwood, 26 and 31 gt pmh-1 (Lambert 1987), 31 gt pmh-1 (Raymond and Franklin 1990), but those production rates may be limited by harvesting rates or truck supply. Detailed measurement from a clean chipping operation showed that about 62% of productive time was involved in chip production that resulted in a rate of 53 gt pmh-1 (Franklin 1992). The resulting specific flail and chipping rate would be 85 gt pmh-1, and we used 80 gt pmh-1 for the maximum rate. For roundwood production, the production rate of a Chambers De-

214

liminator was assumed to be normally distributed with a mean of 414 trees pmh-1 with a standard deviation of 150. These estimates were based on published data (Mooney et al. 2000) and unpublished data used in (Folegatti et al. 2007). The loading rate for the roundwood operation was determined by truck arrivals, the presence of delimbed trees on the landing, or a maximum rate of 4 truckloads pmh-1. Machine costs for the simulation were prepared using a before tax cash flow cost that developed costs for a similar firm with equipment that ranges in age and presents capital for the current year (Smidt et al. 2009). Variable costs estimates were developed using standard rules of thumb (Brinker et al. 2002; Caterpillar 1996). Overhead costs were assumed to be 4% of total costs and a capital return of 10% was estimated. Variable and fixed costs estimates are presented in Table 2.

3. Results with Discussion – Rezultati s raspravom 3.1 Harvesting study – Studija sječe i privlačenja We estimated that the stand had 892 trees ha-1 and 26 m2ha-1 of basal area prior to thinning (Table 3). Stand data indicate that it was available for thinning (Traugott and Dicke 2006). Average total biomass (stem branches and foliage) per tree of pre-harvest, post-harvest, and harvested trees were estimated at 0.21, 0.26, and 0.18 gt, respectively. Post-harvest basal area was estimated at 12 m2ha-1. Post-harvest sampling estimated biomass removal volume at 88 gt ha-1 while the total removal mass from load data was 6 668 gt or 99.5 gt ha-1. With a standard error of 4.5 gt ha-1 the sample mean was significantly different from harvested volume. The difference could be related to a combination of sampling error, the biomass equation, or harvest of small hardwood stems (< 7.5 cm dbh) that were not sampled in the plots. Line transects sampled following the harvest (3.30 m transects at each plot center) found a very little volume per hectare of live branches lost from trees during the harvest (standing or harvested trees). Approximately 1 500 m3 of loosely piled slash remained on the landing following the harvest, which may sum to 300 m3 of solid wood or about 270 gt (Hardy 1996). The slash piles were tree parts that had significant contact with the soil (piled up with blade or driven over) and were not chipped to maintain lower ash content in delivered whole-tree chips. If added to the yield of whole-tree chips, the apparent error of the sample increases. Croat. j. for. eng. 33(2012)2

Utilization, Cost, and Landowner Return from Whole-Tree Chipping Young... (211–223)

On weekdays operators worked 10 hour days from 6 AM to 5 PM with a total break time of about 1 hour. On some weekends, when markets were open, operators worked on-site long enough to load the available trucks and/or set up the equipment or wood for the following workday. While at least one crew member worked one Sunday, no chips were produced. The crew worked 3 Saturdays over 5 weeks and produced just 2 loads of chips. Production from Wednesday and Thursday accounted for nearly 50% of total production. Gross machine time rates were 49 gt pmh-1 for the feller-buncher; 61.7 for the skidder and 72.8 for the loader. Utilization rates for the three machines were

M. F. Smidt and J. McDaniel

36%, 49%, and 42%, respectively. Utilization rate for the feller-buncher was lower because there were two machines on site for most of the harvest. Chipper utilization rate was not measured directly but would be even lower than the loader since the loader was active on the landing when the chipper was idle. The summary of the element and cycle time analysis is presented in Table 4. We observed the production of 18 loads of chips with an average time of 9 minutes per load. Load sizes averaged about 24.5 gt. Skid distances were the straight line distance between the landing and the pile and were well distributed during the observation period. The feller-buncher averaged about

Table 2 Cost estimates for three harvesting systems Tablica 2. Procjene troškova sustava pridobivanja ivera Whole-tree chipping

Description

Roundwood

Opis

Oblo drvo

Iveranje cijelih stabala

Čisto iveranje

Feller-buncher – Sječno vozilo

31.22

Skidder – Skider

20.48

Loader – Utovarivač

35.69

Chain flail delimber – Sječna transportna traka

10.77

–

Chipper – Iverač

–

56.86

–

Chipper with Flail – Iverač s transportnom trakom

–

89.79

Wheeled loader – Kotačni utovarivač

–

0.15

Feller-buncher – Sječno vozilo

39.56

Skidder – Skider

36.21

Loader – Skider

28.21

Chain flail delimber – Sječna transportna traka

22.46

–

Chipper – Iverač

–

117.35

–

Chipper with Flail – Iverač s transportnom trakom

–

154.03

Wheeled loader – Kotačni utovarivač

–

36.91

1 Feller-buncher – 1 sječno vozilo

69.78

92.19

Trošak radnika ($ smh )

2 Feller-bunchers – 2 sječna vozila

92.19

114.59

Overhead ($ smh–1)

1 Feller-buncher – 1 sječno vozilo

17.61

21.03

23.62

Operativni troškovi ($ smh )

2 Feller-bunchers – 2 sječna vozila

18.97

22.78

25.37

Capital return ($ smh–1)

1 Feller-buncher – 1 sječno vozilo

20.68

29.91

35.30

Povrat ulaganja ($ smh )

2 Feller-bunchers – 2 sječna vozila

23.35

32.58

37.97

Beginning of year capital value ($)

1 Feller-buncher – 1 sječno vozilo

500 000

740 000

881 000

Vrijednost na početku rada ($)

2 Feller-bunchers – 2 sječna vozila

569 000

810 000

951 000

Expected book value decline ($)

1 Feller-buncher – 1 sječno vozilo

116 000

174 000

209 000

Funkcionalna amortizacija ($)

2 Feller-bunchers – 2 sječna vozila

133 000

191 000

226 000

Component – Sastavnica

Fixed Cost ($ smh–1) Fiksni troškovi ($ smh–1)

Variable Cost ($ pmh–1) Varijabilni troškovi ($ pmh–1)

–1

Labor and fringe cost ($ smh ) –1

–1

Croat. j. for. eng. 33(2012)2

Clean chipping

215

M. F. Smidt and J. McDaniel

Utilization, Cost, and Landowner Return from Whole-Tree Chipping Young... (211–223)

5.2 trees per minute in the stand and it took 3 or more cycles or more than 4.8 minutes to produce one pile for the skidder. The load sheets kept by the loader operator recorded the departure time, destination and product for each load. Load sheets were used to generate a distribution of loading times and compared to the time study data. Given the expected low precision in time recording intervals by the loader operator, the intervals were estimated in 10 minute classes. Interval data were similar for time study and load sheet (Fig. 1). Since both time study periods were Thursdays, the longer between truck intervals were eliminated. The truck departure frequency was 0 trucks for 31% of operating hours, 1 for 33%, 2 for 24%, 3 for 9%, and 4 for 3% resulting in an average rate of trucks at the landing of 1.2 per hour (variance 1.15). Assuming that arrival and departure rates were similar, the mean time between trucks would be (1/arrival rate) about 50 minutes per truck. For the sake of comparison, the mean time between trucks from the data in Fig. 1 was 37 minutes. With a service rate equal to the mean loading time of 6.7 trucks hr–1 (9 minutes per load), the average utilization would be 18% (1.2/6.7). Since the skidder is only

capable of 61.7 t pmh–1 the maximum service rate was 2.5 loads hr–1 (24.5 t load–1). Utilization rate would then be (1.2/2.5) or 48% which is approximately the utilization rate of the skidder. Using 70% of the system capacity would require a truck arrival rate of 1.75 hr–1 and service time per truck would go from 46 minutes to 80 minutes yielding potential time lost at the landing of 16.5 hours per shift (17.5 * (1.33 – 0.4)). Skidder and feller-buncher model statistics for the delay free cycle and element times are presented in Table 5. The regression of the feller-buncher cycle time was significant but with low R2. Any significance of the feller-buncher cycle time was due to strong relationship between Move and Cut time and the dependent variable. Both skidder models were significant, as they were both dependent variables (straight line distance and loads per cycle). Model parameter estimates for the models used in the simulation are given in Table 6. To simulate skidding roundwood to a landing with a flail delimber, we used skidding data from another study in a similar stand with similar data collection methods (Video and GPS) (Folegatti et al. 2007). The skidder in the roundwood system was required to move the delimbing debris, which increased cycle time.

Table 3 Stand table compiled from pre-harvest and post-harvest sampling; Harvest mass is given in tons ha–1 on a green weight basis; DBH classes are presented in inches since they were collected in 1 inch classes Tablica 3. Prikaz istraživane sastojine prije i poslije sječe stabala Preharvest Prije sječe

DBH (inches) Prsni promjer u inčima

Biomass, t ha–1 – Biomasa, t ha–1 m2 ha–1

Trees, ha–1 –1

Broj stabala, ha

Temeljnica m2 ha–1

Total

Stem only

Merchantable (6.3 cm SED)

Post Harvest

Harvest, t ha–1

Prije sječe

Sječa, t ha–1 m2 ha–1

Trees, ha–1 –1

Ukupno

Debla

Tržišna vrijednost (6.3 cm SED)

Broj stabala, ha

Temeljnica, m2 ha–1

Total Ukupno

0.1

0.5

0.4

0.2

0.0

0.4

1.8

1.5

1.2

0.0

0.1

1.1

5.8

5.0

4.7

0.1

0.8

132

2.4

14.4

12.3

11.9

0.5

3.3

163

4.1

26.5

22.3

22.0

1.6

10.5

183

5.9

41.6

34.5

34.3

2.6

18.3

129

5.3

39.2

32.2

32.0

2.6

19.6

4.1

31.4

25.5

2.7

20.9

1.7

13.5

10.9

0.9

7.3

0.6

5.0

4.0

0.5

3.8

0.7

5.9

4.6

0.5

3.7

Total Ukupno

892

186

153

151

335

216

Croat. j. for. eng. 33(2012)2

Utilization, Cost, and Landowner Return from Whole-Tree Chipping Young... (211–223)

M. F. Smidt and J. McDaniel

Table 4 Machine element and cycle times in minutes and independent variables from analysis of machine video; Standard deviations are given () for the skidder and feller-buncher Tablica 4. Radni zahvati vozila Machine – Stroj

Element – Radni zahvat

Mean – Aritmetička sredina

Max – Maks.

Min – Min.

N – Broj

Load – Utovar

9.01

17.20

7.43

Delay – Kašnjenje

13.5

26.42

–

Pile – Uhrpavanje

14.19

33.94

–

Clean-up – Čišćenje

3.05

7.32

–

Total – Ukupno

28.88

78.62

7.60

Total – Ukupno

4.10 (2.00)

11.48

0.48

Travel unloaded – Vožnja praznoga vozila

1.73 (0.88)

4.97

0.25

–

Skidder

Travel loaded – Vožnja natovarenoga vozila

1.87 (0.94)

5.47

–

Skider

Load – Utoar

0.49 (0.48)

2.5

0.05

–

Bunches/cycle – Snopova / turnus

1.7 (1.1)

–

Skid Distance, m – Udaljenost privlačenja, m

196 (129)

544

–

Total – Ukupno

1.59 (1.03)

8.71

0.31

Move and Cut – Pomicanje i sječa

1.28 (0.55)

2.65

0.18

–

Move and Dump – Pomicanje i ispuštanje

0.29 (0.13)

0.68

0.09

–

Trees per cycle – Br. stabala po turnusu

8.3 (2.8)

–

Trees per bunch – Br. stabala po složaju

28 (7.1)

Loader Utovarivač

Feller-buncher Sječno vozilo

3.2 Simulations – Simulacije The results of the simulations are presented in Table 7. For the whole-tree chipping, the feller-buncher productivity in the simulation was considerably lower than the gross production estimates. Differences could be due to differences in average tree size between that estimated from the cruise versus actual. The differences could be caused by a combination of sampling error or the tree weight equation. In addition, the simulation added more time per pile by making the fellerbuncher produce the exact pile size. So some small bunches were produced to make most piles, which reduced machine efficiency. Finally we had only time study data for one of the two feller-buncher operators and the other one might have been more productive. Lower estimates for skidding productivity could be attributed to the way landing limits were imposed. If some landing space was available the skidder produced enough trees to resupply the landing and that hour was counted as completely productive. The method probably overestimated productive hours compared to the Multidat which would not record idle time. Most the difference in productivity among the scenarios was produced by the change in volume per tree Croat. j. for. eng. 33(2012)2

related to the harvesting system. Additional productivity gains were produced by removing bottlenecks from both felling and trucking. Whole-tree chipping was able to use most of the trucking resources provided. Clean chipping required fewer trucks since residue remained in the woods and roundwood systems were limited by loading and delimbing rather than trucking. Increased trucking resources (increased arrival rate) generally had more improvement from 1.2 to 1.5 than 1.5 to 1.75. While trucks were made available based on the Poisson arrival rate, on some occasions there was not enough wood on the landing to use all of the available trucks. The ratio of truck loaded to trucks available (departure:arrival) ranged from 0.43 to 0.96 with the greatest ratio for whole-tree chipping, where the greatest mass per tree was converted into product. Extra felling capacity was able to reduce harvest scheduled hours for each system and arrival rate combination. Total system costs for roundwood and whole-tree chipping were similar in spite of differences in component costs (Fig. 2). Total cost differences within the roundwood and whole-tree chipping system varied by as much as $4 gt–1. Clean chipping costs had a much larger variation of over $10 gt–1. Some studies in the region have also addressed whole-tree chipping from pine thinnings. An under-

217

M. F. Smidt and J. McDaniel

Utilization, Cost, and Landowner Return from Whole-Tree Chipping Young... (211–223)

Fig. 1 Distribution of time between loads in 10 minute intervals from video data and load sheet data Slika 1. Prikaz vremena između utovara u 10-minutnim intervalima iz videozapisa te baza podataka story removal in loblolly and longleaf pine stands was estimated to have a production rate of 60 gt pmh–1 and a harvest cost of $10.10 gt–1 even though trucking limited the chipper to 25% utilization (Mitchell and Gallagher 2007). A comparable cost from the simulation (less overhead and capital return) would have been nearly $12 gt–1. In similar conditions in loblolly pine thinnings in Alabama, productivity of both skidders (30 gt pmh–1) and feller-bunchers (15 and 23 gt pmh–1) (Klepac et al. 2011) were less than those in the simulation and the gross production study. A comparison of clean chip to roundwood system cost showed that clean chipping costs were about 35% greater than

roundwood costs and the cost of acceptable chips to the digester were within 2% for both harvesting systems (Watson et al. 1991). In the simulations with higher clean chipping productivity, the difference was almost the same here (38%). Stumpage and delivered product prices were south wide average prices from 4th quarter 2011 TimberMartSouth (TimberMart-South 2012). Prices used were FOB at landing of $17.05 gt–1 for pine whole-tree chips, $31.00 for clean chips, and $20.31 for roundwood pulpwood. The average stumpage value for pine pulpwood was $9.02 gt–1 for the same time period. From a landowner perspective revenue per unit area would be a critical comparison and controls for the differen ces in harvest volume. Residual value per hectare (value FOB – Harvesting cost to roadside) ranged from $287 to $564 ha–1 for roundwood. Whole-tree chipping resulted in negative residual value when arrival rate was 1.2. The highest revenue was $246 ha–1 at the low end of the roundwood system. Clean chipping was the most highly variable and ranged from $40 to $656 ha–1. The second feller-buncher was critical for lowering system costs and increasing residual value. The arrival rate had a smaller impact. In order to compete with roundwood pulpwood value (the highest residual value for roundwood) prices for whole-tree chips would have to be similar to pulpwood value at about $22 gt–1 (Fig. 3). Clean chipping was competitive with roundwood when the production rates were high, but had the largest deficit at lower production rates. Average whole-tree chip prices resulted in negative net revenue when a typical stumpage price was paid. To determine the potential for roundwood and clean chip harvesting we estimated net income per schedule hour using a stumpage fee equivalent to $5 gt–1 for roundwood and the high value the whole-tree chips ($19.21 gt–1) (Fig. 4). A lower stumpage fee might be justified since the early harvest may provide a service

Table 5 Model statistics for selected regression models; All times are delay free; Skidder – Roundwood data was from a previous study Tablica 5. Statistička obrada podataka (podaci o skideru s oblovinom iz prijašnjega su istraživanja) F value

P value

F vrijednost

P vrijednost

0.20

273.73

< 0.0001

0.02

0.29

0.7494

0.93

6.42

0.0028

0.420

2.42

27.13

0.0001

0.784

110

0.76

194.3

0.001

DF model

MSE

FB, Move and Cut – Sječno vozilo, micanje i sječa

0.890

FB, Move and Dump – Sječno vozilo, micanje i odlaganje

0.009

FB, Total Cycle – Sječno vozilo, ukupni turnus

0.161

Skidder, chipping – Skider, iveranje Skidder, roundwood – Skider, oblovina

Model – Model

218

Croat. j. for. eng. 33(2012)2

Utilization, Cost, and Landowner Return from Whole-Tree Chipping Young... (211–223)

M. F. Smidt and J. McDaniel

Table 6 Parameter estimates for delay free cycle time regression models for the feller buncher and the skidders from the whole-tree chipping operation and a roundwood pulpwood operation; P values of the T-test, showing that the estimate is not different from 0, are indicated by <0.1a, <0.05b, and <0.01c Tablica 6. Regresijski modeli za turnuse bez vremena kašnjenja B0 (Intercept) Model – Model

Estimate – Procjena

Felling – Obaranje

Name – Naziv

Distance

Skidding – Roundwood

Distance

1.376c

Privlačenje – Oblovina

to the landowner and improve stand return in spite of the reduced thinning revenue. Even with more favorable terms only the most productive whole-tree chipping scenarios resulted in a positive income per SMH.

Load number

0.050 0.571c

Br.utovara Load number

0.014

Udaljenost

Estimate – Procjena

Oblovina

0.007c

Udaljenost

Name – Naziv Hardwood stems

0.148c

Ukupno debala

1.687c

Privlačenje – Iveranje

Estimate – Procjena

Total stems

0.343

Skidding – Chipping

0.516c

Br. utovara

Clean chipping had the greatest potential for earnings per SMH but also included the largest losses. As expected most of the roundwood scenarios had positive net revenue per SMH. Among the scenarios with loss-

Table 7 Simulation productivity for 20 hectare harvest with changes in feller-buncher number (N) and truck availability (Arrival rate) and the trucks loaded (Departure rate) Tablica 7. Simulacije proizvodnosti System

Sustav

Broj

Productivity, gt pmh–1 – Produktivnost, gt pmh–1 SMH

Arrival rate

Departure rate

Iverač

Stopa dostupnosti

Stopa izvršenosti

Feller-buncher

Skidder

Loader

Delimber

Chipper

Sječno vozilo

Skider

Utovarivač

Rezno vozilo

–

1.20

0.82

–

1.50

0.97

Roundwood

–

1.75

1.00

Oblo drvo

–

1.20

0.88

–

1.50

0.99

–

1.75

1.03

–

1.20

1.03

–

1.50

1.18

–

1.75

1.17

–

1.20

1.08

–

1.50

1.34

–

1.75

1.42

–

1.20

0.75

–

1.50

0.76

Clean chipping

–

1.75

0.75

Čisto iveranje

–

1.20

0.98

–

1.50

1.19

–

1.75

1.30

Whole-tree chipping Iveranje cijelih stabala

Croat. j. for. eng. 33(2012)2

219

M. F. Smidt and J. McDaniel

Utilization, Cost, and Landowner Return from Whole-Tree Chipping Young... (211–223)

Fig. 2 Production cost from simulation of roundwood (RW), whole-tree chipping (WT), and clean chipping (CC) systems at different truck arrival rates (1.2; 1.5, and 1.75) and feller buncher number (1 or 2) Slika 2. Troškovi proizvodnje iverja

Fig. 3 Product value (FOB) for each system and the break even value to yield the same residual value per hectare compared to the highest value roundwood harvest, Values are from simulation of roundwood (RW), whole-tree chipping (WT), and clean chipping (CC) systems at different truck arrival rates (1.2; 1.5 and 1.75) and feller-buncher number (1 or 2) Slika 3. Tržišna vrijednost svakoga sustava i točke pokrića

220

Croat. j. for. eng. 33(2012)2

Utilization, Cost, and Landowner Return from Whole-Tree Chipping Young... (211–223)

M. F. Smidt and J. McDaniel

Fig. 4 System net revenue with a stumpage value of $5 per gt and a high value for whole-tree chips ($19.21 per gt), Values are from simulation of roundwood (RW), whole-tree chipping (WT), and clean chipping (CC) systems at different truck arrival rates (1.2; 1.5 and 1.75) and feller-buncher number (1 or 2) Slika 4. Neto prihodi sustava es, the roundwood and whole-tree chipping scenarios ($ smh–1) had losses similar in magnitude to estimated capital return ($20 to $38 smh–1). Contractors that chose to forgo capital return could still pay all their expenses especially if they use the firm to employ themselves.

4. Conclusion – Zaključak Gross time study of a whole-tree chipping system showed that the system productivity was severely limited by truck or market availability. Both detailed timing and gross time study data showed that with 2 feller-bunchers active the skidder would be limiting and would limit overall production to just over about 67 gt pmh-1. Even at low levels of available markets, truck availability is seldom regular and the loss of some productive hours for the harvesting system is likely. If trucking is considered simply as variable cost there is no burden adding trucking capacity to reduce gaps without trucks at the landing. The cost may be transferred to contract haulers as they may wait at the landing to be loaded. An expansion of contractor trucking capacity means that each truck added will lower the utilization of the fleet and lead to higher fixed costs per unit of trucking. Analysis is needed to Croat. j. for. eng. 33(2012)2

rationalize to optimal utilization of both the trucking fleet and the in-woods capacity. It appears unlikely that the whole-tree chipping system described here could compete with roundwood or clean chip harvesting for pulpwood until tree size makes it impractical to load roundwood. Rising prices for energy wood would not reduce the deficiency since those price increases would be reflected in pulpwood prices as well. Limited markets and constrained unloading capacity at mills are also likely to lower the profitability even if prices rise. Many Southern US contractors including the one in the study operate equipment that is nearly completely depreciated so net income can be produced even at low levels of market availability. Large expansion in woody bioenergy production will require a large expansion in demand and large jump in price since investment in in-woods systems and trucking capacity will be required. Prices now often require contractors to forgo capital return. While the practice makes harvesting feasible, it is unlikely to attract the capital investment needed to expand the industry. Since the opportunity for net revenue was so small, especially with limited truck availability or quota, there would be little rationale to select chipping over roundwood harvest.

221

M. F. Smidt and J. McDaniel

Utilization, Cost, and Landowner Return from Whole-Tree Chipping Young... (211–223)

5. References – Literatura Baker, S. A., Westbrook, M. D., Jr., Greene, W. D., 2010: Evaluation of integrated harvesting systems in pine stands of the southern United States. Biomass and Bioenergy 34(5): 720– 727.

west, In: 10th Annual Council on Forest Engineering Meeting, Syracuse, NY 1987, 79–97 p. Mitchell, D., Gallagher, T., 2007: Chipping whole trees for fuel chips: A production study, Southern Journal of Applied Forestry 31(4): 176–180.

Bolding, M. C., Lanford, B. L., 2005: Wildfire fuel harvesting and resultant biomass utilization using a cut-to-length/small chipper system. Forest Products Journal 55(12): 181–189.

Mooney, S. T., Boston, K. D., Greene, W. D., 2000: Production and costs of the Chambers Delimbinator in first thinning of pine plantations. Forest Products Journal 50(4): 81–84.

Brinker, R. W., Kinard, J., Rummer, R., Lanford, B. L., 2002: Machine rates for selected forest harvesting machines. Alabama Agricultural Experiment Station. Circular 296, 22 p.

Puttock, G. D., 1995: Estimating cost for integrated harvesting and related forest management activities. Biomass and Bioenergy 8(2): 73–79.

Caterpillar, Inc., 1996: Caterpillar Performance Handbook. Caterpillar, Inc., Peoria, IL.

Raymond, K. A., Franklin, G. S., 1990: Chain flail delimberdebarkers in eastern Canada: A preliminary assessment. Forest Engineering Research Institute of Canada. TN-153, 8 p.

Clark, A., Saucier, J. R., 1990: Tables for estimating total-tree weights, stem weights, and volumes of planted and natural southern pine in the southeast, GFRP-79, Georgia Forestry Commission. Folegatti, B. S., Smidt, M. F., Loewenstein, E. F., Carter, E., McDonald, T. P., 2007: Analysis of mechanical thinning productivity and cost for use at the wildland urban interface. Forest Products Journal 57(11): 33–38. Franklin, G. S., 1992: Model 23 Flail Chipharvester delimberdebarker-chipper: Productivity and chip quality in hardwood. Forest Engineering Research Institute of Canada, TN-187, 2 p. Greene, W. D., Carruth, J.S. 1994: Log separation economics on in-woods chipping operations. Forest Products Journal 44(10): 68–72. Greene, W. D., Mayo, J. H., DeHoop, C. F., Egan, A. F., 2004: Causes and costs of unused logging production capacity in the southern United States and Maine. Forest Products Journal 54(5): 29–37. Han, H. S., Lee, H. W., Johnson, L. R., 2004: Economic feasibility of an integrated harvesting system for small-diameter trees in southwest Idaho. Forest Products Journal 54(2): 21–27. Hardy, C. C., 1996: Guidelines for estimating volume, biomass, and smoke production from piled slash. USDA Forest Service Pacific Northwest Research Station, Portland, OR. GTR-364, 17 p. Klepac, J., Rummer, B.,Thompson, J., 2011: Harvesting small trees for bio-energy, In: Council on Forest Engineering Annual Meeting, Quebec City, CA 2011, 11 p. Lambert, M. B., 1987: Harvesting and processing biomass of small diameter stands for multiple products in the north-

222

Shrestha, S. P., Lanford, B. L., 2002: Comparison of timber utilization between a tree-length and an in-wood chipping harvesting operations, In: Council on Forest Engineering Conference, Auburn, AL 2002, 5 p. Smidt, M. F., Tufts, R. A., Gallagher, T. V., 2009: Cost of Fiber: Final Report to Wood Supply Research Institute. South, D. B., 2003: »Correct« planting density for loblolly depands on your objectives and who you ask. Forest Landowner Manual 34: 46–51. Spinelli, R., Hartsough, B.R., Moore, P.W., 2008: Recovering sawlogs from pulpwood-size plantation cottonwood. Forest Products Journal 58(4): 80–84. Stiff, C. T., Stansfield, W. F., 2003: Thinning guidelines for loblolly pine plantations in eastern Texas based on alternative management criteria. http://www.forsightresources. com/library/Stiff_Stansfield2003.pdf. (Access 15 June 2012) Stokes, B. J., Watson, W. F., 1991: Wood recovery with inwoods flailing and chipping. Tappi Journal 74(9): 109–112. TimberMart-South, 2012: South-wide quarterly summary report, 4th Quarter 2011. Traugott, T. A., Dicke, S., 2006: Are my pine trees ready to thin? Mississippi State University Extension Service. Watson, B., Stokes, B., 1994: Cost and utilization of above ground biomass in thinnings. In: 17th Annual Meeting of the Council on Forest Engineering. Portland, OR 1994, 192–201 p. Watson, W. F., Stokes, B., Flanders, L. N., Straka, T. J., Dubois, M. R., Hottinger, G. J., 1991: Cost comparison at the woodyard chip pile of clean woodland chips and chips produced in the woodyard from roundwood. In: 1991 TAPPI Pulping Conference, 163–189 p.

Croat. j. for. eng. 33(2012)2

Utilization, Cost, and Landowner Return from Whole-Tree Chipping Young... (211–223)

M. F. Smidt and J. McDaniel

Sažetak

Iskoristivost, troškovi i povrat sredstava u pridobivanju iverja iz mladih sastojina teda-bora U južnom dijelu Sjedinjenih Američkih Država prorede u kulturama teda-bora za proizvodnju iverja u budućnosti vode prema proizvodnji oblovine za industriju celuloze. Troškovi su pridobivanja iverja visoki, a najvažniji su utjecaj ni čimbenici prijevoznici (dostupnost kamiona) i kretanja na tržištu. Cilj je istraživanja bio troškovno opisati proiz vodni lanac iverja od cijelih stabala teda-bora, posječenih u proredama te usporediti prihode dobivene od iveranja cijelih stabala, čistoga iveranja u sastojini i celuloznoga drva. Bruto proizvodnja i studij rada i vremena provedeni su na površini od 67 ha, u dvanaestogodišnjoj sastojini tedabora (Pinus taeda). Nakon sječe temeljnica sastojine iznosila je 12 m2ha-1 (26 m2ha-1 prije sječe), a gustoća stabala 335 po ha (892 po ha prije sječe). Ukupna je biomasa procijenjena na 195 tona svježe tvari po hektaru, a ukupno je pridobiveno 98 gt ha-1drvnoga iverja. Iverje je potom dostavljeno pogonu za proizvodnju celuloze i papira. Upotrijebljena su ova vozila: dva sječna vozila 720 Tigercat, skider 620 Tigercat, gusjenični utovarivač T250B Tigercat, stariji model kotačnoga utovarivača, iverač Precision Husky 3086. Udaljenost prijevoza kretala se od 48 do 61 km. Tržišne potrebe za iverjem iznosile su oko 1400 gt/tjedno. Bruto stope proizvodnje iznosile su: za sječno vozilo 49 gt pmh-1, za skider 61,7 gt pmh-1, za utovarivač 72,8 gt pmh-1. Simulacije pridobivanja provedene su pomoću dinamičkoga simulacijskoga programa Stella 9,03. Program simu lira svaki sat proizvodnje na temelju proizvodnih jednadžbi i raspodjele iz trenutačne studije te u nekim slučajevima iz prethodnih studija sa sličnim uvjetima. Ovdje opisan sustav iveranja cijelih stabala nije se mogao natjecati sa sustavima pridobivanja oblovine ili čistoga iveranja za potrebe industrije celuloze zbog nepraktičnosti utovara oblovine. Rastuće cijene energijskoga drva vjero jatno će dovesti i do rasta cijena drva za proizvodnju celuloze. Mnogi izvođači radova na jugu SAD-a, uključujući i one u ovom istraživanju, koriste se opremom koja je gotovo u potpunosti amortizirana, pa neto dobit može biti proiz vedena čak i pri niskim razinama tržišne dostupnosti. Veliko širenje proizvodnje bioenergije zahtijevat će veliko širenje potražnje te povećanje cijene jer će biti potrebna ulaganja u sustave pridobivanja i prijevoza. Neto prihodi bili su izrazito mali (pogotovo s ograničenom dostupnošću kamiona) pa su slaba opravdanja za odabir iveranja nasuprot pridobivanju oblovine za proizvodnju celuloze. Ključne riječi: stablovna metoda, iveranje, biomasa, sječa, ekonomija, prorede, teda-bor

Authors’ address – Adresa autorâ:

Received (Primljeno): June 25, 2012 Accepted (Prihvaćeno): August 31, 2012 Croat. j. for. eng. 33(2012)2

Assoc. Prof. Mathew F. Smidt, PhD. e-mail: smidtmf@auburn.edu John McDaniel, Research Assistant e-mail: mcdaniele5@bellsouth.net Auburn University School of Forestry and Wildlife Sciences Duncan Drive, Auburn Alabama 36849 USA

223

Original scietific paper – Izvorni znanstveni rad

Predicting and Controlling Moisture Content to Optimise Forest Biomass Logistics Mauricio Acuna, Perttu Anttila, Lauri Sikanen, Robert Prinz, Antti Asikainen Abstract – Nacrtak Wood fuel quality attributes have to be considered by logistics planners if fuel procurement from forests and energy production at the plant are considered simultaneously. The single most important quality attribute is the moisture content (MC) of chips or raw material delivered to energy plants. It affects heating value, storage properties, chipping and transport costs of the fuel. To assess the impact of forest biomass moisture content on supply chain costs, we devel oped a linear programming-based tool for optimization decision support that minimizes sup ply chain costs including harvesting, storage, chipping, and transportation of fuels. A CHP plant in Finland was used as the study case and three biomass raw materials (supply chains) were used for the analysis: whole trees from early thinnings, stemwood from early thinnings, and logging residues from final fellings. Our results indicate that both the proportion and volume of the biomass material delivered to the plant are very sensitive to specifications on MC range limits and the length of the storage (drying) period. Compared to a scenario with no storage, a reduction in volume harvested of up to 33% can be achieved to meet the month ly energy demand if proper drying methods, such as covering of biomass material, are imple mented before chipping and delivering the biomass materials to the energy plant. Keywords: moisture content, biomass supply chain, optimal storage and transportation, linear programming, biomass covering

1. Introduction – Uvod To sustainably meet the increasing demand for forest biomass, proper harvesting technology (including comminution, handling and storage) and work methods must be developed and implemented with the goal to produce high quality fuels (Röser et al. 2010). Fuel quality is assessed based on the properties that affect the energy yield and costs. Moisture content (MC), gross calorific value and ash content of the logging residue are three properties commonly used to assess the fuel quality, as each of these properties determine the viability of biomass procurement for energy production (Gautam et al. 2012; Brand et al. 2011), as well as the usability of the plant and efficiency and economy of combustion (Röser et al. 2011). The most important single quality factor is the MC of chips. It affects the heating value, storage properties and transport costs of the fuel (Asikainen et al. 2001; Röser et al. 2011). MC is a direct cost factor, and it is taken into Croat. j. for. eng. 33(2012)2

account in the pricing of the fuel. An excessive MC results in a price reduction, while a low MC brings a bonus. The procurement of logging residue for energy production can be uneconomical due to high MC and low calorific value. High MC in biomass lowers the energy density and transportation becomes less efficient (Gautam et al. 2012). Therefore, natural drying of timber during the procurement processes should be promoted to facilitate the drying process and ensure the availability of high quality fuel in the short and long term (Röser et al. 2011). Some considerations have to be made by logistics planners if fuel procurement from forests and energy production at the plant are considered simultaneously. On one hand the extension of storage time lowers system costs by lowering transportation costs and by improving efficiency at the plant. The efficiency improvement is the result of increasing heating value and lower needle and impurity

225

M. Acuna et al.

Predicting and Controlling Moisture Content to Optimise Forest Biomass Logistics (225–238)

content in the material. On the other hand, extension of storage time increases system costs because the wood starts to decompose causing material losses and because of the capital invested in the stored fuel (Asikainen et al. 2001; Pettersson and Nordfjell 2007). Drying (MC) curves are a vital piece of information when implementing optimal storage, harvesting, chipping and transportation decisions in a biomass planning model. Although mathematical models have been developed to optimize biomass logistics for short periods (Eriksson and Bjorheden 1989; Gunnarson et al. 2004; Kanzian et al. 2009), we are not aware of any logistics model or study that investigated the effect of MC on supply chain costs for longer periods (> 1 year) through the use of an optimization model that explicitly uses MC as an input parameter. Taking previous studies as a reference point, we developed a linear programming-based decision support tool to conduct an investigation into the effect of biomass MC on supply chain costs for an operational planning under different scenarios.

2. Materials and methods – Materijal i metode 2.1 Supply chains used for the analysis – Lanci dobave The study is based on the parameters of a large scale combined heat and power plant (CHP) in Joensuu, Finland, and the described supply chains apply to the Joensuu facility. This plant has a total CHP output of 180 MW, out of which 30 MW is electrical output. The produced heat is distributed to the city through a district heating (DH) system. In the DH area of Joensuu, 41 640 inhabitants (or 57% of the municipality population) are connected to the 200 km pipeline DH network (Energiateollisuus 2011). In our study, a number of forest energy supply chains were selected and included in the optimization logistics model. The biomass materials selected and their corresponding supply chains were adapted and simplified to meet the objectives of the investigation using the Joensuu CHP plant study case. Supply chain I: Whole tree supply chain from thinnings with chipping at the roadside. In Finland, logging systems in small diameter whole tree thinning can be classified into those based on the motor-manual or mechanized cutting of trees. Supply chain I is based on the traditional two machine system with a harvester and a forwarder, where both machines utilize the multi-tree processing technique. Since the fell-

226

ing phase is still the most demanding and costly part of the production chain in energy thinning, the cutting of whole trees can be done with purpose built accumulating felling heads or by normal harvester heads equipped with multi-tree handling accessories (Laitila 2012). After felling, thinning trees are forwarded to roadside where they are left (stored) to dry for a few months. In the last step, comminution of whole trees is done at the roadside using heavy truck-mounted chippers or crushers in large-scale operations. To avoid delays caused by the interaction of machines, the truck mounted chipper and chip truck can be replaced by a single chipper truck. That blows the chips directly into a container and then hauls the load to the plant. As the chipper truck is equipped with its own chipping device and crane, load capacity suffers and the operation radius around the plant is reduced (Hakkila 2004). Supply chain II: Delimbed stems from thinnings with chipping at the plant. The harvesting units are equipped with delimbing knives and feed rollers that are suitable for multiple-tree processing (Laitila 2012). The costs of multi-stem harvesting to produce delimbed shortwood are, on average, 23% greater than harvesting whole trees due to the difference in productivity (Heikkilä et al. 2005; Laitila et al. 2010; Laitila and Väätäinen 2011). In this supply chain, logging residues (branches, needles, etc.) are left in the stand, and only the delimbed stems are forwarded to roadside for storage. Wood stems are dried for a few months at the roadside after which they are transported to the plant for comminution. This is performed using highly efficient chippers or crushers which makes road transportation and chipping operations independent of each other and very cost efficient, especially with short transportation distances. This provides some advantages such as increased technical and operative availability of the equipment, an independent and less labor intensive procurement process and the improvement of fuel quality (Hakkila 2004). Supply chain III. Logging residues with chipping at roadside. In Finland, logging residues are obtained primarily during final fellings conducted each year from November to March (winter months). In this case, the procurement of logging residues is based on the loose residue system. The harvesting method is adapted so that logging residues are left (stored) in the stand during the timber processing. After drying, the logging residues are forwarded to roadside, normally by a forwarder with an enlarged load space and a residue grapple. In a next step, the logging residues are comminuted at the roadside landing close to the logging site in a similar fashion to supply chain I. Croat. j. for. eng. 33(2012)2

Predicting and Controlling Moisture Content to Optimise Forest Biomass Logistics (225–238)

M. Acuna et al.

Fig. 1 BIOPLAN 1.1 user’s interface Slika 1. Računalni program BIOPLAN 1.1 – sučelje

2.2 Biomass supply chain optimization system Optimizacija sustava dobave biomase 2.2.1 Implementation of the system – Primjena sustava We developed and implemented a linear programming, MS Excel©-based tool for optimization decision support named BIOPLAN to investigate the effect of MC on storage, chipping and transportation costs of biomass material delivered to an energy plant (Fig. 1). The objective function of the strategic, non-spatial optimization model minimizes the total costs associated with harvesting, storage, chipping and transport according to MC curves, which are included as explicit parameters in the optimization model. In its current version (1.1), BIOPLAN includes whole trees from thinning, delimbed stems from thinning, and logging residues as biomass materials along with their corresponding supply chains. BIOPLAN is structured in such a way that all the information is displayed in a series of matrices which include: Croat. j. for. eng. 33(2012)2

Þ Decision variables on tons and corresponding solid volume of biomass material to harvest in each period, Þ Loose volume of wood chips produced in each period (roadside or energy plant), Þ Number of truck loads delivered to the energy plant, Þ Energy content of chips, Þ Harvesting and forwarding costs (including subsidies for whole trees and stem wood from thinning operations), chipping costs (roadside or energy plant), storage cost of material at the roadside, and transportation cost (stemwood or chips). The model considers a 2-year planning horizon and decisions on tons or volume of biomass material to harvest are made on a monthly basis (24 periods) over that time horizon. Except in supply chain II, the biomass material is stored for a number of periods and then chipped at the roadside. Chips with a determined

227

M. Acuna et al.

Predicting and Controlling Moisture Content to Optimise Forest Biomass Logistics (225–238)

Table 1 Sets, parameters, and variables used in the mathematical formulation of the model Tablica 1. Grupe, parametri i varijable primijenjeni u matematičkom formuliranju modela Term – Termin

Definition – Definicija

Set – Grupa

i,j = periods

i  I = {1...24}, j  J = {13...24} Parameters – Parametri Conversion factors from m3 solid to m3 loose for whole trees, stem wood and logging residues, respectively

a, b, g

Faktori pretvorbe iz rasute u čvrstu drvnu tvar dobivenu od cijelih stabala, debla i drvnoga ostataka Energy content for chips produced from whole trees, stem wood and logging residues, respectively

ECwt, ECsw, EClr

Energija iz iverja cijelih stabala, debla i drvnoga ostatka Harvesting cost (€/m3 solid) for whole trees, stem wood, and logging residues harvested in period i

HC1wt, HC1sw, HC1lr

Troškovi sječe (€/m3 čvrste tvari) za cijela stabla, debla i šumski ostatak u godini i

ST1,jwt, ST1,jsw, ST1,jlr

Storage cost (€/m3 solid) for whole trees, stem wood, and logging residues stored at roadside from period i to j (i ≤ j) Troškovi skladištenja uz šumsku cestu (€/m3 čvrste tvari) za cijela stabla, debla i šumski ostatak u vremenu od i do j (i ≤ j) Chipping cost (€/m3 solid) for whole trees, stem wood, and logging residues harvested in period i and chipped in period j at roadside or plant

wt 1,j

sw 1,j

lr 1,j

CH , CH , CH

TR1,jwt, TR1,jsw, TR1,jlr

Troškovi iveranja (€/m3 čvrste tvari) cijelih stabala, debala i drvnoga ostatka, sječenih u vremenu i, iveranih uz šumsku cestu ili u energani, u vremenu j Transportation cost (€/m3) for whole trees chips (solid volume), stem wood (solid volume), and logging residues chips (loose volume) harvested in period i and transported to plant in period j Troškovi prijevoza iverja (€/m3) cijelih stabala (čvrsta tvar), debala (čvrsta tvar) i drvnoga ostatka (rasuta tvar) posječenih u vremenu i te prevezenih u energanu u vremenu j Variables – Varijable

xi,j yi,j zi,j xi,jr zi,jr

Solid volume of whole trees harvested in period i and stored at roadside until period j for chipping Obujam čvrste tvari iz cijelih stabala posječenih u vremenu i te skladištenih uz šumsku cestu do vremena iveranja j Solid volume of stem wood trees harvested in period i and stored at roadside until period j for transport and chipping at the plant Obujam čvrste tvari iz deblovine pridobivene u vremenu i te skladištenih uz šumsku cestu do vremena j uz iveranje u energani Solid volume of logging residues harvested in period i and stored at roadside until period j for chipping Obujam čvrste tvari iz drvnoga ostatka pridobivenoga u vremenu i te uskladištenoga uz šumsku cestu do vremena iveranja j xi,j × a = Loose volume of chips from whole trees harvested in period i and stored at roadside until period j for chipping xi,j × a = Obujam rasutoga iverja dobivenoga iz cijelih stabala posječenih u vremenu i te uskladištenih uz šumsku cestu do vremena iveranja j zi,j × g = Loose volume of chips from logging residues harvested in period i and stored at roadside until period j for chipping zi,j × g = Obujam rasutoga iverja dobivenoga iz drvnoga ostatka u vremenu i te uskladištenoga uz šumsku cestu do vremena iveranja j

moisture and energy content are then transported to the energy plant for consumption. In supply chain II, stemwood is stored at the roadside for a number of periods and then transported to the energy plant for chipping and consumption. Storage of any biomass material at the roadside is allowed for a period of up to 24 months (from January Year 1 to December Year

228

2) and all the material supplied must meet the plant monthly demand for energy (MWh) in the year 2 (Energy Generation Year) at minimum cost. That means that any biomass produced in the year 1 will be combusted in the year 2. In its basic formulation, the supply chain model can be expressed as follows. Sets, parameters, and variables are presented in Table 1: Croat. j. for. eng. 33(2012)2

Predicting and Controlling Moisture Content to Optimise Forest Biomass Logistics (225–238)

Objective function (FO) Equation 1 minimizes the total supply chain costs (€), associated with biomass harvesting, storage, chipping and transport.

M. Acuna et al.

ing at the plant. Within each supply chain, the system reports the cost per m3 solid, cost per green ton, cost per m3 loose, cost per truck load and cost per MWh. 2.2.2 Optimization model parameters – Parametri optimizacijskoga modela Table 2 shows the parameters used in the control scenario for the three supply chains analysed.

(1)

Constraints Equation 2 ensures that the energy content of the chips supplied satisfy the monthly energy demand at the plant (MWh).

(2) Equation 3 ensures that an even volume of whole trees is harvested evenly in each year. This allows for continuous work for the harvesting and haulage contractors. (3) Equation 4 ensures that an even volume of stem wood is harvested evenly each year. This allows for continuous work for the harvesting and haulage contractors. (4) Equation 5 ensures that an even volume of logging residues is harvested evenly during the winter months of each year. (5) The model assumes that in any period the chips arriving at the energy plant must be consumed in the same period, and therefore, there are no costs associated with the storage of chips at the plant. Other constraints modeled in the system include MC limits for chips and stemwood arriving at the plant, and drying period limits for the materials that are stored at the roadside. Stumpage price is not included as part of the total supply chains costs. As outputs, BIOPLAN reports total cost for the whole operation and total cost by activity (harvesting, storage, chipping, and transportation), as well as total energy of the fuel supplied to the plant in MWh. Additionally, for each supply chain, the system reports solid volume and fresh tons of biomass material harvested, loose volume of chips produced at the roadside or plant, total energy content (MWh), and total number of truck loads with chips or stem wood arrivCroat. j. for. eng. 33(2012)2

Several references were consulted to get the parameters for BIOPLAN. Basic density, calorific values, and solid content for the main species in Finland were obtained from Nurmi (1993); Hakkila (2004); and Laurila and Lauhanen (2012). Harvesting, chipping, and transport costs are presented in Table 3. These costs were preliminarily calculated and validated from costing spreadsheets developed by METLA (Heikkila et al. 2005; Laitila 2006; Laitila 2008). These values were subsequently updated by using results from more recent studies (Laitila et al. 2010; Laitila and Väätäinen Table 2 Parameters and conversion factors used in BIOPLAN Tablica 2. Paramateri i faktori pretvorbe primijenjeni u BIOPLAN-u Parameters & conversion factors

SCH

Parametri i faktori pretvorbe

III

19.5

19.0

20.0

410

400

415

172.2

168.0

174.3

0.42

2.38

35.0

Energy content at 0% MC, MJ/kg Količina energija pri 0 % udjela vlage, MJ/kg Basic density, kg/solid m3 Osnovna gustoća, kg/čvrste tvari, m3 Bulk density, kg/solid m3 Obujamna gustoća, kg/čvrste tvari, m3 Solid content Udio čvrste tvari Ratio loose – m3 to solid m3 Odnos (m3) rasute prema čvrstoj tvari Truck payload, tons Nosivost kamiona, t Truck volume, m3 Obujam tovarnoga prostora, m3 Round trip distance, km Puni obilazak, km Dry matter loss rate, %/month Stopa suhe tvari, % /mjesečno Interest rate, %/month Kamatna stopa, % /mjesečno

130.0* 47.0** 130.0* 128

150

185

1.0

2.0

0.5

* m3 loose, ** m3 solid * m3 rasuta tvar, ** m3 čvrsta tvar

229

M. Acuna et al.

Predicting and Controlling Moisture Content to Optimise Forest Biomass Logistics (225–238)

Table 3 Harvesting, chipping, transport cost parameters used in BIOPLAN Tablica 3. Parametri troškova sječe, iveranja i prijevoza primijenjeni u BIOPLAN-u Parameters – Parametri Mechanized felling, bunching & forwarding, €/m3 Strojna sječa i izrada te privlačenje, €/m3

SCH I

SCH II

SCH III

13.0

14.6

10.0*

5.8

3.8**

7.4

Chipping, €/m3 – Iveranje, €/m3 MC% <=35 – Udio vlage, <=35 % 36=MC%<=50 – Udio vlage, 36 – ≤ 50 %

5.1

3.1**

6.7

MC%>50 – Udio vlage, >= 50 %

4.3

2.3**

5.9

Transport, €/km – Prijevoz, €/km

2.4

1.8

2.5

* It includes just forwarding – Uključuje privlačenje forvarderom ** Chipping at the energy plant – Iveranje u energani

2011; Tahvanainen and Anttila 2011). Storage costs in the model are based on the assumption that there have been costs associated with harvesting and transporting the material to roadside and that these costs have been paid for at the time of harvest. Thus, storage costs are then the interest charge on the harvesting and transport to roadside costs since the wood owner incurs a delay due to storage in being reimbursed for these. An annual interest rate of 6% was used for the analysis, as it has been suggested and used in previous biomass studies in Finland (Laitila 2006).

2.2.3 Drying curves and algorithm – Krivulje i algoritam MC of biomass materials for any given month was calculated with the drying models developed by Sikanen (2012) based on heuristic fitting. The models predict daily moisture change for seven drying periods of a year based on previous empirical experiments and historical data. For whole trees and delimbed stemwood, values by Hakkila (1962) were used. As no data on the annual variation in MC for crown biomass from final fellings were found, Hakkila’s (1962) values for small-sized Norway Spruce were used, but increased by 2% units. For the purpose of the analysis, and due to the lack of predictive MC models for each biomass material, drying curves were assumed to be the same for whole trees, stem wood, and logging residues (Fig. 2). In addition, dry matter loss rates per month due to storage were assumed to be 1% for whole trees and stem wood, and 2% for logging residues (Laitila 2006).

2.3 Data analysis – Obrada podataka The analysis in this paper focuses primarily on the effect of MC on supply chain costs. Three aspects of the biomass supply chain and their impact on total costs and the optimal fuel supply solution were investigated with the BIOPLAN system: Þ Effect of different MC limits on the material arriving at the CHP plant,

Fig. 2 Drying curves for biomass materials at different felling times Slika 2. Krivulje udjela vlage tijekom obaranja i skladištenja

230

Croat. j. for. eng. 33(2012)2

Predicting and Controlling Moisture Content to Optimise Forest Biomass Logistics (225–238)

Þ Effect of drying period limits on the material stored at the roadside, Þ Effect of more efficient drying methods. The results of the analysis in each supply chain are presented in terms of the monthly volume of biomass material that is harvested, as well as the total supply chain costs and the costs of different operational activities (harvesting, storage, chipping, and transportation). Statistical differences of monthly volume harvested and costs by biomass material between the scenarios analyzed were checked with t-tests conducted at the p=0.05 level of significance. 2.3.1. Effect of different MC limits on total supply chain costs – Utjecaj graničnih sadržaja vlage na ukupne troškove lanca dobave biomase Three scenarios were set up to investigate the effect of MC limits on supply chain costs. The first (basic) scenario did not include specific constraints to limit the MC of the biomass material (chips in supply chains I and III, and stem wood in supply chain II) arriving at the CHP plant. In the second and third scenario, we added a constraint to limit the MC of the biomass material arriving at the plant, which was set to 41–49%, and 41–43%, respectively. The latter MC range is the Joensuu CHP plant preference in their logistics operations. 2.3.2. Effect of storage time on total supply chain costs – Utjecaj duljine skladištenja materijala na ukupne troškove lanca dobave biomase In BIOPLAN, the storage of any biomass material at the roadside is allowed for a period of up to 24 months (from January Year 1 to December Year 2). The optimization model determines the monthly volume that is harvested and stored at the roadside so that the supply chain costs are minimized during that period. To investigate the effect of biomass storage time at the roadside on total supply chain costs, a constraint was added to limit the number of periods (months) of storage. There was particular interest in knowing if storage for a period shorter than 12 months had a substantial effect on the biomass supply chain costs. Therefore, in addition to the unconstrained scenario (no limits on storage), two scenarios were established which constrained the storage period to a maximum of 6 periods, and a minimum of 3 periods, respectively. 2.3.3 Effect of more efficient drying methods on supply chain costs – Utjecaj djelotvornijih načina sušenja biomase na ukupne troškove lanca dobave Effective drying of biomass material, including covering, is essential to decrease the MC of biomass Croat. j. for. eng. 33(2012)2

M. Acuna et al.

material. MC can be reduced in a few months by 10–20% using only solar and wind power if biomass material is covered (Röser et al. 2010) To analyze the effect of covering on biomass supply chain costs, and in the absence of accurate drying curves for covered material, we established two hypothetical scenarios where the drying rates for all the biomass materials were adjusted by 10% and 15% in the autumn and winter months (October to April), assuming a lower rate of rewetting of the biomass materials when they are covered during these months. Due to the lack of information, covering costs were assumed to be € 1/m3.

3. Results and Discussion – Rezultati s raspravom 3.1 Effect of different MC limits on supply chain costs – Utjecaj graničnih sadržaja vlage na ukupne troškove lanca dobave biomase For the three scenarios analyzed (unconstrained, 41–49% MC, 41–43% MC), Fig. 3 shows the MC (%) curves of biomass material (chips or stem wood) delivered to the plant during year two. In the unconstrained scenario, the biomass materials present a much higher variation in MC throughout the year, especially for whole trees and logging residues. Maximum and minimum MC values were 52.3% and 41.5% for whole trees, 53.7% and 39.6% for logging residues, and 51.2% and 42.6% for stem wood. As expected, the biomass material delivered to the plant during the summer had a much lower MC than in autumn and winter. During the winter months (November to March), logging residues had a lower MC (average 48.6% MC) compared to whole trees and stem wood (average 50.3% MC), which resulted from much longer average drying periods (> 1 year) at the roadside for logging residues (7 months compared to less than 1.5 months for whole trees and stem wood). Variation in post-storage MC for logging residues was much higher than for stem wood and whole trees (std. dev. 5.4 versus 2.8 for stem wood and 3.4 for whole trees) in the unconstrained scenario. Conversely, there was no variation for any of the biomass materials produced (whole trees and logging residues) (std. dev. = 0) when MC was constrained to 41–43%, as the MC of these materials was always at the top of the range (43%). Fig. 4 shows the volume harvested per year by biomass material for the three scenarios analyzed. Only the mean volume differences between whole trees and logging residues harvested in the year 1 and 2 were statistically significant (p<0.05). In the unconstrained

231

M. Acuna et al.

Predicting and Controlling Moisture Content to Optimise Forest Biomass Logistics (225–238)

Fig. 3 Variation in post-storage MC for whole trees, stem wood, and logging residues used in the second year for energy generation in the three MC scenarios analyzed Slika 3. Promjene u sadržaju vlage kod cijelih stabala, debla i šumskoga ostatka pri dužem skladištenju (korištenje tek druge godine)

Fig. 4 Solid volume (m3) harvested for whole trees, stem wood, and logging residues in the three MC scenarios analyzed Slika 4. Obujam čvrste tvari za cijela stabla, debla i drvni ostatak u tri različita scenarija scenario, logging residues were mostly produced in the year 1 (47 118 m3 or 100% of the total biomass material harvested in that year) and stored at the roadside for longer periods (average of 7 months) before being

232

delivered to the plant in the year 2. This is the result of lower harvesting costs for logging residues, which allows for longer storage periods as the interest on the harvesting cost (storage cost) is reduced in comparison Croat. j. for. eng. 33(2012)2

Predicting and Controlling Moisture Content to Optimise Forest Biomass Logistics (225–238)

to whole trees and stem wood. Conversely, due to the cost structure and lower energy content of whole trees (and to a lesser extent stem wood), the optimal solution only allows for short periods of storage at the roadside, with chipping and transportation occurring just a few periods after the harvest and storage. These supply chains were used primarily in the year 2, when approximately 41% of the material harvested (94 770 m3) corresponded to whole trees and 57% of the material harvested (130 979 m3) corresponded to stem wood. When MC was constrained to 41–43%, the solution precluded the harvest of stem wood and the energy demand was mostly met with logging residues harvested during the first year and whole trees harvested during the second year. Similarly, when MC was constrained to 41–49%, the volume harvested of stem wood was just marginal and accounted for less than 1% of the total volume harvested in the year 1, and about 3% of the total volume harvested in the year 2. The volume of logging residues harvested in the year 1 was 43 170 m3 (87% of the total volume harvested) when the MC ranged between 41–49% and 152 161 m3 (92% of the total volume harvested) when the MC ranged between 41–43%. Whole trees was the predominant biomass material harvested in the year 2 with 151 161 m3 (69% of the total volume harvested) when the MC ranged between 41–49% and 99 209 m3 (88.2% of the total volume harvested) when the MC ranged between 41–43%. The predominance of log-

M. Acuna et al.

ging residues and whole trees is due to their lower harvesting costs (which is the main cost component in the three supply chains) in comparison to whole trees. This reduction in harvesting costs offsets the higher chipping and transportation costs of these two biomass materials. In addition, the reduction in harvesting costs makes storage more cost effective for logging residues and whole trees, allowing for longer storage periods of these biomass materials (average of 1.5 months for whole trees and 7 months for logging residues, versus average of 1 month for stem wood). Despite the effect of MC limits on the volume and distribution of the biomass harvested, no statistical differences were obtained in terms of supply chain costs (Fig. 5) in the three scenarios analysed (p < 0.05). In part, this is explained by the fact that the harvesting and extraction of biomass materials to roadside are the major supply chain cost component, which however, are not affected by moisture content. Also, in constrained scenarios, the solution tends to adjust the proportion of the materials that minimizes the supply chain costs, based on the contribution of their operational activities (harvesting, storage, chipping, and transportation) to the objective function. On a per m3 solid basis, the total supply chain costs in the unconstrained scenario were a bit lower than those in the 41–49% MC, and the 41–43% MC constrained scenarios. This was basically due to the balanced proportion of stem wood in relation to whole trees and stem logging

Fig. 5 Supply chain costs per solid m3 (left) and MWh (right) by operational activity in the three MC scenarios analysed Slika 5. Troškovi lanaca dobave u tri različita scenarija Croat. j. for. eng. 33(2012)2

233

M. Acuna et al.

Predicting and Controlling Moisture Content to Optimise Forest Biomass Logistics (225–238)

residues, and the lower supply chain costs associated. Due to its high harvesting and storage costs, stem wood volume was marginal in the 41–49% MC scenario, and was nil when MC was constrained to a tight 41–43%. On a per MWh basis, there was no major difference in supply chain costs between the three scenarios, although chipping and storage costs were bigger in the 41–43% MC scenario. Total transportation costs were similar in the three supply chains. Despite the fact that transportation cost per km was cheaper for logging residues, the longer distance assumed for this supply chain offset the savings that resulted from a lower cost per km.

3.2 Effect of drying period limits on supply chain costs – Utjecaj vremena sušenja materijala na troškove lanca dobave biomase Fig. 6 shows the total volume harvested in the year 1 and 2 for whole trees, stem wood, and logging residues, when the drying period was limited to a maximum of 6 and a minimum of 3 months, and compared to the unconstrained scenario. In the unconstrained scenario, stem wood and whole trees were the predominant materials, accounting for approximately 47% and 34% of the total volume harvested in the year 1 and 2, respectively. It is clear that increasing the stor-

Fig. 6 Volume harvested of whole trees, stem wood, and logging residues, when the drying period is limited to a min. of 3 months (a), max. of 6 months (b), and unconstrained (c) Slika 6. Obujam posječenih stabala, debala i drvnoga ostatka kada je vrijeme sušenja materijala najmanje 3 mjeseca (a), najviše 6 mjeseci (b) i bez ograničenja (c)

234

Croat. j. for. eng. 33(2012)2

Predicting and Controlling Moisture Content to Optimise Forest Biomass Logistics (225–238)

age (drying) period to lower the MC of the biomass materials favors the production of whole trees and logging residues, since the high harvesting and storage costs of stem wood preclude the production and delivery of this material to the energy plant. For example, when storage was limited to a maximum of 6 months, whole trees and logging residues accounted for 10.2% and 89.0% of the total volume harvested, respectively. The same pattern was observed when storage was constrained to a minimum of 3 months. In this case, whole trees and logging residues accounted for 8.2% and 91.8% of the total volume harvested, respectively. In terms of the average storage period, most of the volume of stem wood is only stored for a reduced number of periods. In the unconstrained scenario, 79% of the stem wood volume is stored for less than one month, whereas when the maximum storage period is constrained to 6 months, 90% of the volume is stored for less than one month. In the case of whole trees, shorter storage periods are obtained in the unconstrained scenario and when the storage period is constrained to a maximum of 6 periods. In these two scenarios, 100% of the volume harvested of whole trees is stored for less than 5 months, and a high proportion of this material (about 75% in each scenario) is not stored at the roadside but delivered to the plant right after the harvesting. Conversely, logging residues are stored for longer periods. In the unconstrained scenario, 52% of the volume harvested of logging residues is stored for a period between 10 and 12 months. Likewise, when the drying period is constrained to a minimum of 3 months, 100% of the volume harvested of logging residues is stored for a period longer than 8 months, with a maximum and an average storage period of 22 and 12 months, respectively. The longer storage period of logging residues is explained by the lower harvesting costs, as well as by the lower storage costs, which are calculated as the interest charge on the harvesting and transport to roadside costs. Despite the differences in production that resulted from the constraints added to limit the storage period, no substantial differences in supply chain costs were observed across the scenarios. In comparison with the unconstrained scenario, the cost per m3 and per MWh only increase by less than 1% when the storage period is limited to a maximum of 6 months or to a minimum of 3 months. These results indicate that the volume distribution of the biomass materials is quite sensitive to the storage period, although no major differences are obtained in terms of the supply chain costs per m3 harvested. Croat. j. for. eng. 33(2012)2

M. Acuna et al.

3.3 Effect of more efficient drying methods on supply chain costs – Utjecaj djelotvornijih načina sušenja biomase na troškove lanca dobave The positive effect of proper piling, covering, and handling methods, to lower the MC and maintain the lower MC over the winter months has been noted in several earlier studies (Ryan 2009; Anheller 2009; Röser et al. 2011). These methods provide direct financial benefits by increasing the energy content and facilitating a more efficient chipping and grinding operation. Table 4 presents the results of simulated covering on supply chain costs. When the drying rates of the materials are adjusted by 10% and 15% (less rewetting in autumn and winter months), supply chain costs increase by about € 0.43/ m3 and € 0.81/ m3, respectively. This rise is basically explained by the addition of covering costs as well as the higher chipping costs of whole trees (the predominant biomass material in both scenarios) that result from a lower MC. To some extent, the greater cost is also explained by the harvesting costs of whole trees and stem wood. In the unconstrained scenario, the solution includes a substantial volume of logging residues, due to their lower harvesting costs in comparison with whole trees and logging residues. Conversely, when the drying rates of the biomass materials are adjusted by 10% and 15%, the solution precludes the harvest of logging residues because of their high chipping and transportation costs, and the procurement system relies entirely on whole trees and stem wood. Although the total supply chain costs per m3 solid increase when the MC is adjusted by 10% and 15%, the results of the analysis show that on a per MWh basis, these costs are offset by the higher energy content of whole trees and logging residues, and overall, the covering of biomass materials does not result in any substantial change in the total supply chain costs per MWh. The use of more efficient methods has an important effect on the volume of biomass materials that need to be harvested to meet the monthly demand of the energy plant. When compared to a scenario with no storage, a reduction in the volume harvested of up to 33% (from about 406 000 m3 to 269 000 m3) can be achieved if proper drying methods, such as covering of biomass material, are implemented before chipping and delivering the biomass materials to the energy plant.

4. Conclusions – Zaključci In this study, a linear programming-based optimal decision support tool, named BIOPLAN, has been developed and implemented to assess the effect of bio-

235

M. Acuna et al.

Predicting and Controlling Moisture Content to Optimise Forest Biomass Logistics (225–238)

Table 4 Costs per m3 and MWh by operational activity for different rewetting rate scenarios Tablica 4. Troškovi različitih radnih aktivnosti po scenarijima Unconstrained – Bez ograničenja

10% LRR*

15% LRR**

Activity – Radnje

€/m

€/MWh

€/m

€/MWh

€/m

€/MWh

Harvesting – Pridobivanje drva

13.19

7.04

13.87

7.32

13.43

6.94

Storage – Skladištenje

0.24

0.13

0.15

0.08

0.15

0.08

0.35

0.18

0.35

0.18

Covering – Prekrivanje

Chipping – Iveranje

4.02

2.15

4.11

2.17

4.79

2.48

Transport – Prijevoz

6.99

3.74

6.39

3.37

6.54

3.38

Total – Ukupno

24.44

13.06

24.87

13.13

25.25

13.06

*10% lower rewetting rate – * 10 % niži udio vlaženja **15% lower rewetting rate – ** 15 % niži udio vlaženja

mass moisture content on supply chain costs. In complex supply chains, there are a number of factors associated with the production and procurement of biomass materials to energy plants that drive the solutions when using an optimization approach. These considerations have to be taken into account by logistics planners if fuel procurement from forests and energy production at the plant are considered simultaneously. Getting a better understanding of the tradeoffs associated with the storage of biomass materials before they are delivered to energy and heating plants is key to meet the energy demand at minimum cost. These considerations include an understanding of the effect of the length of time in storage and its positive impact on transport costs and the efficiency at the plant, as well as the negative implications such as biomass material decomposition, higher chipping and storage costs associated with the capital that is bound to the storages. Results of this study indicate that in the majority of the scenarios assessed, whole trees and logging residues are the predominant delivered biomass material to the CHP plant. This is the result of the lower harvesting costs and the higher energy content associated with this biomass materials. The lower harvesting and storage costs of logging residues allows for longer storage periods as opposed to whole trees and stem wood which are mostly stored for short periods. In constrained scenarios (including a limit of the MC of the biomass materials delivered to the plant, and a limit of the storage period), the solution precludes the production of stem wood due to its high harvesting and storage costs and lower heating value in comparison to logging residues and whole trees. In regard to the volume of biomass materials required to meet the monthly demand of the energy plant, a reduction of

236

up to 33% can be achieved if covering of biomass material is implemented before chipping and delivering the biomass materials to the energy plant. Despite the limitations of our study (limited number of supply chains, same drying curves for all the biomass material, generic parameters in each supply chain), the results of the study show the importance of counting on optimal decision support tools that allows for the assessment of key fuel quality attributes and their effect on supply chain costs.

Acknowledgements – Zahvala The authors thank the following people and institutions for their support in carrying out this research project: Þ AFORA – University of the Sunshine Coast, Australia, Þ METLA, Joensuu, Finland, Þ University of Eastern Finland, Joensuu, Finland, Þ FORTUM, Joensuu, Finland.

5. References – Literatura Anheller, M., 2009: Biomass losses during short-term storage of bark and recovered wood. Department of Energy and Technology Uppsala. Swedish University of Agricultural Sciences, Faculty of Natural Resources and Agricultural Sciences, Department of Energy and Technology. Examensarbete ISSN 1654-9392. Asikainen, A., Ranta, T., Laitila, J., Hämäläinen, J., 2001: Hakkuutähdehakkeen kustannustekijät ja suurimittakaavaisen hankinnan logistiikka (Cost factors and large scale procurement of logging residue chips). In: Research notes 131. University of Joensuu, Faculty of Forestry; (In Finnish). Croat. j. for. eng. 33(2012)2

Predicting and Controlling Moisture Content to Optimise Forest Biomass Logistics (225–238) Brand, M. A., Bolzon de Muñiz, G. I., Ferreira, W., Brito, J. O., 2011: Storage as a tool to improve wood fuel quality. Biomass and Bioenergy 35(7): 2581–2588. Energiateollisuus, 2011: District Heating in Finland 2010. Energiateollisuus ry (Finnish Energy Industries). ISSN 07864809, ET-Kaukolämpökansio 7/6. 72 p. Available at: http:// www.energia.fi/sites/default/files/district_heating_in_finland_2010_web.pdf Eriksson, L. O., Bjoerheden, R., 1989: Optimal storing, transport and processing for a forest-fuel supplier. European Journal of Operational Research 43(1): 26–33. Gautam, S., Pulkki, R., Shahi, Ch., Leitch, M., 2012: Fuel quality changes in full tree logging residue during storage in roadside slash piles in Northwestern Ontario. Biomass and Bioenergy 42: 43–50. Gunnarsson, H., Rönnqvist, M., Lundgren, J., 2004: Supply chain modelling of forest fuel. European Journal of Operational Research 158(1): 103–123. Hakkila, P., 1962: Polttohakepuun kuivuminen metsässä. Summary: Forest seasoning of wood intended fot fuel chips. Communicationes Instituti Forestalis Fenniae 54.4. 82 p. Hakkila, P., 2004: Developing technology for large-scale production of forest chips. Wood Energy Technology Programme 1999-2003, Final report. Helsinki 99 p. Heikkilä, J., Laitila, J., Tanttu, V., Lindblad, J., Sirén, M., Asikainen, A., Pasanen, K., Korhonen, K. T., 2005: Karsitun energiapuun korjuuvaihtoehdot ja kustannustekijät. [Harvesting alternatives and cost factors of delimbed energy wood]. Metlan työraportteja / Working Papers of the Finnish Forest Research Institute 10. 56 p. (In Finnish). Kanzian, C., Holzleitner, F., Stampfer, K., Ashton, S., 2009: Regional energy wood logistics – Optimizing local fuel supply. Silva Fennica 43(1): 113–128. Laitila, J., 2006: Cost and sensitive analysis tools for forest energy procurement chains. Forest Studies / Metsanduslikud Uurimused 45: 5–10. ISSN 1406-9954. Laitila, J., 2008: Harvesting technology and the cost of fuel chips from early thinnings. Silva Fennica 2008, 42(2):267–283. Laitila, J., 2012: Methodology for choice of harvesting system for energy wood from early thinning. Dissertationes Forestales 143. 68 p. Available at: http://www.metla.fi/dissertationes/df143.htm

M. Acuna et al.

Laitila, J., Heikkilä, J., Anttila, P., 2010: Harvesting alternatives, accumulation and procurement cost of small-diameter thinning wood for fuel in Central-Finland. Silva Fennica 44(3): 465–480. Laitila, J., Väätäinen, K., 2011: Kokopuun ja rangan autokuljetus ja haketustuottavuus. Metsätieteen aikakausikirja 2: 107–126. Laurila, J., Lauhanen, R., 2012: Weight and volume of smallsized whole trees at different phases of the supply chain. Scandinavian Journal of Forest Research 27(1):46–55. Nurmi, J., 1993: Heating values of the above ground biomass of small-sized trees. Acta Forestalia Fennica 236. 30 p. Pettersson, M., Nordfjell, T., 2007: Fuel quality changes during seasonal storage of compacted logging residues and young trees. Biomass and Bioenergy 31(11–12): 782–792. Röser, D., Erkkilä, A., Mola-Yudego, B., Sikanen, L., Prinz, R., Heikkinen, A., Kaipainen, H., Oravainen, H., Hillebrand, K., Emer, B., Väätainen, K., 2010: Natural drying methods to promote fuel quality enhancement of small energywood stems. Metlan työraportteja / Working Papers of the Finnish Forest Research Institute 186. 60 p. ISBN 978-951-40-2279-1. Available at: http://www.metla.fi/julkaisut/workingpapers/2010/mwp186.htm. Röser, D., Mola-Yudego, B., Sikanen, L., Prinz, R., Gritten, D., Emer, B., Väätäinen, K., Erkkilä, A., 2011: Natural drying treatments during seasonal storage of wood for bioenergy in different European locations. Biomass & Bioenergy 35(10): 4238–4247. Ryan, M., 2009: Quality issues in the forest biomass supply chain. In: Canbio Annaual conference. Oct. 20-21, 2009. Edmonton, Canada. Presentation available at: http://canbio.ca/ upload/documents/edmonton09ryans.pdf Sikanen, L., Röser, R., Anttila, P., Prinz, R., 2012: Forecasting algorithm for natural drying of energy wood in forest storages. Paper to be submitted to the Journal of Forest Energy (in press) Tahvanainen, T., Anttila, P., 2011: Supply chain cost analysis of long-distance transportation of energy wood in Finland. Biomass and Bioenergy 35(8):3360–3375.

Sažetak

Predviđanje i praćenje sadržaja vlage radi poboljšanja u logistici pridobivanja biomase Kako bi se zadovoljila sve veća potražnja za šumskom biomasom, moraju se usavršiti pravilne tehnologije prido bivanja biomase (uključujući usitnjavanje, rukovanje i skladištenje biomase) radi proizvodnje biogoriva visoke kakvoće. Potrebno je bolje poznavanje skladištenja biomase prije nego što se ona dopremi na glavno stovarište ili u energanu kako bi se postigli viši energijski učinci uz manje troškove. To uključuje poznavanje razdoblja skladištenja biomase s Croat. j. for. eng. 33(2012)2

237

M. Acuna et al.

Predicting and Controlling Moisture Content to Optimise Forest Biomass Logistics (225–238)

mogućim pozitivnim utjecajem na troškove prijevoza i energijsku učinkovitost u energani, te predviđanje mogućih negativnih učinaka kao što su razlaganje drvnoga materijala i viši troškovi iveranja i skladištenja nastalih zbog ve zanoga kapitala u skladištima. Najvažnija značajka kakvoće iverja ili sirovine za proizvodnju iverja jest udio vlage (MC). Ona utječe na toplin sku vrijednost, svojstva skladištenja, troškove iveranja i prijevoza. Za procjenu utjecaja sadržaja vlage biomase na troškove lanca dobave razvijen je alat (linearno programiranje) za potporu odlučivanja koji smanjuje troškove lanca dobave, uključujući troškove pridobivanja biomase, skladištenja, iveranja i prijevoza. Pri istraživanju je korištena postojeća bioenergana u Finskoj te su proučena tri lanca dobave biomase: biomasa od cijelih stabla iz ranih proreda, deblovina iz ranih proreda i šumski ostatak iz dovršnoga sijeka. Najveći udio biomase bio je iz cijelih stabala te iz drvnoga ostatka (zbog nižih troškova pridobivanja i više energijske vrijednosti). Niži troškovi pridobivanja i skladištenja drvnoga ostatka omogućuju dulje razdoblje skladištenja, za razliku od skladištenja cijelih stabala i deblovine koji se uglavnom skladište u kraćim razdobljima. Prilikom postavljenih ograničenja u pridobivanju biomase (uključujući granične vrijednosti udjela vlage te dopuštenoga vremena skladištenja) isključuje se proizvodnja biomase iz deblovine zbog troškova pridobivanja i skladištenja te niže energijske vrijednosti. Moguće je smanjiti mjesečnu potražnju en ergane za biomasom za 33 % ako bi se materijal pokrivao prije samoga iveranja i prijevoza. Unatoč ograničenjima u istraživanju (ograničen broj lanaca dobave, iste krivulje sušenja biomase, generički parametri u svakom lancu dobave) rezultati istraživanja pokazuju važnost pravoga alata za podršku odlučivanju koji omogućuje procjenu ključnih značajki te njihov utjecaj na troškove lanca dobave biomase. Ključne riječi: udio vlage, lanac dobave biomase, razdoblje skladištenja i prijevoza, linearno programiranje, prekrivanje biomase

Authors’ address – Adresa autorâ:

Received (Primljeno): July 20, 2012 Accepted (Prihvaćeno): August 19, 2012

238

Mauricio Acuna, PhD. e-mail: macuna@usc.edu.au AFORA – University of Sunshine Coast Private bag 12 Hobart, TAS AUSTRALIA Perttu Anttila, PhD. e-mail: perttu.anttila@metla.fi Robert Prinz, MSc. e-mail: robert.prinz@metla.fi Prof. Antti Asikainen, PhD. e-mail: antti.asikainen@metla.fi Finnish Forest Research Institute P.O. Box 68 FI-80101, Joensuu FINLAND Prof. Lauri Sikanen, PhD. e-mail: lauri.sikanen@uef.fi University of Eastern Finland P.O. Box 111 FI-80101, Joensuu FINLAND Croat. j. for. eng. 33(2012)2

Original scietific paper – Izvorni znanstveni rad

Crown Biomass Relationships of Lebanon Oak in Northern Zagros Forests of Iran Sheyda Khosravi, Manouchehr Namiranian, Hedayat Ghazanfari, Anoushirvan Shirvani Abstract – Nacrtak Allometric relationships for estimating the biomass of the crown of Lebanon oak (Quercus libani Oliv.) have been developed through using biometric parameters such as the diameter at breast height, tree height, crown length, and crown width. The study was conducted in Ar mardeh forests in Iran’s northern Zagros mountains; for a long time the local people have been pollarding the crown of oak trees in these forests for their traditional uses. After the local people entirely cut the crowns of 48 sample trees, the weight of the crowns and their constitut ing components (leaves and branches) were measured. The results showed that the amount of the crown biomass of Lebanon oak at the stand level is about 4.98±0.81 tons ha–1 (95% confidence interval), 79% of which are branches and the rest are leaves. All the equations, represent ing the relationships between biometric parameters and the biomass of the crown or its com 2 ponents, were highly significant (p<0.001), and the adjusted coefficient of determination (Radj ) was in the range of 0.51–0.65. The most suitable relationship was a multiple regression between the crown width and the tree height, as independent variables, with the crown biomass, as the dependent variable (R2adj = 0.65). These relationships can be helpful for evaluating the crown biomass production of Lebanon oak, and can be useful for planning a sustainable forest man agement. Keywords: Allometric relationships, Biomass, Crown, Lebanon oak, Northern Zagros

1. Introduction – Uvod The dominant tree species of Baneh forests in the northern Zagros, one of the richest ecosystems in Iran, is Quercus spp. As forest dwellers depend strongly on forests, specific relationships, the so-called traditional forestry, have been developed in these areas for a long time (Valipour et al. 2009), mostly aimed providing fodder for farm animals, firewood and also wood (Ghazanfari et al. 2004). In view of the fact that most of the forest inhabitants live, in poverty, by farming animals, they are highly dependent on oak forests for feeding these animals. Therefore, each traditional owner of a forest divides the whole area of his forest into 3 or 4 parts, and each year in late August, he cuts all the crowns of the trees in one of these parts (Jazirei and Ebrahimi-Rastaghi 2003; Moradi et al. 2010). The branches cut from each tree are, along with their leaves, divided into a number of different sets named »Bakhe« by the local people. These Bakhes are kept within the crowns of trees or in nearby houses so that, Croat. j. for. eng. 33(2012)2

during winter, they can be used as a food source for farm animals, and their wood can be used as firewood for heating. Hence, short-rotation forestry is traditionally done on the crowns of oak trees. On the other hand, animals grazing in the forest for more than seven months have threatened the natural regeneration (Ebrahimi-Rastaghi 2003; Valipour et al. 2011). Though the Forest and Rangeland Organization’s policy of Iran is to conserve these forests, it has not been successful so far due to the lack of social acceptance and local participation (Ebrahimi-Rastaghi 2003; Ghazanfari et al. 2004). At the moment, the sustainability of these forests is facing a serious problem (Valipour et al. 2011), as the investigation of the correct management strategy lacks fundamental information, such as the information on forest biomass. The tree biomass plays an important role in sustainable management of forests (Zianis and Mencucini 2004), and can be regarded as an indicator for biological and economic productivity of the site (Cole and

239

S. Khosravi et al.

Crown Biomass Relationships of Lebanon Oak in Northern Zagros Forests of Iran (239–247)

Fig. 1 The location of the study area within the Kurdistan province, Iran Slika 1. Istraživano područje pokrajine Kurdistan, Iran Evel 2006). In most researches, tree biomass was estimated with the use of regression equations, which were derived through performing regression between the destructively measured dry weight of trees, as the dependent variable, and the tree dimensions as the independent variables (Nelson et al. 1999). Independent variables are easily-measurable biometric parameters such as the diameter breast height (DBH), height, basal area, and the crown dimensions. Dependent variables can be above-ground biomass including some components such as stem, branch, bark, leaf and/ or needle, bud, cone, and under-ground biomass in the form of coarse and fine roots. For instance, in a research done on two species of Quercus variabilis and Q. mongolica in central Korea, highly significant allometric equations were found between DBH and the above-ground biomass, wood and bark, live and dead branches, twigs, foliage and annual production (Son et al. 2004). In another study, Cienciala et al. (2008) developed allometric relationships between the aboveground biomass and the variables such as age, DBH, tree height, the length, and width of the crown for 51 destroyed samples from two species of oak trees (Q. robur and Q. petraea) in Czech Republic. In general, conducting researches on the biomass of oak trees is necessary to develop an understanding of natural oak tree ecosystems (Son et al. 2004). In view of specific conditions of our forests, which are closely connected with tree crowns, as mentioned above, the present study aims at establishing some useful allometric relationships through using destructive

240

samples for predicting the crown biomass in a stand of natural oak trees in Baneh forests. Our research is focused on Lebanon oak (Quercus libani Oliv.), as it is one of the main species constituting Zagros forests.

2. Materials and Methods – Materijal i metode 2.1 Study Area – Područje istraživanja The study was conducted in Armardeh forests (35˚51’40“–35˚57’55“ N and 45˚44’20“–45˚49’55“ E) 12 kilometers SW of Baneh, which is a town in the Kurdistan province in the western part of Iran (Fig. 1). The dominant species of these forests include Q. libani, Q. infectoria, and Q. brantii. The annual precipitation is about 760 mm (Valipour et al. 2009), most of which is in late autumn to early spring.

2.2 Data Collection – Prikupljanje podataka After choosing the suitable stand, the traditional owner’s prior consent was obtained before the study began. The density of trees in the stand of interest was estimated to be 370 tree ha–1, 302 of which were Q. liba ni and the others were Q. infectoria (Khosravi et al. 2012). Usually, regression researches for the direct estimation of biomass are made on the basis of a small number of trees, and therefore, not representative of the whole forest (Sawadogo et al. 2010). Hence, in this study, 50 Lebanon oaks were randomly selected in orCroat. j. for. eng. 33(2012)2

Crown Biomass Relationships of Lebanon Oak in Northern Zagros Forests of Iran (239–247)

der to determine with acceptable accuracy the crown biomass at the stand level. The trees were measured for DBH, tree height, crown length, and crown width. It should be pointed that two perpendicular diameters of the crown were measured, the arithmetic mean of which was considered as the crown width. Two trees with highly lopsided crowns, heavily defoliated, and broken tops were excluded from our calculations (Brown 1978). Statistical analyses were carried out on the remaining 48 trees. In the late summer, the forest owners cut off all the crowns of the sample trees in such a way that all the branches of the trees were pollarded, and bunched these cut branches plus leaves in a number of Bakhes. After that, all the Bakhes were weighed, and their whole sum was considered as the entire fresh crown weight for the corresponding tree from which the crowns were cut. For example, the number of 10 Bakhes with the weights of 3.26, 2.38, 4.08, 2.40, 3.46, 2.54, 2.34, 2.16, 3.02 and 2.62 kg were obtained through pollarding the sample tree 1; therefore, the entire fresh crown weight of this tree amounted to 28.26 kg. Twenty branches of each tree were randomly selected from different Bakhes, and the weights of the leaves and branches were measured separately. A number of leaves and branches were also sent to the laboratory for determining moisture contents. »Samples of leaves and branches were oven-dried at 80°C (Starr et al. 1998; Burger and Delitti 2008; Blujdea et al. 2012). During the drying process, their weights were measured every 24 hours to see whether they had reached the constant weight or not. We established that the leaves stopped losing weight after 48 hours, and the branches did the same after 72 hours. Then, their constant weights were used for determining the moisture contents«. The mean value of the moisture contents of leaves and branches was estimated to be 44% and 33%, respectively.

2.3 Data Analysis – Obrada podataka The crown biomass (Bt: ton ha-1) of Lebanon oak trees at the stand level was calculated with the use of Eq. (1): (1) Where:

s um of the crown biomass of sample trees (in terms of tons), density of the Lebanon oaks in the investigated stand,

Croat. j. for. eng. 33(2012)2

S. Khosravi et al.

t he number of the sample trees (n=48).

± 95% p robable limit of error (PLE), equivalent to half of the confidence interval, was calculated with the use of Eq. [2]: (2) Where: t

student’s t,

the standard deviation,

t he number of the samples (Batcheler and Craib 1985).

After plotting the crown biomass versus the biometric parameters, five simple regression equations including linear (y = c + αx), quadratic (y = c + αx2 + βx), multiplicative (lny = c + αlnx), exponential (lny = c + αx) and sigmoid (lny = c + α/x) were tested between the independent and dependent variables. The independent variables (x) include the DBH, tree height, crown length, crown width, and the number of Bakhes, while the dependent variable (y) is the crown biomass. Moreover, α and β are the regression coefficients, and c is the interception of the line with the y axis (Zar 1996). To predict the biomass of crowns, leaves and branches, a multiple stepwise regression approach was employed with the use of different combinations of biometric parameters. The adjusted coefficient of determination (R2adj), the root mean square error (RMSE), and the relative RMSE (RMSE%) were calculated, and (R2adj) was tested for significance. The appropriate equations with smaller values of RMSE and RMSE% and higher values of (R2adj) were selected and reported in the Result section. We excluded from our calculations any data point with the studentized residual, i.e. the ratio of the residual to its standard error, larger than ±3 (Cole and Evel 2006; Návar 2009), and recalculated the parameters of the equation. In the present study, statistical analyses were carried out with the use of SPSS 19.

3. Results – Rezultati The descriptive statistics of the biometric parameters and the crown biomass of the sample trees are shown in Table 1. The average DBH and height of these trees were 28.1 and 9.5 cm, respectively. The average crown biomass at the level of sample trees is 16.5 kg (ranging between 2.4 and 41.4 kg), 13.1 kg of which is from branches (ranging between 2.0 and 31.8 kg), and the rest from leaves (ranging between 0.4 and 12.1 kg). The average number of the Bakhes ob-

241

S. Khosravi et al.

Crown Biomass Relationships of Lebanon Oak in Northern Zagros Forests of Iran (239–247)

Table 1 Descriptive statistics of Lebanon oak sample trees (n= 48) Tablica 1. Deskriptivna statistika uzorkovanih stabala libanonskoga hrasta (n = 48) DBH

Mean – Aritmetička sredina

28.1

9.5

6.5

3.8

16.5

13.1

3.4

Range – Raspon

36.4

8.4

9.1

8.9

39.0

29.8

11.7

Maximum – Maksimum

49.4

12.9

10.6

9.6

41.4

31.8

12.1

Minimum – Minimum

13.0

4.5

1.5

0.7

2.4

2.0

0.4

S. D.

8.51

1.82

2.05

2.16

4.40

9.25

7.32

2.36

S. E.

1.23

0.26

0.30

0.31

0.64

1.33

1.06

0.34

DBH (diameter at breast height), cm – DBH (prsni promjer), cm TH (tree height), m – TH (visina stabla), m CL (crown length), m – CL (duljina krošnje), m CW (crown width), m – CW (širina krošnje), m BN (Bakhe number) – BN (svežnjevi krošanja) CB (crown biomass), kg – CB (masa krošanja), kg BB (branch biomass), kg – BB (masa grančica), kg LB (leaf biomass), kg – LB (masa lišća), kg S.D. (standard deviation) – S.D. (standardna devijacija) S.E. (standard error) – S.E. (standardna pogreška)

tained from pollarding the crown of each tree is 8, within the range from 2 to 22. We estimated the crown, branch, and leaf biomass of Lebanon oak at the stand level to be 4.98±0.81, 3.94±0.64, and 1.04±0.21 ton ha–1, respectively (95% confidence interval). The regression equations of the tree parameters and crown biomass, and its components, are shown in Table 2. All of these equations were highly significant (p<0.001). The simple regression relationships between the crown biomass and biometric parameters are illustrated in Fig. 2 (a, b, c and d). R2adj of these equations varies from 0.51 (a) to 0.57 (d). Eq. (e), which represents the relationship between the number of Bakhes and 2 and a small the crown biomass, has a high value of Radj RMSE. The results of multiple regression show better goodness of fit than the results obtained by simple regression, as the crown biomass was estimated with the use of two parameters – the crown width and tree height (f in Table 2). The regression equation for estimating the branch biomass was also calculated in the same way (g in Table 2); however, in this case, R2adj was slightly lower and RMSE higher than the previous one. But, as compared with relationships f and g, the predictor parameters for the estimation of the leaf biomass in equation h were different, and in general, the leaf biomass is less predictable than the crown and branch biomasses (R2adj = 0.54, RMSE% = 43.78).

242

4. Discussion – Rasprava The lack of information along with socio-economic problems place obstacles in the way of providing appropriate management of Zagros forests (Valipour et al. 2009). To the best of our knowledge, the present study is the first research conducted on the crown biomass of Lebanon oak trees in Zagros forests, and it was developed on the basis of destructive samples of individual trees (n=48). Hence, there are no available results of other similar researches in these forests to be compared with our results. The data analysis showed that the crown biomass of Lebanon oak constituting 81.5% of the stand trees is equal to 4.98±0.81 ton ha–1 (95% confidence interval), 79% of which is kept in branches, and the rest is in the form of leaves. Using 33 destructive samples of Quercus brantii in a study into the southern Zagros’ forests, which differ from the northern Zagros’ forests due to the impact of the local people, and the type and density of species, Adl (2007) estimated the leaf biomass to be 1 317.3 kg ha–1, which is higher than the value of the leaf biomass we estimated. The independent variable in relationship e (Fig. 2 and Table 2) is the number of Bakhes obtained from the crowns at the tree level cut by the local people. This variable, counted after the trees were pollarded, is a typical traditional parameter of high capability in pre2 = 0.88). A very It should dicting the crown biomass (Radj Croat. j. for. eng. 33(2012)2

Crown Biomass Relationships of Lebanon Oak in Northern Zagros Forests of Iran (239–247)

S. Khosravi et al.

Table 2 Equation description for the estimation of crown biomass of Lebanon oak Tablica 2. Opis formule za procjenu biomase iz krošanja libanonskoga hrasta Coefficients – Koeficijenti

Allometric equation

2 Radj

RMSE

RMSE%

–

0.51***

7.18

43.37

–16.53 (2.16)***

–

0.55***

7.10

42.89

0.65 (0.56)**

1.10 (0.16)***

–

0.51***

7.20

43.49

CB = c + a(CW)

4.45 (1.69)*

2.99 (0.38)***

–

0.57***

5.52

34.44

CB = c + α(BN2) + b(BN)

–2.09 (1.66)

–0.05 (0.16)*

2.81 (0.36)***

0.88***

2.90

18.09

ln(CB) = c + a(CW) + bln(TH)

–0.85 (0.60)

0.11 (0.03)**

1.38 (0.30)***

0.65***

5.96

35.99

ln(BB) = c + a(CW) + bln(TH)

–1.14 (0.61)

0.10 (0.03)**

1.42 (0.30)***

0.64***

4.99

38.06

LB = c + a(CW ) + bln(CL)

–0.22 (1.13)

0.06 (0.01)***

1.25 (0.67)

0.54***

1.45

43.78

Alometrijska jednadžba

c (S.E.)

a (S.E.)

b (S.E.)

ln(CB) = c + (a/DBH)

3.93 (0.19)***

–32.59 (4.58)***

ln(CB) = c + (a/TH)

4.48 (0.25)***

ln(CB) = c + aln(CL)

The statistical analyses are significant at 95% confidence interval (***p<0,001; **p<0,01; *p<0,05; and non-significant) – Statistička je analiza značajna za interval pouzdanosti od 95 % (***p<0,001; **p<0,01; *p<0,05; nije značajan) nsp>0,05, RMSE (root mean square error), kg – RMSE (srednja pogreška korijena), kg RMSE% (relative RMSE) – RMSE % (relativni RMSE) CB (crown biomass), kg per tree – CB (masa krošnje), kg po stablu BB (branch biomass), kg per tree – BB (masa grančica), kg po stablu LB (leaf biomass), kg per tree – LB (masa lišća), kg po stablu DBH (diameter at breast height), cm – DBH (prsni promjer), cm TH (tree height), m – TH (visina stabla), m; CL (crown length), m – CL (duljina krošnje), m CW (crown width), m – CW (širina krošnje), m BN (Bakhe number) – BN (svežnjevi krošanja)

be noted that the local people, experts at traditional pollarding, could predict, with an acceptable accuracy, the obtainable number of Bakhes from pollarding a tree before cutting off the crown, just by seeing the crown spread area and the diameter of each tree. DBH has been used in most studies conducted on allometric relationships as a common parameter for predicting the crown, branch and leaf biomass (Bartelink 1997; Son et al. 2004; Mitsopoulis and Dimitrakopoulos 2007; Singh et al. 2011; Zečić et al. 2011). However, in the present study, this parameter is less capable of predicting the crown biomass (R2adj = 0.51). Among simple regression equations, linear equation d, which describes the relationship between the crown width and the crown biomass, was more acceptable. By adding the height as another variable, a better goodness of fit for the estimation of the crown and branch biomass was observed (f and g in Table 2). However, the crown width and length were more suitable for predicting the leaf biomass, which is less predictable (h in Table 2). All the obtained equations were highly significant (p<0,001); however, even through quite a number of biometric parameters were used, we were not able to Croat. j. for. eng. 33(2012)2

estimate very accurately the biomass of crown or its components; this can be due to different reasons, and some of them are as follows: Þ the upper structure, i.e. the crown of a tree and the crown geometry (as compared to stem) are influenced by individual tree habitat (microclimate and competition with neighbors), resulting in large variability at this level from tree to tree, and then making it difficult to predict the crown biomass (Sawadogo et al. 2010). Generally speaking, in some species, the predictability of small components like branches and leaves is less accurate than that of larger components (Návar 2009; Sawadogo et al. 2010; Blujdea et al. 2012), Þ the poorer predictability for some species could be the result of their intrinsic physical structure due to their genetic behavior (Sawadogo et al. 2010), Þ the high dependence of forest dwellers on the forest, and the process of cutting the crowns off the trees in short time intervals have brought all the trees of either small or large diameters to have branches younger than 4 years.

243

S. Khosravi et al.

Crown Biomass Relationships of Lebanon Oak in Northern Zagros Forests of Iran (239–247)

Fig. 2 Allometric relationships of the crown biomass to tree parameters such as DBH (a), tree height (b), crown length (c), crown width (d) and the Bakhe number (e). In relationships d and e, the studentized residuals of the data points enclosed in squares are higher than ±3, and hence, they were omitted from our calculations Slika 2. Prikaz alometrijske povezanosti biomase krošanja stabala s ovim parametrima: prsnim promjerom stabla (a), visinom stabla (b), duljinom krošnje (c), širinom krošnje (d) i brojem svežnjeva krošanja tzv. Bakheov broj (e)

244

Croat. j. for. eng. 33(2012)2

Crown Biomass Relationships of Lebanon Oak in Northern Zagros Forests of Iran (239–247)

Therefore, the highly significant relation between the tree age and the crown biomass (Socha and Wezyk 2007) could be changed, and highly accurate predictions of the crown biomass become more difficult. Though the exact effects of the traditional forest management are not known, it can be seen that this management influences the forest structure and tree characteristics (Valipour et al. 2009). The results of this study can be helpful in appraising the crown biomass of Lebanon oaks in the study area, and may also prove useful for planning to replace the traditional management with the sustainable forest management. The future researches should focus on comparing the biomass of crown and other components of trees in these forests with the biomass of those components of trees in other stands of these forests, which are less influenced by the local people, so that the effects of traditional impact on biomass of the northern Zagros’ forests can be comprehensively understood.

5. Conclusion – Zaključak In the present study, the crown biomass of Lebanon oak was estimated at the stand level, and the allometric relationships were presented at the tree level for calculating the crown biomass on the basis of biometric parameters. The value of R2adj in the obtained relations varied between 0.51 and 0.65. On the basis of the results of this study, policy makers in forest management can take effective new measures to improve the traditional way of forests harvesting or can evaluate the feasibility of the replacement of other energy resources with leaf and branch biomass. In view of the fact that these forests are affected by traditional pollarding in short-term periods, in order to make use of the information gained through the present study on a larger scale, the traditional management and the level of the agreement between the existing conditions in these forests and other habitats of Lebanon oak should be considered.

Acknowledgement – Zahvala The present study partly uses the results from Sh. Khosravi’s master thesis done at the University of Tehran, Karaj, Iran. The authors are thankful to M. and A. Rashidi families, who are inhabitants of Armardeh, and to the Center for Research and Development of Northern Zagros Forests for their valuable cooperation during the research. Moreover, we thank A. Khosravi and G. Khosravi for their help in field measurement, A. Valipour for his valuable comments on the paper, and A. Safaei for his editing the manuscript. Croat. j. for. eng. 33(2012)2

S. Khosravi et al.

Finally, we are grateful to both of the anonymous reviewers for their valuable suggestions regarding the manuscript.

6. References – Literatura Adl, H. R., 2007: Estimation of leaf biomass and leaf area index of two major species in Yasuj forests. Iranian Journal of Forest and Poplar Research 15(4): 417–426. Bartelink, H. H., 1997: Allometric relationships for biomass and leaf area of beech (Fagus sylvatica L.). Annals of Forest Science 54(1): 39–50. Batcheler, C. L., Craib, D. G., 1985: A variable area plot method of assessment of forest condition and trend. New Zealand Journal of Ecology 8: 83–95. Blujdea, V. N. B., Pilli, R., Dutca, I., Ciuvat, L., Abrudan, I. V., 2012: Allometric biomass equations for young broadleaved trees in plantations in Romania. Forest Ecology and Management 264: 172–184. Brown, J. K., 1978: Weight and density of crowns of Rocky Mountains conifers. United States Department of Agriculture Forest Service, Intermountain Forest and Range Experiment Station, Research Paper INT-197, Ogden, UT. Burger, D. M, Delitti, W. B. C., 2008: Allometric models for estimating the phytomass of a secondary Atlantic Forest area of southeastern Brazil. Biota Neotropica 8(4): 131–136. Cienciala, E., Apltauer, J., Exnerová, Z., Tatarinov, F., 2008: Biomass functions applicable to oak trees grown in CentralEuropean forestry. Journal of Forest Science 54(3): 109–120. Cole, T. G., Evel, J. J., 2006: Allometric equations for four valuable tropical tree species. Forest Ecology and Management 229 (1–3): 351–360. Ebrahimi-Rastaghi, M., 2001: The rule of baneh (Pistacia atlantica) in management of beyond northern Iranian forests. Technical Report. Forest and Rangeland Organization, Tehran, Iran. 23 pp. Ghazanfari, H., Namiranian, M., Sobhani, H., Mohajer, R. M., 2004: Traditional forest management and its application to encourage public participation for sustainable forest management in the northern Zagros mountains of Kurdistan province, Iran. Scandinavian Journal of Forest Research 19 (Supp 004): 65–71. Jazirei, M. H., Ebrahimi-Rastaghi, M., 2003: Silviculture of Zagros forest. 1rd edn., University of Tehran Press, Tehran, Iran, 560 p. Khosravi, S., Namiranian, M., Ghazanfari, H., Shirvani, A., 2012: Estimation of leaf area index and assessment of its allometric equations in oak forests: Northern Zagros, Iran. Journal of Forest Science 58(3): 116–122. Mitsopoulos, I. D., Dimitrakopoulos, A. P., 2007: Allometric equations for crown fuel biomass of Aleppo pine (Pinus ha lepensis Mill.) in Greece. International Journal of Wildland Fire 16(5): 642–647.

245

S. Khosravi et al.

Crown Biomass Relationships of Lebanon Oak in Northern Zagros Forests of Iran (239–247)

Moradi, A., Oladi, J., Fallah, A., Fatehi, P., 2010: An evaluation of the capability of IRS-P6 data for monitoring pollarding forest areas of northern Zagros (Case study: Kurdistan, pollarded forests of Baneh). Journal of Agricultural Science and Technology 12: 299–308. Návar, J., 2009: Allometric equations for tree species and carbon stocks for forests of northwestern Mexico. Forest Ecology and Management 257(2): 427–434. Nelson, B. W., Mesquita, R., Pereira, J. L. G., de-Souza, S. G. A., Batista, G. T., Couto, L. B., 1999: Allometric regressions for improved estimate of secondary forest biomass in the central Amazon. Forest Ecology and Management 117(1–3): 149–167. Sawadogo, L., Savadogo, P., Tiveau, D., Dayamba, S. D., Zida, D., Nouvellet, Y., Oden, P. C., Guinko, S., 2010: Allometric prediction of above-ground biomass of eleven woody tree species in the Sudanian savanna-woodland of West Africa. Journal of Forestry Research 21(4): 475–481.

Son, Y., Park, I. H., Yi, M. J., Jin, H. O., Kim, D. Y., Kim, R. H., Hwang, J. O., 2004: Biomass, production and nutrient distribution of a natural oak forest in central Korea. Ecological Research 19(1): 21–28. Starr, M., Hartman, M., Kinnunen, T., 1998: Biomass functions for mountain birch in the Vuoskojärvi Integrated Monitoring area. Boreal Environment Research 3: 297–303. Valipour, A., Namiranian, M., Etemad, V., Ghazanfari, H., 2009: Relationships between diameter, height and geographical aspects with bark thickness of Lebanon oak tree (Quercus libani Oliv.) in Armardeh, Baneh (Northern Zagros of Iran). Research Journal of Forestry 3(1): 1–7. Valipour, A., Ghazanfari, H., Namiranian, M., Shakeri, Z., 2011: Lack of regeneration and sustainability crisis of northern Zagros forests of Iran. Development on the margin conference, Bonn, October 5–7, Germany. Zar, J. H., 1996: Biostatistical analysis, 3rd edn. Prentice-Hall Inc. New Jersey, USA, 662 p.

Singh, V., Tewari, A., Kushwaha, S. P. S., Dadhwal, V. K., 2011: Formulating allometric equations for estimating biomass and carbon stock in small diameter trees. Forest Ecology and Management 261(11): 1945–1949.

Zečić, Z., Vusić, D., Štimac, Z., Cvekan, M., Šimić, A., 2011: Aboveground Biomass of Silver Fir, European Larch and Black Pine. Croatian Journal of Forest Engineering 32(1): 369–377.

Socha, J., Wezyk, P., 2007: Allometric equations for estimating the foliage biomass of Scots pine. European Journal of Forest Research 126(2): 263–270.

Zianis, D., Mencuccini, M., 2004: On simplifying allometric analyses of forest biomass. Forest Ecology and Management 187(2–3): 311–332.

Sažetak

Odnos biometrijskih parametara i biomase iz krošanja libanonskoga hrasta na sjevernom Zagrosu u Iranu Hrastove su šume u blizini gradića Baneh među najvažnijim prirodnim ekosustavima sjevernoga Zagrosa. Zbog siromaštva lokalno stanovništvo siječe krošnje stabala libanonskoga hrasta koje im potom služe kao stočna hrana. Postupak pridobivanja krošanja stabala je sljedeći: svaki šumovlasnik podijeli šumu u tri do četiri dijela te svake godine u kolovozu na pojedinom dijelu šume kreše grane cijele krošnje stabala. Potkresane se grane slažu u svežnjeve tzv. Bakheove svežnjeve. Svežnjevi se pohranjuju i koriste zimi za prehranu stoke. Istraživanje se bavi mjerenjem biomase dobivene iz krošanja libanonskoga hrasta te uočava alometrijsku poveza nost određenih parametara stabla s krošnjom stabla, a sve radi buduće procjene biomase iz krošanja. U tu je svrhu 48 stabala nasumično odabrano te su izmjereni neki od biometrijskih parametara poput prsnoga promjera, visine stabla, duljine i širine krošnje stabala. U kasno ljeto, nakon što je lokalno stanovništvo okresalo grane krošanja te ih složilo u tzv. Bakheove svežnjeve, svežnjevi su vagani. Uzeti su uzorci lišća i grana koji su također vagani te je izmjeren sadržaj vlage. Uzorci su potom sušeni 48 do 72 sata na temperaturi od 80 °C i ponovno vagani. Analiza podataka pokazala je da biomasa iz krošnje stabala libanonskoga hrasta predstavlja 81,5% dijela stabala i iznosi 4,98 ± 0,81 tona po hektaru (95 % interval pouzdanosti), 79 % otpada na grane, a ostatak je sačuvan u lišću. Sve su dobivene jednadžbe signifikantne (p < 0,001), a prilagođeni koeficijent determinacije bio je u rasponu od 0,51 do 0,65. Međutim, primjenom brojnih biometrijskih parametara i njihova usporedba ipak nije olakšala procjenjivan je biomase krošanja stabala. Lako mjerljivi prsni promjer stabla kao jedan od parametara nije pokazao visoku korel aciju s razvojem krošnje i njezine biomase. Najprikladnija povezanost dobivena je između širine krošnje i visine stabla, kao nezavisnih varijabli, s biomasom iz krošnje stabla kao zavisnom varijablom (= 0,65). Regresijska jednadžba za procjenu biomase iz grana također je izračunata na isti način, međutim, u ovom slučaju, bila je nešto niža od

246

Croat. j. for. eng. 33(2012)2

Crown Biomass Relationships of Lebanon Oak in Northern Zagros Forests of Iran (239–247)

S. Khosravi et al.

prethodne. Širina i duljina krošnje stabala pokazale su se kao pogodniji parametri za predviđanje biomase iz lista, što je i teže procjenjiva značajka od procjenjivanja biomase iz krošnje odnosno grana. Rezultati ovoga istraživanja mogu biti od pomoći u procjeni biomase iz krošanja stabala libanonskoga hrasta istraživanoga područja te pri planiranju održivoga gospodarenja šumama. Ključne riječi: alometrijska povezanost, biomasa, krošnja stabala, libanonski hrast, sjeverni Zagros

Authors’ address – Adresa autorâ: Sheyda Khosravi, MSc. e-mail: khosravi.sheyda@gmail.com Prof. Manouchehr Namiranian, PhD. e-mail: mnamiri@ut.ac.ir Asst. Prof. Anoushirvan Shirvani, PhD. e-mail: shirvany@ut.ac.ir University of Tehran Faculty of Natural Resources Department of Forestry and Forest Economics P.O. Box 31585-4314 Karaj IRAN

Received (Primljeno): January 31, 2012 Accepted (Prihvaćeno): July 18, 2012 Croat. j. for. eng. 33(2012)2

Asst. Prof. Hedayat Ghazanfari, PhD. e-mail: hedayat@uok.ac.ir University of Kurdistan Faculty of Natural Resources Department of Forestry P.O. Box 15175-66177 Sanandaj IRAN

247

Original scietific paper – Izvorni znanstveni rad

Quantifying the Use of Brush Mats in Reducing Forwarder Peak Loads and Surface Contact Pressures Eric R. Labelle, Dirk Jaeger Abstract – Nacrtak Forest biomass from timber harvesting residues is often used during mechanized forest opera tions to improve trafficability of strip roads (machine operating trails). In particular, during cut-to-length operations brush mats from harvesting residues are created on operating trails to reduce rutting. However, forest biomass is becoming increasingly important as a source of renewable energy. To maintain its full calorific value as a biofuel, brush (tree limbs, tops, and foliage) needs to be free of any mineral soil, which is considered a contaminant in this context. In cut-to-length operations, this eliminates any use of brush as a mat to improve trafficability on machine operating trails since it gets in direct contact with mineral soil. Using brush ex clusively for biofuel will leave operating trails uncovered and can result in severe damage to forest soils. To manage the two competing uses of brush, it would be helpful to determine minimum brush amounts needed for efficient soil protection as it would potentially allow utilizing remaining brush as biofuel. This study assessed brush mats for their ability to dis tribute applied loads. As load distributing capacity of a brush mat increases, so does the resulting soil protecting effect. A total of 15 test scenarios were performed with a forwarder to analyze differences in peak loads recorded underneath brush mats of 5, 10, 15, 20, 25, and 30 kg m–2 (green mass) each subjected to 12 traffic cycles of a forwarder including unloaded and loaded movements. Highest loads were recorded within the first few forwarding cycles located on the 5 kg m–2 brush mat and then decreased on average by 23.5% as brush amount increased up to 30 kg m–2. When no brush was used (0 kg m–2) and the forwarder was in direct contact with the steel surface of the load test platform, we noticed that 97% of all peak surface contact pres sures recorded exceeded the 150 kPa pressure threshold, compared to only 41% when the forwarder was driven over the 30 kg m–2 brush mat. Keywords: biomass, brush mats, forest machinery, surface contact pressure, soil protection

1. Introduction – Uvod Until the frequent use of heavy machinery in forest industry starting in the 1960s, anthropogenic disturbances on forest soils during timber harvesting operations were quite low, both in frequency and magnitude, and were most often limited to the damages caused by horse traffic. Currently, to be productive, efficient, and safe, mechanized forest operations depend on heavy equipment to harvest, process, and transport trees. Soil disturbances are predominantly associated with in-stand timber extraction processes when mineral soil is exposed, compacted, and/or disCroat. j. for. eng. 33(2012)2

placed while machines transport timber from the felling site to a roadside landing. The Canadian forest industry applies two main mechanized harvesting methods (cut-to-length, CTL; and full tree) to harvest and extract wood efficiently and safely. The gross mass of machinery ranges from 15 to 40 metric tons and exerts static surface contact pressures of 70–180 kPa (Kozlowski 1999). This machinery operates directly on the forest floor, thus having the potential to cause severe soil disturbance (Nugent et al. 2003). The most frequent and depleting disturbance is soil compaction, which is defined as an increase in soil density (Craig

249

E. R. Labelle and D. Jaeger

Quantifying the Use of Brush Mats in Reducing Forwarder Peak Loads ... (249–274)

2004). By increasing a soil’s mechanical resistance to penetration, the densification process can have a direct impact on plant growth through a reduction of air exchange and infiltration rate (Forristall and Gessel 1955; Froehlich and McNabb 1984; Corns 1988). The mechanized CTL harvesting method, which dominates in Atlantic Canada, usually requires a harvester to fell and process trees and a forwarder to transport the logs from the machine operating trails to roadside. When applying the CTL method, harvesting equipment travels on trails typically covered by a layer of forest biomass in form of harvest residues (limbs, tops, and foliage of trees) resulting from the processing of felled trees (Fig. 1). This layer of debris acts as a so-called brush mat, which helps to improve traction and prolong trafficability. The brush mat distributes machine loads over a greater area, thereby lowering peak loads exerted on forest soils and, as such, mitigates soil disturbances and related negative impacts on plant growth (Wronski et al. 1990; Bettinger and Kellogg 1993; Richardson and Makkonen 1994). However, the desire to reduce carbon emissions from fossil fuels (oil and natural gas) in combination with their high and volatile price has increased the focus of forest stakeholders in using harvest residues as a source of bioenergy. A pre-requisite for any viable bioenergy operation is that biomass remains free of contaminating mineral soil, thus maintaining its full calorific value (Eliasson 2005). To avoid such contamination, operators have changed the common practice of delimbing trees in front of the harvester thereby creating a brush mat on the trail to delimbing trees on the side of the trail, which leaves the trails bare and unprotected (Eliasson

2005). With the absence of brush, a machine’s surface contact pressure is directly and fully exerted to the ground, often resulting in increases of soil density and other disturbances. Conversely, brush used on machine operating trails for soil protection cannot be reused for bioenergy generation due to contaminants. In order to optimize the two competing uses of brush, knowledge of minimum quantities and qualities of brush for effective soil protection on machine operating trails is needed. With this knowledge, brush amounts necessary for soil protection could be allocated and the remaining brush, if any, utilized as biofuel without compromising forest soil integrity along machine operating trails. The study attempts to provide necessary information in this respect.

1.1 Research objectives – Ciljevi istraživanja Prevention of high surface contact pressures exerted by forest machinery is discussed with the following objectives: 1) Quantify (dynamic) forwarder loads transmitted below brush mats of 5, 10, 15, 20, 25, and 30 kg m –2 (green mass) compared to loads without applying any brush mat (0 kg m–2 scenario). 2) Determine minimum amount of brush required to limit the dynamic surface contact pressure acting on the soil to a certain threshold. 3) Quantify the impact of repetitive loadings of brush mats on their ability to distribute machine loads affecting the soil below the mats. In this article, all brush amounts presented, including those used as reference from previous studies, are expressed in green mass. We also defined a loading as the impact associated with a tire passing over the load test area.

1.2 Impacts of forest machinery on soil physical properties – Utjecaj šumskih vozila na fizikalne značajke tla

Fig. 1 Forwarder operated over a brush covered machine operating trail Slika 1. Forvarder na traktorskoj vlaci pokrivenoj zastorom granjevine

250

Most machinery used during forest operations are operated in off-road conditions directly on the forest floor. In CTL operations, both harvester and forwarder regularly share the same off-road machine operating trails (normally 3.5 m wide and spaced by 15–20 m from adjacent trail centrelines), which cover approximately 17–23% of the total area to be logged (McNeel and Ballard 1992). This area, directly exposed to machine traffic, is often subject to severe soil property alterations (density increase and/or soil displacement resulting in reduced pore space and lowered hydraulic conductivity) due primarily to the normal load exerted by forest machinery (150 to 400 kN; Kozlowski Croat. j. for. eng. 33(2012)2

Quantifying the Use of Brush Mats in Reducing Forwarder Peak Loads ... (249–274)

1999). As a soil is being compacted, soil particles are moved closer together, which reduces air voids, in particular macro pores essential for plant growth (Adams and Froehlich 1984; Taiz and Zeiger 1998; Brady and Weil 1999). This reorganization of soil particles usually leads to increased density, which directly increases the soil’s resistance to penetration and lowers infiltration rates (Vepraskas 1988). Soil compaction by a forwarder or skidder in the range of 15–25% compared to pre-impact levels, following a single traffic cycle is not uncommon and has been reported by Voorhees et al. (1978); Bock and Van Rees (2002); and Labelle and Jaeger (2011). The highest increase of soil density usually occurs within the first few traffic cycles (Froehlich 1979; Bradford and Peterson 2000; Han et al. 2006; Syunev et al. 2009). Continuation of traffic further increases soil density but at lower rates (Lockaby and Vidrine 1984; Froehlich et al. 1986; Brais and Camiré 1998; Jamshidi et al. 2008; Han et al. 2009; Labelle and Jaeger 2011). The effect of machine induced soil disturbance may also persist for many years. Residual forest soil compaction has been identified for periods ranging from 15 (Froehlich et al. 1986) to 55 years (Power 1974) on sites in temperate climate zones without winter frost. Five years after mechanized CTL operations on soils exposed to frequent freeze/thaw cycles in New Brunswick, Canada, Labelle and Jaeger (2011) reported no sign of soil bulk density natural rehabilitation. A technique commonly used in CTL operations to avoid rutting on machine operating trails and to keep them operational during high traffic frequencies is to follow the corduroy road construction principle and reinforce the bearing capacity of the trails by having the harvester create a brush layer of limbs, branches, and tree tops (McDonald and Seixas 1997).

1.3 Effects of brush on reducing soil compaction Djelovanje zastora granjevine na smanjenje zbijanja tla The CTL method was developed in the Nordic countries in the early 1970s (Vidrine et al. 1999). The innovation of this method was the immediate processing of felled trees on-site, directly in the stand. Since its beginning, trees have been delimbed, bucked to appropriate length, and topped on the machine operating trail. This on-trail delimbing process started due to convenience for the operator to process trees in front of the harvester rather than on the side of the trail where boom operating space might be restricted by residual trees. Operating forest machinery on a brush mat proved to be beneficial in lowering rut depth by increasing floatability, which also helped to extend the Croat. j. for. eng. 33(2012)2

E. R. Labelle and D. Jaeger

window of operations into seasons with increased soil moisture content (Tufts and Brinker 1993; McMahon and Evanson 1994; Eliasson and Wästerlund 2007). Since the early 1980s brush has been studied for its potential to reduce soil disturbances. First, Jakobsen and Moore (1981) investigated the impact of brush on soil resistance to penetration (soil impedance) and have been followed by McMahon and Evanson (1994); McDonald and Seixas (1997); Han et al. (2006); and Han et al. (2009). These studies found that brush mats used on machine operating trails had a beneficial impact on mitigating the increase of soil resistance to penetration, in particular when the mats consisted of a thick brush layer of approximately 20 kg m–2. Additionally, brush mats placed on top of soils with high moisture content helped to reduce the increase of penetration resistance by forwarder traffic to a larger extent compared to brush mats applied on dryer soils (McDonald and Seixas 1997; Han et al. 2006). The aforementioned studies are important contributions in understanding the significance of brush mats on maintaining soil integrity but specific effects depend both on site and machine characteristics. Due to the high variability of soil conditions at harvesting sites (soil texture, soil depth, soil moisture content, and organic matter content), combined with varying characteristics of brush, it appears difficult to predict the general reinforcing effect that brush might have on soil bearing capacity. Ideally, it would be preferable to test the effects of various brush amounts on an array of soil types, soil moisture contents, and machine configurations, but current budget, equipment, and logistic restrictions did not allow for these kinds of tests. However, determining the effect of brush mats of varying quantity and quality on alteration of machine load distribution below brush mats uncoupled of soil property variation is one approach to enhance knowledge about brush mat capabilities independent of site characteristics. Shifting the use of brush away from soil protection, in particular with its new competing use as a source of biofuel, could prove costly considering the associated risks of reduced plant growth and the slow natural recovery rate of compacted soils.

1.4 Functions of a brush mat – Svrha zastora granjevine During the processing phase of CTL forest operations, branches placed directly on the ground in front of forest machines can provide several key functions to protect forest soils. Having branches intertwined within a brush mat increases friction and creates a reinforced surface, which can expand the contact area between machine running gear and the forest soil. At

251

E. R. Labelle and D. Jaeger

Quantifying the Use of Brush Mats in Reducing Forwarder Peak Loads ... (249–274)

the broadest level, this increased contact area should reduce average loads per unit area below the entire brush mat. This reduction of average loads is attributed to reduced peak loads directly located underneath tires of the equipment and load transfers in extended areas adjacent to the loading impact in both lateral and longitudinal directions.

2. Methodology – Metode rada 2.1 Testing device – Naprava za istraživanje To measure and record dynamic loads exerted by forest machines, a prototype load test platform was designed and constructed. This platform allowed for

full-scale tests of loaded forwarders driving over various brush mats to analyze load distribution below the mats. The entire load test platform was composed of three separate sections: ramps, in- and out-feed, and the load test area itself (Fig. 2A). The principal part of the structure was the load test area measuring 4.1 m by 2.5 m for a total area of 10.3 m2 and equipped with 24 high capacity (450 kN) compression load cells, each able to measure independent loads on a 30.5 cm by 30.5 cm resolution. This resolution corresponds to the size of loading plates, a matrix of 104 steel plates forming an engineered surface below which load cells were positioned. This section was constructed and mounted to a concrete base. In- and out-feed sections were built at the same height (19 cm) as the load test area to per-

Fig. 2A Three main sections of load test platform including in- and out-feed sections and load test area and 2B Forwarder traffic over the brush covered load test area Slika 2A. Tri glavna dijela ispitne platforme (ulazni dio, mjerni dio, izlazni dio). 2B. Kretanje forvardera po mjernom dijelu platforme pokrivenom zastorom granjevine

252

Croat. j. for. eng. 33(2012)2

Quantifying the Use of Brush Mats in Reducing Forwarder Peak Loads ... (249–274)

E. R. Labelle and D. Jaeger

Fig. 3 Top view schematic of test layouts used during testing. Black circles represent active load cells. Virtual active zones of the clustered test layout (four groups of six load cells) are identified by diagonal lines, while virtual active zones of the transect test layout (two full-width adjacent rows) are shown by horizontal lines Slika 3. Tlocrt skice korištenoga ispitnoga rasporeda. Crni krugovi prikazuju dinamometre. Mjerna područja grupimičnoga ispitnoga rasporeda (četiri grupe po šest dinamometara) prikazana su kosim crtama, a poprečni je mjerni raspored (cijelom širinom, jedan do drugoga dva reda dinamometara) prikazan ravnim crtama. mit testing of the forwarder in a level position(zero percent gradient), thus avoiding potential wheel slip and a change in machine centre of gravity, while also minimizing variations in brush mat compressibility. Furthermore, each in- and out-feed section was of sufficient length to allow the full wheel-base of the forwarder to completely pass the load test area, a key design feature enabling the distinction between discrete loading events.

test layout offered the highest resolution to capture machine footprints with a total active area (91.5 cm by 61.0 cm) per cluster. In addition to obtaining a high load resolution underneath machine tires, it was also of interest to understand how the brush could distribute applied loadings laterally. Therefore, load cells were also installed in a transect layout on two adjacent rows throughout the full width of the load test area (12 load cells wide; Fig. 3).

Depending on the required load resolution, load cells could be placed in different arrangements, in socalled test layouts, within the platform. Two different test layouts (clustered and transect) were used during testing. To specifically quantify the impact of the forwarder, the 24 load cells were first positioned in a clustered test layout (four clusters of six load cells each arranged in two adjacent rows of three load cells wide), directly located in forwarder tracks (Fig. 3). This

2.2 Sampling procedure – Način uzorkovanja

Croat. j. for. eng. 33(2012)2

2.2.1 Control parameters – Nadzor parametara To establish control parameters, the Timbco forwarder (specifics in section 2.3) was first driven unloaded over the bare load test platform (without brush cover using clustered test layout). The resulting dynamic loads from each tire were measured by the load cells, enabled by a 25 input channel System 5000 scan-

253

E. R. Labelle and D. Jaeger

Quantifying the Use of Brush Mats in Reducing Forwarder Peak Loads ... (249–274)

ner, and stored in a StrainSmart data acquisition system (Vishay Micro-Measurements 2011). For data analysis, loads (kN) recorded by each load cell were also converted to surface contact pressure (kPa) by relating the measured load to the surface area of a socalled virtual active zone (930.3 cm2, consisting of four quarters of four adjoining loading plates; (Fig. 3). The control test was replicated once with the forwarder driven over the platform in the same position to verify the accuracy and precision of the load recording system. The same procedure was then repeated with the forwarder loaded with 6,680 kg of dry logs. In order to avoid weight fluctuations of the loaded logs during the four week long testing period, dry logs with relatively stable moisture content were chosen. As a result, we only had a partial load compared to the log bunk’s full capacity of 20 metric tons. 2.2.2 Brush mat construction and forwarder traffic Izrada zastora granjevine i kretanje forvardera After control parameters were assessed, actual testing with brush of varying quantity and quality was performed. Fresh softwood brush (balsam fir Abies balsamea (L.) Mill. and black spruce Picea mariana (Mill.) B.S.P.) collected from on-going mechanized CTL clear felling operations in Fredericton, New Brunswick, Canada was stored inside the unheated storage hall to reduce air drying, while protecting it from precipitation. Prior to any brush amount test, branches used to create the brush mats were characterized individually by specie, diameter, and length. Aside from species identification, branches were assigned to one of four diameter classes (x ≤ 10 mm, 10 mm < x < 30 mm, 30 mm ≤ x ≤ 60 mm, and x > 60 mm) and to one of five length classes (y ≤ 1 m, 1 m < y < 2 m, 2 m ≤ y ≤ 3 m, 3 m < y < 4 m, and y ≥ 4 m). Following classification, branches were weighed with a digital scale and placed perpendicular to the direction of travel on the platform to simulate branch positioning of in-wood delimbing by a harvester until the target brush amount was reached (McMahon and Evanson 1994; Wood et al. 2003; Han et al. 2006; Akay et al. 2007). In each test, brush samples (approx. 150 g) were collected randomly from the brush (branches and needles) to determine branch moisture content by oven drying the material at 105°C until constant mass was reached. Prior to any forwarder traffic, the thickness of each brush mat was assessed. Since the top boundary of the mat was difficult to determine because of its irregular shaped surface, a standard measuring method was developed by placing a known and constant load on a measuring board on top of the mat. The thickness of the mat was then determined by measuring the vertical distance between the bottom of

254

the measuring board and the surface of the steel load test area. Brush thickness was recorded at six locations, four positioned in track areas and two in outside-track areas before traffic as well as after two, six, and 12 forwarding cycles (a cycle is comprised of two forwarder passes over the load test platform) to quantify brush mat compressibility after repetitive loadings. Once a brush mat was completed and its pre-impact thickness recorded, all load cells were zeroed and forwarder traffic commenced. Due to space limitations at the testing site, the unloaded forwarder was driven backwards (at a speed of 1.5 km h –1) into the hall over ramps and in-feed section onto the load test area and further on to the out-feed section until the front bogie axle passed the load test area (Fig. 2B). After a short interval, the forwarder was driven from the out-feed section at the same speed in a forward movement, again, onto the load test area and over the in-feed section with ramps outside the hall. During these machine passes, all 24 load cells were activated to record dynamic loads. Afterwards, the forwarder was loaded using the same load as in control tests (6,680 kg) and driven over the platform in the same pattern as explained previously. These two unloaded and two loaded passes over the brush mat constituted two forwarding cycles. Similar to in-stand forest operations, where the combination of one unloaded forwarder pass going into the cut block to gather wood and one loaded pass as the forwarder returns to the landing constitutes one forwarding cycle, we also considered one unloaded and one loaded pass over the platform as one complete cycle. However, for easiness of operation we always paired the inward and outward movements of the unloaded and loaded forwarder. A total of 12 forwarding cycles were tested to determine the capacity of the brush mat to attenuate surface contact pressure over repetitive loadings. When all traffic frequencies were completed on a specific brush mat (combination of 12 unloaded and 12 loaded passes), the forwarder was driven outside the hall, the platform was cleared of the compressed brush, and a new brush mat was created from fresh brush for the next test. Replacing brush between tests was essential since the properties of branches (compressibility, yield point, tensile strength, etc.) were likely altered by the effect of machine loadings. Each testing scenario with one of the six brush amounts (5, 10, 15, 20, 25, and 30 kg m–2) over the clustered test layout was replicated twice (three tests in total) to increase statistical power of analyses. Following these tests, load cells were re-positioned into the transect test layout covering the full width of the load test area to Croat. j. for. eng. 33(2012)2

Quantifying the Use of Brush Mats in Reducing Forwarder Peak Loads ... (249–274)

assess the ability of brush mats to distribute loads laterally. With this transect layout, three brush amounts (10, 20, and 30 kg m–2) were tested directly over the steel load test area. Replicates were omitted for this analysis since these tests were mainly conducted to verify the operation and accuracy of the load cells in their new positions prior to testing brush mats on the sand covered load test area. After testing the three brush mats with the transect layout, brush mats of 10, 20, and 30 kg m–2 (each brush amount replicated twice) were tested on top of a 20 cm thick layer of sand placed on the load test area to test the load distributing capabilities of brush mats when applied on a flexible surface. Prior to any tests, the sand layer located on the load test area was compacted using a plate compactor for three minutes. Following the compaction phase, we placed a brush mat on top of the sand layer. Before any trafficking commenced, all load cells were zeroed to account for the added mass of either the brush mat (in clustered layout over steel load test area) or mass of sand and brush mat for all tests performed over the sand covered load test area. Once load cells were zeroed, the forwarder was driven over the mat in the same sequence as for tests without sand layer described before. Once all forwarder traffic cycles had been completed for a respective scenario, brush was removed from the load test area and discarded. Sand was then loosened with a shovel and re-compacted with the plate compactor before testing the next brush mat amount. Lastly, during testing of brush mat scenarios over sand, we collected sand samples to obtain gravimetric moisture content by oven drying the material at 105°C until constant mass was reached. The approach of dividing a peak load from a single load cell by the surface area of a loading plate to derive surface contact pressure (kPa or kN m–2) mentioned previously was straight forward. However, the presence of brush and the sand layer over the load test area made it more difficult to maintain constant positioning of the forwarder in relation to the platform during traffic. To differentiate recorded loads affected by the load distributing capability of varying brush amounts from those loads simply affected by machine positioning, where the load was distributed between adjacent load cells, we summed the peak load and the second highest load (obtained from an adjacent load cell).

2.3 Forwarder specifications – Značajke forvardera A 2000 Timbco TF820-D forwarder with a tare mass of 23,500 kg and a load capacity of 20,000 kg was used for all tests. This eight-wheel (28L-26 tire size) forwarder had two independent bogie axles. Two OlofsCroat. j. for. eng. 33(2012)2

E. R. Labelle and D. Jaeger

fors steel flexible tracks with widening plates, weighing 1,100 kg each, were installed on the rear bogie axle during all test scenarios. Cross members (10 cm wide) of these steel flexible tracks were spaced by 25 cm. The front bogie axle remained without steel flexible tracks during testing for better manoeuvrability over the load test area of the platform and associated in-/out-feed sections. According to the PASCAL software (FPInnovations’ ground pressure calculator), nominal surface contact pressure underneath the front rubber tired bogie axle was 65 kPa loaded and 35 kPa below the rear loaded bogie axle equipped with tracks (Makkonen 2007). These nominal surface contact pressures are based on a 6,680 kg load. Tire inflation pressure in all eight tires remained constant at 157 kPa throughout the experiment. Unlike the machine movements during regular in-stand forest operations. where the boom of the Timbco is articulated in an upright position in front of the machine, the boom was straight and fully extended in front of the forwarder during all tests to accommodate the limited height clearance inside the testing facility.

2.4 Statistical analyses – Statistička analiza Statistical analyses were performed with SPSS® (SPSS 2007) and Minitab® (Minitab Inc. 2010) statistical software. Dependant variables were any type of load readings (peak loads, sum of peak and second highest loads, third highest loads) obtained directly from the load cells or peak surface contact pressures once loads were related to the surface area of a loading plate. To determine the impact of an independent variable (brush amount, log bunk load status, forwarder traffic frequency, etc.) on the chosen dependant variable, a series of one-way analysis of variance (ANOVA) tests were performed and a probability level of 0.05 was chosen during all statistical tests.

3. Results – Rezultati This section is structured according to testing scenarios and will first focus on results obtained when the forwarder was driven directly over the steel load test area, without brush mats. These initial analyses will address: 1) peak loads, 2) the relationship between peak and second highest loads, 3) sum of peak and second highest loads, and 4) the effect of sand on stress propagation. Afterwards, general brush mat characteristics will be presented, followed by analyses of peak loads, sum of peak and second highest loads, and dynamic surface contact pressures recorded below varying brush amounts when exposed to different forwarder traffic frequencies.

255

E. R. Labelle and D. Jaeger

Quantifying the Use of Brush Mats in Reducing Forwarder Peak Loads ... (249–274)

Fig. 4 Peak load frequency distributions for all three test scenarios (no brush). Numbers shown inside dotted lines indicate the percentage of loads within the respective 10 kN load class and the curves represents a normal distribution Slika 4. Raspodjele učestalosti vršnih opterećenja za sve tri inačice istraživanja, bez zastora granjevine. Brojevi unutar isprekidanih crta označuju udjele opterećenja unutar pojedinih razreda (raspona 10 kN), a krivulje predstavljaju normalne distribucije

3.1 Effects of machine related variables on load distribution (no brush) – Djelovanje varijabli vozila na raspodjelu opterećenja (bez zastora granjevine) 3.1.1 Peak loads – Vršna opterećenja In the following section, all peak loads presented were recorded when the forwarder was driven directly over the load test area, without any brush mat. To assess

256

recorded loads, we first analyzed peak load frequency distribution per scenario (clustered layout over steel, transect layout over steel, and transect layout over sand) and presented these results in Fig. 4. In addition to the frequency distribution, we also determined the percentage of peak loads within four 10 kN load classes (numbers between dashed lines) and included normal distribution curves as reference. Croat. j. for. eng. 33(2012)2

Quantifying the Use of Brush Mats in Reducing Forwarder Peak Loads ... (249–274)

Peak loads varied between 7 kN (1 observation) and 32 kN (1 observation) for the clustered test layout (Fig. 4A). Most of the peak loads (91%) were located between 10 and 30 kN and only 9% were below 10 kN. Peak loads recorded from the transect test layout on steel varied from 7 kN (1 observation) to 35 kN (2 observations). With this test layout, 5% of peak loads (10 observations) exceeded 30 kN whereas < 1% were exceeding this load during the clustered test layout. The percentage of peak loads located between 10 and 20 kN (90%) was almost identical to the one obtained from the clustered test layout. Results were different with the transect layout over sand scenario. With the addition of the sand layer, peak loads varied from 5 kN (2 observations) to 28 kN (3 observations) and 27% (35 observations) of peak loads were below 10 kN. The higher percentage of peak loads located below 10 kN in combination with no loads exceeding 30 kN clearly demonstrate the effect of the added sand layer in distributing applied loads. 3.1.2 Relationship between peak and second highest loads – Odnos između vršnoga i drugoga najvećega opterećenja As the area of contact underneath single tires or tracks of the forwarder was always exceeding the surface area of a single loading plate and since machine location in relation to the load test area had the potential to vary between loading events, it was necessary to consider the loads of adjacent loading plates and of the related load cells during the analyses. Simply due to machine positioning, the wheel load exerted by the forwarder could either be recorded by a single load cell with a relatively low load on an adjacent load cell in the case where most of the load was located directly above a single load cell or it could be shared more equally between two or more adjacent load cells. This variation made it difficult when analyzing peak loads of single cells to differentiate between the effect of machine positioning and the effect of brush mats on the recorded loads. Therefore, we first analyzed the relative load distribution between adjacent load cells to identify potential similar pattern between testing scenarios without brush. This was done by first identifying the highest recorded load during a pass of a tire or tracked tire and then identifying the second highest load of an adjacent load cell. In a second step, the second highest peak load was divided by the peak load to derive the relative magnitude of the second highest load in percent of the peak load. Fig. 5 shows the frequency distribution of the relative second highest recorded loads for all three test Croat. j. for. eng. 33(2012)2

E. R. Labelle and D. Jaeger

layouts. If high frequencies of the relative second highest loads had been located in the lower or upper limits of the abscissa, it would indicate relatively low or equally high second highest loads compared to peak loads, respectively. This could also indicate low variation of machine positioning between different tests. It would have then been appropriate to concentrate the analyses mainly on peak loads since either the load exerted to the second load cell is relatively low or similar to the peak load. Results from the clustered layout showed a lower percentage of second highest loads to peak loads in the < 20% class and a higher percentage in the upper > 80% class compared to the transect layout over steel scenario (Figs. 5A and B). This indicates that loads were shared more equally during scenario 1 than they were in scenario 2. When the forwarder was operated directly over the sand covered load test area, only 15% of loads were located below 40% and the number of observations located between 80–100% increased up to 31% (Fig. 5 C) again showing that the sand layer was able to distribute loads more effectively between two adjacent loading plates. However, since the majority of the second highest loads (approx. > 70%) was within the 20 to 80% range of the peak loads, using only peak loads or conversely second highest loads for analyzing the effects of varying brush mats would not have been sufficient. For this reason, the following analyses will mostly focus on the sum of peak and second highest loads since single peaks may be biased by machine positioning, which could hide the effect different brush mats have on load distribution. 3.1.3 Sum of peak and second highest loads – Zbroj vršnoga i drugoga najvećega opterećenja Loads exerted by a forwarder are directly related to the mass acting on the running gear. Consequently, quantifying loads from unloaded and loaded front and rear single tires and separating them per axle was essential in determining accurate loads. Combining loads from front and rear tires could potentially overshadow trends that might be present. When individualizing loads by log bunk load status (unloaded and loaded), we noticed a statistical difference (p = 0.001) of the mean sum of peak and second highest loads between unloaded and loaded rear tires (Fig. 6). Mean sum of peak and second highest loads obtained from single tires of the rear unloaded axle were 16.2, 16.1, and 12.5 kN and increased to 28.9, 29.1, and 20.6 kN when loaded for Figs. 6A, B, and C, respectively. These sum of peak and second highest loads were therefore 78.4, 80.2, and 64.6% higher for loaded than unloaded scenarios for Figs. 6A, B, and C, respectively. However, as the majority of the load on the log bunk is distrib-

257

E. R. Labelle and D. Jaeger

Quantifying the Use of Brush Mats in Reducing Forwarder Peak Loads ... (249–274)

Fig.5 Percent of second highest load to peak load frequency distributions. Numbers shown inside dotted lines indicate the percentage of loads within the respective percentage class Slika 5. Raspodjela učestalosti udjela drugoga najvećega opterećenja u odnosu na vršno opterećenje. Brojevi unutar isprekidanih crta označuju udjele opterećenja unutar pojedinih postotnih razreda uted to the rear bogie axle, tires from the front axle showed no statistical difference (p = 0.305) between the mean sum of peak and second highest loads recorded from unloaded and loaded test scenarios. Mean sum of peak and second highest loads from front tires showed only a difference of 0.3 to 1.2 kN, (0.5 to 3.3%) between the unloaded and loaded test scenarios.

258

A priori, we were expecting highest loads to be recorded underneath the rear loaded tires. However, the modest load of 6,680 kg on the log bunk was not enough to exert loads equal to those recorded underneath the front tires, which are located in proximity to the heavy boom, hydraulic rotary system, and the operator cabin. Nevertheless, our analyses mainly foCroat. j. for. eng. 33(2012)2

Quantifying the Use of Brush Mats in Reducing Forwarder Peak Loads ... (249–274)

cused on the loads exerted by the rear tires because most forwarders with rear bogie axles are equipped with steel flexible tracks. The difference in loads of the rear tires between loading regimes (unloaded vs. loaded) was less pronounced for the tests performed over the sand covered load test area and reasons for this will be explained in the discussion. After further analysis, we also determined that loads exerted by single rear axle tires were similar between the two machine sides, thus not requiring machine sides to be separated in upcoming analyses.

3.2 Effect of the sand layer on stress propagation Djelovanje sloja pijeska na rasprostiranje naprezanja Thus far, mean loads recorded amongst different test layouts have been rather difficult to compare due to the added sand layer, which in itself reduced loads by placing further distance between the actual load and the load cells. Since load cells were zeroed prior to any test, the vertical stress or dead load caused by the sand did not impact the load readings recorded by the load cells. However, because of shearing resistance within the sand layer, applied loads to the sand mass were spread laterally with increasing depth from the area of application. In short, as depth increases, the area over which new stresses occur will increase but the magnitude of the stresses will decrease (McCarthy 2002). Based on Boussinesq’s (1885) equation for stress propagation, Steinbrenner (1936) developed influence coefficients (Id) for calculating the vertical stress distribution in soils caused by a rectangular loaded area using the load per unit area (q) and the vertical stress increase (Δσz) resulting at a certain depth (z) from a loaded area of length (L) and width (B) as obtained by equation 1. Δσz = q · Id

(1)

where: ,

Based on equation 1, we calculated that on a loaded area of 0.195 m2 (L = 0.273 m, B = 0.714 m), the 20 cm layer of sand (z = 0.20 m) added on top of the steel load test area lowered applied pressures by 34% compared to pressures recorded directly below the tire (Fang and Daniels 2006). The loaded area described above relates to the contact area below a 28L-26 Firestone tire as Croat. j. for. eng. 33(2012)2

E. R. Labelle and D. Jaeger

equipped on the Timbco forwarder and assumes a 7.6 cm vertical penetration within the sand layer (Firestone 2010). For example, a wheel load of 30 kN applied to a 28L-26 tire (same contact area as stated above), would equal a nominal ground pressure of 153.9 kPa or kN m–2 directly below the tire. At a depth of 20 cm (vertical distance from the tire’s contact area to the location where the stress is calculated) the pressure located underneath the tire would be reduced by 34% to equal 101.6 kPa or kN m–2. It is noteworthy to mention that Boussinesq’s (1885) stress propagation theorem, based on elastic theory, is only an approximation since it was derived for homogeneous, isotropic materials of semi-infinite extent. Steinbrenner’s (1936) influence coefficient also assumes uniform load distribution below the loaded area (Liu and Evett 1992). The mineral sand used as a covering layer on top of the load test area was relatively homogeneous (80% of particles were within 0.3 mm ≤ x ≤ 2.0 mm and average moisture content was 6.7%) but was located on top of a rigid surface, thus not having the same properties in all directions. Unlike the basic assumption of equation 1, load distribution below a forwarder tire is not uniform and is highly influenced by the operating surface. These conditions likely affected the percent reduction described above.

3.3 General brush mat characteristics – Glavne značajke zastora granjevine A total of 5,511 branches for a combined green mass of 5,550 kg were characterized by species that created the brush mats required for all tests scenarios. Besides differing between species, we also categorized branches in diameter and length classes. We attempted to compose brush mats using similar degrees of branch size and tree tops. The highest frequency of branch diameter, regardless of brush amount, was within class two (1 to 3 cm diameter; Fig. 7). Over 80% of branches used for all brush mats were within diameter class one and two and a maximum of 4% were in the diameter class four (> 6 cm). When averaging all brush mat replicates of the same amount, seven out of 12 mats showed a higher frequency distribution in branch length class two (1 to 2 m), while the other five brush mats had a higher distribution in length class one. The percentage of branches greater than 3 m in length used during testing was quite low and only varied between 0 and 4% of total branches used. Branch and needle moisture content was also monitored for each brush mat and varied between 45.3 and 55.2% of the green mass. In general, brush mats used directly over the steel load test area had an average moisture content of 52.1% compared to 48.8% for

259

E. R. Labelle and D. Jaeger

Quantifying the Use of Brush Mats in Reducing Forwarder Peak Loads ... (249–274)

Fig. 6 Mean sum of peak and second highest loads presented by single tire and log bunk status directly over load test area (no brush). A different letter indicates a significant difference based on a one-way ANOVA test at the 0.05 probability level Slika 6. Sredina zbroja vršnoga i drugoga najvećega opterećenja prikazanoga za pojedine kotače i (ne)utovarenost vozila neposredno na ispitnom dijelu platforme bez zastora granjevine. Različita slova označuju značajnost razlike na osnovi jednostruke analize varijance za razinu vjerojatnosti od 5 % brush mats used over the sand covered load test area. Explanations for this slight variation will be presented in the discussion. Once the target amount of brush mat was reached (e.g. 20 kg m–2) and branches were placed on the load

260

test area, brush thickness measurements were recorded pre impact and also following two, six, and 12 forwarding cycles in both track- and outside-track locations. The thickness of all brush mats pre impact was similar for both track and outside-track areas and Croat. j. for. eng. 33(2012)2

Quantifying the Use of Brush Mats in Reducing Forwarder Peak Loads ... (249–274)

E. R. Labelle and D. Jaeger

Fig. 7 Branch diameter and length class frequency distributions. Each figure represents the total frequency for all replicates. Numbers shown inside dotted lines indicate the percentage of branches within the respective branch diameter or length class (rear loaded axle only). Branches were assigned to one out of four diameter classes 1 (x ≤ 10 mm), 2 (10 mm < x < 30 mm), 3 (30 mm ≤ x ≤ 60 mm), and 4 (x > 60 mm) and to one out of five length classes 1 (y ≤ 1 m), 2 (1 m < y < 2 m), 3 (2 m ≤ y ≤ 3 m), 4 (3 m < y < 4 m), and 5 (y ≥ 4 m) Slika 7. Raspodjele učestalosti razreda promjera i duljina grana. Svaki grafikon prikazuje ukupnu učestalost svih ponavljanja. Brojevi unutar isprekidanih crta označuju udjele unutar pojedinoga razreda promjera i duljina grana (samo opterećena stražnja osovina). Grane su pridružene u jedan od četiriju razreda promjera 1 (x ≤ 10 mm), 2 (10 mm < x < 30 mm), 3 (30 mm ≤ x ≤ 60 mm) i 4 (x > 60 mm) te jedan od pet duljinskih razreda 1 (y ≤ 1 m), 2 (1 m < y < 2 m), 3 (2 m ≤ y ≤ 3 m), 4 (3 m < y < 4 m) i 5 (y ≥ 4 m) ranged from 9 cm for the 5 kg m–2 brush mat to 70 cm for the 30 kg m–2 brush mat in the track areas and from 9 cm for the 5 kg m–2 brush mat to 74 cm for the 30 kg Croat. j. for. eng. 33(2012)2

m–2 brush mat for outside-track areas (Fig. 8). Logically, brush mat thickness increased with increasing brush amount. The largest decrease in thickness was

261

E. R. Labelle and D. Jaeger

Quantifying the Use of Brush Mats in Reducing Forwarder Peak Loads ... (249–274)

Fig. 8 Mean brush mat thickness measured over repetitive loadings at two locations (track and outside track) on the load test area for each brush amount Slika 8. Srednja debljina zastora granjevine mjerena nakon višekratnih opterećenja na dva mjesta (u tragu i izvan traga) ispitne platforme za svaku gustoću zastora granjevine measured between pre impact and following the first two forwarding cycles. Beyond this point, a further increase in loading frequency per test seemed to only slightly reduce brush mat thickness. Regardless of initial brush mat amount, the mean brush mat thickness in track locations was reduced to < 10 cm following 12 forwarding cycles.

262

3.4 Effects of brush on machine load distribution Djelovanje zastora granjevine na raspodjelu opterećenja vozila 3.4.1 Peak loads – Vršna opterećenja Following an assessment of machine related variables on loads exerted by the forwarder, we focused Croat. j. for. eng. 33(2012)2

Quantifying the Use of Brush Mats in Reducing Forwarder Peak Loads ... (249–274)

E. R. Labelle and D. Jaeger

Fig. 9 Peak load frequency distributions for varying brush amounts. Numbers shown inside dotted lines indicate the percentage of loads within the respective 10 kN load class and the curves represent a normal distribution Slika 9. Raspodjele učestalosti vršnih opterećenja za različite gustoće zastora granjevine. Brojevi unutar isprekidanih crta označuju udjele opterećenja unutar pojedinih razreda (raspona 10 kN), a krivulje predstavljaju normalne distribucije on analyzing the effect of brush quantity on machine load distribution. As a first analysis, peak load frequency distributions were assessed per brush amount and scenario (Fig. 9). Similar to Fig. 4, percentages of peak loads within 10 kN load classes are shown between dashed lines and normal distribution curves are identified for reference. In the clustered test layout, Croat. j. for. eng. 33(2012)2

frequency distributions increased in the lower peak load classes as brush increased from 5 to 30 kg m–2, and 45 peak loads were below 10 kN. Conversely, frequency distributions of loads greater than 30 kN decreased with increasing brush amount. Specifically, five peak loads exceeded 30 kN with the 5 kg m–2 brush mat and no peak load was beyond 30 kN for the 20, 25, and

263

E. R. Labelle and D. Jaeger

Quantifying the Use of Brush Mats in Reducing Forwarder Peak Loads ... (249–274)

30 kg m–2 brush mats (Fig 9 A). During the transect layout over steel, the percentages of peak loads exceeding 20 kN were 36, 21, and 3% for 10, 20, and 30 kg m–2, respectively (Fig. 9B). Generally, peak load distribution for respective brush amounts was similar between scenarios 1 and 2. Once the forwarder was driven over brush mats placed on the sand covered load test area, a considerable reduction in peak loads was observed. In fact, only 1% of all peak loads exceeded 20 kN with the 10 and 20 kg m–2 brush mats and none were beyond 19 kN with the 30 kg m–2 mat. At the maximum brush amount tested, close to a quarter of the peak loads were below 10 kN. 3.4.2 Mean peak loads and mean sum of peak and second highest loads – Srednja vršna opterećenja i sredina zbroja vršnoga i drugoga najvećega opterećenja As mentioned previously, due to the size of single loading plates, machine positioning could influence peak loads by either having the majority of the wheel load recorded by a single load cell or shared more equally between two or three adjacent load cells. However, for a more comprehensive analysis, this study also observed mean peak loads because of their direct impact on soil physical properties. Mean peak loads decreased with the presence of brush compared to tests done directly over the steel load test area (Table 1). On average, the use of a 30 kg m–2 brush mat lowered mean peak loads by 32% compared to when the equipment was driven directly over the steel load test area (averaging both test layouts over steel surface). A similar trend of reduced load with an increase in brush amount was revealed when analyzing both the absolute maximum peak loads and the mean of all loads exceeding the mean load and two standard deviations (Table 1). We also found that mean sum of peak and second highest loads decreased with an increase in brush amount from 5 to 30 kg m–2 (clustered and transect test layouts over steel surface). The percent reduction of mean peak and second highest loads recorded for the steel load test area varied from 2.5 to 5.7% under a 5 kg m–2 brush mat and from 22.1 to 24.9% underneath a 30 kg m–2 brush mat in relation to the no brush scenario (Table 1). Absolute maximum sum of peak and second highest loads recorded underneath the steel load test area covered with a 5 kg m–2 brush mat ranged from 35.0 to 37.6 kN and from 28.8 to 34.3 kN when it was covered with the 30 kg m–2 brush amount. Even when the forwarder was driven over a 10 kg m–2 brush mat placed on the sand covered load test area, absolute maximum sum of peak and second highest loads ranged between 26.2 and 28.5 kN and varied

264

between 25.6 and 28.1 kN with the 30 kg m–2 brush mat. When combining all replicates, loads from single tires (mean sum of peak and second highest loads from each group of three load cells) from the rear loaded axle recorded during the clustered layout tests decreased from 28.9 kN during the no brush scenario to 22.1 kN for the 30 kg m–2 brush mat, equaling a 23.5% reduction in mean peak load (Fig. 10A). Modifying load cell position from a clustered to a transect test layout provided similar results as mean sum of peak and second highest loads decreased from 29.1 kN with the no brush scenario to 22.0 kN for the 30 kg m–2 brush mat, which translated to a 24.6% load reduction (Fig. 10B). Based on a one-way ANOVA, there was a statistical difference of mean load between 0 and 10 kg m–2 brush mats indicating a beneficial effect of having a minimum of 10 kg m–2 of brush to statistically lower machine loads (Figs. 10A–B). A further increase of brush also lowered mean loads up to the maximum brush amount studied of 30 kg m–2, with statistical differences between each brush amount. During testing with both of these test layouts (Figs. 10 A–B), the mean percentage of the third highest load to the sum of three load cells wide (scenario 1), and the sum of four load cells wide (scenario 2), a so-called half cluster (right ordinate of Fig. 10) increased with additional brush. This result supports the assumption of enhanced lateral load distributing capability of brush mats as brush amounts increase. From our data, we calculated that the addition of a sand layer (20 cm thick) on top of the steel load test area, on average, lowered mean loads recorded below the sand by 25% in comparison to tests done directly over the steel load test area (Fig. 10). Single tires from the rear axle exerted lower mean sum of peak and second highest loads when they were in direct contact with the sand than when the load test area was covered with 10 and 20 kg m–2 brush mats placed on top of the sand (Fig. 10C). This was surprising since we were expecting brush placed on top of the sand layer to further decrease mean loads. However, upon further investigation, we determined that the percentage of the third highest average load to the sum of the half cluster was much higher when the machine was in direct contact with the sand (17%) than when brush was added (11, 11, and 13% for the 10, 20, and 30 kg m–2 brush mats, respectively). A higher percentage here indicates that a higher load is being distributed to the third load cell. Therefore, combining all three highest loads per loading (indicated with dashed lines in Fig. 10C) showed the expected decrease of load from 25.3 kN for the no brush scenario to 21.8 kN for 30 kg m–2 brush amount. Croat. j. for. eng. 33(2012)2

Table 1 Descriptive statistics of peak loads and sum of peak and second highest loads for all tests (rear loaded axle only) Tablica 1. Opisna statistika vršnih opterećenja i zbroja vršnih i drugoga najvećega opterećenja za sva ispitivanja (samo stražnja opterećena osovina) Quantifying the Use of Brush Mats in Reducing Forwarder Peak Loads ... (249–274)

Croat. j. for. eng. 33(2012)2 E. R. Labelle and D. Jaeger

265

E. R. Labelle and D. Jaeger

Quantifying the Use of Brush Mats in Reducing Forwarder Peak Loads ... (249–274)

Fig. 10 Mean sum of peak and second highest loads per brush amount. A different letter indicates a significant difference based on a oneway ANOVA test at the 0.05 probability level. Mean sum of peak, second, and third highest loads are indicated with dashed lines. Right ordinate identified with rhombus corresponds to the percentage of the mean third highest load to half cluster sum (rear loaded axle only) Slika 10. Sredina zbroja vršnoga i drugoga najvećega opterećenja za različite gustoće zastora granjevine. Različita slova označuju značajnost razlike na osnovi jednostruke analize varijance za razinu vjerojatnosti od 5 %. Sredina zbroja vršnoga, drugoga i trećega najvećega opterećenja označena je isprekidanom crtom. Desna ordinata predstavljena podacima u obliku romba odnosi se na udjel sredine trećih najvećih opterećenja u polovici zbroja klastera (samo stražnja opterećena osovina) 3.4.3 Peak surface contact pressures –Vršni površinski dodirni tlakovi To gain further insight on the capacity of brush mats to mitigate the effects of machine traffic, we re-

266

lated ground pressures (obtained by dividing a peak load by the surface area of a single loading plate) to a fixed value, referred to as a dynamic surface contact pressure threshold (DSCPT), thereby making comCroat. j. for. eng. 33(2012)2

Quantifying the Use of Brush Mats in Reducing Forwarder Peak Loads ... (249–274)

E. R. Labelle and D. Jaeger

Fig. 11 Peak ground pressures per brush amount in relation to a 150 kPa dynamic surface contact pressure threshold as identified by vertical dashed line. Percentages indicate the number of surface contact pressures that exceed the threshold. (Rear loaded axle only) Slika 11. Vršni površinski tlak za različite gustoće zastora granjevine u odnosu na granični prag dinamičkoga površinskoga dodirnoga tlaka od 150 kPa (prikazan uspravnim isprekidanim crtama). Postotni udjeli pokazuju broj površinskih dodirnih tlakova koji nadilaze granični prag (samo stražnja opterećena osovina) parisons amongst brush mats, test layouts, and operating surfaces more equitable. Static surface contact pressure thresholds for maintaining soil integrity available from the literature vary from 35 to 80 kPa depending on soil type and soil strength (Olsen and Wästerlund 1989; Owende et al. 2002). However, loads Croat. j. for. eng. 33(2012)2

exerted by the forwarder during testing at the load test platform included dynamic loads and were therefore considerably higher than their static counterpart. To compensate for additional forces (vibratory, rolling, and transient) associated with dynamic testing, we assumed a DSCPT of 150 kPa for a single loading plate.

267

E. R. Labelle and D. Jaeger

Quantifying the Use of Brush Mats in Reducing Forwarder Peak Loads ... (249–274)

We derived the DSCPT of 150 kPa used for one loading plate by applying a multiplication factor of three to a conservative threshold of 50 kPa. The intention was not to determine an exact threshold at which soil properties are negatively affected, but rather compare all

surface contact pressure scenarios to the same reference value. Forwarder peak surface contact pressures were reduced and generally more concentrated (lower variation) as brush amounts increased from 0 to 30 kg m–2 (Fig. 11). When focusing on tests performed

Fig. 12 Mean sum of peak and second highest loads per brush amount and loaded forwarder passes. A different letter indicates a significant difference based on a one-way ANOVA test at the 0.05 probability level per loaded forwarder pass and brush amount (rear loaded axle only) Slika 12. Sredina zbroja vršnoga i drugoga najvećega opterećenja za različite gustoće zastora granjevine i prolaske opterećenoga forvardera. Različita slova označuju značajnost razlike po prolascima forvardera i gustoći zastora granjevine na osnovi jednostruke analize varijance za razinu vjerojatnosti od 5 % (samo stražnja opterećena osovina)

268

Croat. j. for. eng. 33(2012)2

Quantifying the Use of Brush Mats in Reducing Forwarder Peak Loads ... (249–274)

over the steel load test area (no sand) with load cells placed in the clustered test layout, the percentage of recorded surface contact pressures exceeding the 150 kPa DSCPT decreased from 97 to 43% for 0 and 30 kg m–2 brush amounts, respectively. A similar trend was apparent with the transect test layout over steel where 96% of pressures exceeded the DSCPT threshold when the forwarder was operated directly on steel and only 51% when it was operated over the heaviest 30 kg m–2 brush mat (Fig. 11B). Peak surface contact pressures exceeding the 150 kPa DSCPT with the transect test layout over sand, decreased from 54 to 21% for the 10 and 30 kg m–2, respectively (Fig. 11C). However, even with the 30 kg m–2 brush mats, the percentage of peak pressures exceeding the threshold (21%) was slightly greater than when the forwarder was operated directly over sand without any brush (19%).

3.5 Effects of repetitive loadings on load distribution below brush mats – Djelovanje višekratnoga opterećivanja na raspodjelu opterećenja ispod zastora granjevine All results presented so far have combined loadings recorded from all forwarding traffic frequencies per test. However, machine trafficability throughout a cut block is a function of wood location and volume making traffic frequency highly variable ranging from a single cycle to 12 cycles or more near a main landing where timber is being accumulated. To determine the ability of a brush mat to distribute loads over repetitive loadings, we averaged loads recorded during 1–2, 3–6, and 7–12 loaded forwarder passes and identified them as two, six, and 12 passes in Fig. 12. In the majority of cases, mean loads increased slightly with an increase of traffic frequency and based on a one-way ANOVA test were statistically significant between two and 12 loaded passes for brush mats of 15, 20, 25, and 30 kg m–2 (Fig. 12A). However, no statistical difference was found in loads recorded between the different passes for the lighter 5 and 10 kg m–2 brush amounts. We also noted that the differences between load distributing/reducing effects of brush mats of varying quantity decreased as traffic frequency increased from two to 12 loaded passes. This indicates that brush mats of higher amounts (e.g. 30 kg m–2) were most beneficial in reducing mean peak loads at lower traffic frequencies. Nevertheless, even after 12 loaded passes the 30 kg m–2 brush mats showed a higher capability to distribute machine loads than the lighter 10 kg m–2 brush mats after just two passes through a statistically significant reduction of mean loads (p = 0.000, p = 0.009, and p = 0.000, for Figs. 12A, B, and C, respectively). Croat. j. for. eng. 33(2012)2

E. R. Labelle and D. Jaeger

4. Discussion – Rasprava 4.1 Brush mat properties – Značajke zastora granjevine Despite several studies relating the effect of brush mats to changes of soil properties, there is limited literature available on the mechanical properties of branches and none available on the properties of brush mats for mitigating soil disturbances. For effective load distribution, branches used to compose brush mats should have high plasticity and bending strength under loading. Unlike stem wood, branches are composed of reaction wood that is classified as compression or tension wood. Compression wood forms within the lower portion of leaning branches, whereas tension wood is created in the upper portion (Hsu et al. 2005). The modulus of elasticity and tensile strength of compression wood are lower than those of normal wood (Tsoumis 1968). However, compression parallel to the grain and bending strength are higher (Hakkila 1989 in Gurau et al. 2008). Gurau et al. (2008) found that scots pine (Pinus Sylvestris L.) branches (randomly collected from a delimbing operation) had average compression strength of 31.8 MPa, which was 56% less than that of stem wood. Branch wood usually demonstrate higher plasticity than stem wood by bending under compression rather than failing in shearing rupture typically observed for stem wood (Gurau et al. 2008). Overall, branch wood seems to have adequate mechanical properties that would allow brush mats to be effective at lateral and longitudinal load distribution. Generally, a brush mat composed of varying branch sizes and lengths will not react in the same manner to repetitive loadings compared to a single branch. Branches as part of a mat create a reinforced grid through increased internal friction and interlocking effect, thus increasing the overall bending and tensile strengths relative to individual branches. Branch species, through varying density, strength, and moisture content can potentially influence how a specific mat will react to loading forces. In this project, branches from balsam fir and black spruce were used to create brush mats because these are the most common species harvested using a CTL method in New Brunswick, Canada. In this province, softwoods comprise 63% of all Crown land forests (representing 51% of all New Brunswick forests) and from this amount, 80% are classified as balsam fir and spruce (Natural Resources Canada, 2008). This is the main reason why brush mats were composed of branches from balsam fir and black spruce in this study. Furthermore, black spruce is often located on

269

E. R. Labelle and D. Jaeger

Quantifying the Use of Brush Mats in Reducing Forwarder Peak Loads ... (249–274)

sites with wet soils such as low lying areas with a shallow depth to water table where the need of placing a brush mat on the machine operating trail is increased due to the greater risk of soil disturbance. The composition of brush mats (species, diameter and length of branches, and quantity) likely impacts their ability to reduce machine surface contact pressures. A study focusing on softwood and hardwood brush mat load distribution will be presented in a separate journal article. Branches used to create brush mats for tests on the sand covered load test area had an average moisture content of 48.8% (percent green mass), which was 3.3% lower than the average brush used for tests over the steel load test area. Lower branch moisture content would imply that more material is required to reach a target brush amount, thereby increasing the average mat thickness, than with branches of higher moisture content. This trend was observed in our results as higher brush thickness and it was recorded for most brush mats constructed over the sand covered load test area (lower moisture content) than the ones created directly over steel load test area (higher moisture content). This difference in moisture content is mainly related to the length of time required for testing. Brush amounts tested over the sand covered load test area were performed after all brush mats tested directly over the steel load test area had been completed. This study showed that average brush thickness recorded before machine impact (all scenarios combined) was 19 and 40 cm for 10 and 20 kg m–2 brush amounts, respectively (Fig. 8), which is similar to the results obtained by McMahon and Evanson (1994) where 20 and 40 cm thick radiata pine (Pinus radiata D. Don) brush mats corresponded to 9.2 and 18.6 kg m–2, respectively. Brush diameter can also contribute to the efficiency of brush mats in distributing machine loads. According to Jakobsen and Moore (1981), brush mats with branches of 20 cm diameter or more tend to impede the performance of the steel flexible track and can result in increased soil disturbance. Even though we used four diameter classes, one of which was > 6 cm, no branches or tree tops used in this study exceeded 10 cm in diameter, since this is the operational threshold for merchantable volume of most coniferous tree species in Eastern Canada. In fact, more than 60% of all branches used were in the range of 1 to 6 cm diameter. Based on a sample size of 133 branches collected during a commercial thinning of a radiata pine plantation, McMahon and Evanson (1994) recorded an average branch diameter of 3.3 cm (measured 10 cm away from the end of the stem) in material used to create their brush mats. Han et al. (2006) used Douglas-fir (Pseudotsuga menziesii var. glauca) and western hem-

270

lock (Tsuga heterophylla) branches < 7.6 cm in diameter to create brush mats of 7.5 kg m–2. Ideally, we believe that having a combination of branch diameters and lengths would increase the overall stability and efficiency of the mat to distribute the applied loads because of increased friction between branches.

4.2 Effects of brush on machine load distribution Djelovanje zastora granjevine na raspodjelu opterećenja vozila Previous studies focused on determining the effect of machine traffic over brush on soil physical conditions mainly through the assessment of soil mechanical resistance and soil density changes between pre- and post-impact measurements. Even though this study concentrated on loads recorded underneath different brush amounts, our results offered similar trends to what was reported by Han et al. (2006), where a 15 kg m–2 brush mat statistically lowered penetration resistance on a soil of medium moisture condition at a 10 cm depth. Poltorak (2011) also reported benefits of using a 20 kg m–2 brush mat to statistically lower soil density increase (compaction), soil displacement, and soil penetration resistance caused by mechanized forest operations compared to machine traffic directly over bare soil. In a similar study performed by Wronski et al. (1990), it was determined that for every increase of 10 kg m–2 beyond an initial 10 kg m–2 brush density, soil bearing capacity was increased by 25%. In both scenarios of the present study, where brush was in direct contact with steel loading plates, an increase in brush amount lowered the peak loads and also the sum of peak and second highest loads. However, the same could not be concluded for the sand covered scenario since lower mean sum of peak and second highest loads were observed when the forwarder was driven directly over the sand compared to when the sand was covered with 10 and 20 kg m–2 of brush. We assume that the reason for this is the increased surface contact area between both track and tire to the sand in combination with the ability of sand to distribute loads diagonally within the 20 cm thick sand layer. However, expanding the zone of analyses from two to three load cells wide provided similar results (i.e. lower loads as brush increased from 0 to 30 kg m–2) as to when the forwarder was driven over steel rather than sand covered load test area. The amount of brush required to protect forest soils is largely dependent on site characteristics (soil moisture, soil texture, organic content, stand type, etc.) and is therefore difficult to predict (Han et al., 2006). Jakobsen and Moore (1981) suggested that the critical amount of brush required to protect forest soils is 18 kg m–2. Based on the results Croat. j. for. eng. 33(2012)2

Quantifying the Use of Brush Mats in Reducing Forwarder Peak Loads ... (249–274)

obtained from the load test platform, we would recommend leaving a minimum brush layer of 15 to 20 kg m–2 on sensitive sites to lower machine surface contact pressure. Operating heavy equipment on highly susceptible soils (silty clay, clay at high water contents) could require the maximum brush amount tested of 30 kg m–2 or more depending on the number of passes required for timber extraction. It is interesting to explore how much brush is needed to create the recommended brush mat of 20 kg m–2 on sensitive sites and how much biomass would still be remaining for biofuel use. Assuming average softwood stand conditions in New Brunswick with a merchantable volume of 130 m3 ha–1 the total available brush from clear-cut felling operations would amount to 50 green tons per hectare. Approximately 35 green tons per hectare or 70 % of the total brush available would be needed to create a brush mat of 20 kg m–2 covering the entire network of operating trails (average trail width of 3.5 m and spacing of 20 m between adjacent trail centrelines). If site conditions allow for biomass removal and related export of nutrients, this would leave 15 green tons per hectare of forest biomass available for biofuel use.

4.3 Effects of repetitive loadings on load distribution below brush mats – Djelovanje višekratnoga opterećivanja na raspodjelu opterećenja ispod zastora granjevine In forest operations, the frequency of off-road machine traffic is a function of harvested wood volume, its location throughout a cut block, and the locations and capacities of the landings, and can vary between a single cycle to 12 cycles or more near a main landing where wood is being accumulated (Jakobsen and Moore 1981; McDonald and Seixas 1997; Han et al. 2009). For this reason it was of interest to quantify the received loads below varying brush mats during repetitive loadings. The ability of a brush mat to lower loads transferred below the mat was reduced when traffic frequency increased from two to 12 loaded forwarder passes. Other studies in forested settings also showed similar results where initial brush mats provided adequate protection but deteriorated with increasing machine passes, which in turn made them less effective at minimizing soil disturbance from machine traffic (Han et al. 2006; Akay et al. 2007). In particular, Wronski et al. (1990) reported that brush mats less than 10 kg m–2 had a negligible effect on minimizing rut depth following the first two machine passes. As a brush mat was being compacted by repetitive loadings of the forwarder, some branches were sheared off by the machine running gear which decreased the overall strength of the mat. The difference Croat. j. for. eng. 33(2012)2

E. R. Labelle and D. Jaeger

between mean peak loads recorded after two and 12 loaded passes increased with increasing brush amount. Brush mats of higher amounts (20–30 kg m–2) were more efficient in reducing mean peak loads at lower traffic frequencies but their load distributing capacity remained beneficial even after 12 loaded passes. According to a field study by Han et al. (2006), brush mats of 7.5 and 15 kg m–2 were only effective in minimizing the compactive energy of a forwarder (eight-wheel Valmet 890.1 equipped with steel flexible tracks weighing 31,434 kg loaded) for the first two to three loaded passes after which the brush mats deteriorated and no longer provided adequate support. Considering the similarity between the Valmet used in the study by Han et al. (2006) and the Timbco used in our study, we were not expecting such a difference between the ability of brush mats to lower machine impact. However, during the Han et al. (2006) study, over 70% of the machine operating trail was covered with brush greater than 7.6 cm in diameter and the maximum brush amount tested was only 15 kg m–2. As a comparison, only 4% of total branches used in this study were classified as greater than 6 cm in diameter and our maximum brush amount tested was 30 kg m–2.

5. Conclusion – Zaključak In this study we intended to quantify the capability of brush mats to laterally distribute induced loads from forwarder traffic to reduce peak loads received below the mats, which have a high potential to severely disturb forest soils. This was done by analyzing the recorded loads below mats of varying brush quantity by using a load test platform. Softwood brush mats > 10 kg m–2 were proven to be beneficial in statistically reducing peak loads of an eight-wheel forwarder compared to a no brush scenario. However, based on results from the dynamic surface contact pressure analyses, we would recommend leaving 15–20 kg m–2 of softwood brush on sensitive sites and up to the maximum amount available for segments of trails located over highly susceptible soils. We also concluded that increasing traffic frequency from two to 12 loaded forwarder passes caused brush mats to lose some of their ability to distribute the applied loads but still remained beneficial at the highest traffic frequency tested. Besides the use of brush for mats on machine operating trails to lower machine impacts, its alternative use as biofuel will further increase. However, leaving sufficient amounts of brush on machine operating trails is essential for proactively mitigating soil disturbances during mechanized forest operations and needs to remain an integral part of best management practices.

271

E. R. Labelle and D. Jaeger

Quantifying the Use of Brush Mats in Reducing Forwarder Peak Loads ... (249–274)

Acknowledgements – Zahvala This work was financially supported by the Natural Sciences and Engineering Research Council of Canada, the New Brunswick Innovation Foundation, FPInnovations, the University of New Brunswick Research Fund, and the New Brunswick Department of Transportation. Forwarder and operator were generously provided by Debly Forest Services Limited. The authors are grateful for manuscript revisions provided by Dr. Robert J. Rogers from the Mechanical Engineering Department at the University of New Brunswick and Mr. Mark Partington from FPInnovations. Assistance with statistical analyses and programming tasks was obtained from Dr. William Knight from the Mathematics Department at the University of New Brunswick. We are also appreciative of the lab and field assistance obtained from Mr. Marcel Labelle, Mrs. Hélène Rioux, Mr. Benjamin J. Poltorak, and Mr. Scott Fairbairn.

6. References – Literatura Adams, P. W., Froehlich, H. A., 1984: Compaction of forest soils. USDA For. Serv. Res. Pap. PNW-217. Akay, A. E., Yuksel, A., Reis, M., Tutus, A., 2007: The impacts of ground-based logging equipment on forest soil. Polish J. Environ. Stud. 16(3): 371–376. Bettinger, P., Kellogg, L. D., 1993: Residual stand damage from cut-to-length thinning of second-growth timber in the Cascade Range of western Oregon. For. Prod. J. 43(11–12): 59–64. Bock, M. D., Van Rees, K., 2002: Forest harvesting impacts on soil properties and vegetation communities in the Northwest Territories. Can. J. For. Res. 32: 713–724.

Eliasson, L., Wästerlund, I., 2007: Effects of slash reinforcement of strip roads on rutting and soil compaction on a moist fine-grained soil. For. Ecol. Manage. 252: 118–123. Fang, H.-Y., Daniels, J. L., 2006: Introductory Geotechnical Engineering. An Environmental Perspective. Spon text. Taylor and Francis. 545 p. Firestone, 2010: <http://www.firestoneag.com/tirelist. asp?ref=46> (Accessed April 20, 2010). Forristall, F. F., Gessel, S. P., 1955: Soil properties related to forest cover type and productivity on the Lee Forest, Snohomish County, Washington. Soil. Sci. Soc. Am. Proc. 19: 384–389. Froehlich, H. A., 1979: Soil compaction from logging equipment: Effects on growth of young ponderosa pine. J. of Soil and Water Conserv. 34: 276–278. Froehlich, H. A., McNabb, D. S., 1984: Minimizing soil compaction in Pacific Northwest forest. pp. 159–192. In E.L. Stone (ed.) Forest soils and treatment impacts. Proceedings of the 6th North American Forest Soils Conference. Knoxville, TN. Froehlich, H. A., Miles, D. W. R., Robins, R. W., 1986: Growth of young Pinus ponderosa and Pinus concorta on compacted soils in Central Washington. For. Ecol. Manage. 15: 285–294. Gurau, L., Cionca, M., Mansfield-Williams, H., Sawyer, G., Zeleniuc, O., 2008: Comparison of the mechanical properties of branch and stem wood for three species. Wood and Fiber Science 40(4): 647–656. Hakkila, P., 1989: Utilization of residual forest biomass. Springer Verlag, Berlin, Heidelberg, Germany. 568 p. Han, H.-S., Page-Dumroese, D., Han, S.-K., Tirocke, J., 2006: Effects of slash, machine passes, and soil moisture on penetration resistance in a cut-to-length harvesting. Int. J. For. Eng. 17(2): 11–24.

Boussinesq, J., 1885: Application des potentiels à l’étude de l’équilibre et du mouvement des solides élastique. GauthierVillais, Paris. 721 p.

Han, S.-K., Han, H.-S., Page-Dumroese, D. S., Johnson, L., 2009: Soil compaction associated with cut-to-length and whole tree harvesting of coniferous forest. Can. J. For. Res. 39: 976–989.

Bradford, J. M., Peterson, G. A., 2000: Handbook of soil science. Edited by M. E. Sumner. Taylor and Francis. CRC Press. p. 247.

Hsu, Y.-S., Chen, S.-J., Lee, C.-M., Kuo-Huang, L.-L., 2005: Anatomical characteristics of secondary phloem in branches of Zelkova serrata Makino. Bot. Bull. Acad. Sin. 46: 143–149.

Brady, N. C., Weil, R. R., 1999: The nature and properties of soils. Twelfth edition. Prentice-Hall, Inc. United States of America. 881 p.

Jakobsen, B. F., Moore, G. A., 1981: Effects of two types of skidders and of a slash coveron soil compaction by logging of mountain ash. Australian J. For. Res. 11: 247–255.

Brais, S., Camiré, C., 1998: Soil compaction induced by careful logging in the claybelt region of northwestern Quebec (Canada). Can. J. Soil Sci. 78: 197–206.

Jamshidi, R., Jaeger, D., Raafatnia, N., Tabari, M., 2008: Influence of two ground-based skidding systems on soil compaction on different slope and gradient conditions. Int. J. For. Eng. 19(1): 9–16.

Corns, I. G., 1988: Compaction by forestry equipment and effects on coniferous seedling growth on four soils in the Alberta foothills. Can. J. For. Res. 18: 75–84. Craig, R.F., 2004: Soil Mechanics. Seventh edition. Spon press. British Library Cataloguing in Publication Data. 447 p. Eliasson, L., 2005: Effects of forwarder tyre pressure on rut formation and soil compaction. Silva Fennica 39(4): 549–557.

272

Kozlowski, T. T., 1999: Soil compaction and growth of woody plants. Scandinavian J. For. Res. 14(6): 596–619. Labelle, E. R., Jaeger, D., 2011: Soil compaction caused by cut-to-length forest operations and possible short-term natural rehabilitation of soil density. Soil Sci. Soc. Am. J. 75(6): 2314–2329. Croat. j. for. eng. 33(2012)2

Quantifying the Use of Brush Mats in Reducing Forwarder Peak Loads ... (249–274) Liu, C., Evett, J. B., 1992: Soils and Foundations. Third edition. Prentice Hall, Englewood Cliffs, N.J. p. 140–149. Lockaby, B. G., Vidrine, C. G., 1984: Effect of logging equipment traffic on soil density and growth and survival of loblolly pine. South. J. Appl. For. 8: 109–112. Makkonen, I., 2007: PASCAL ground pressure spreadsheet. FPInnovations, Pointe-Claire, QC. <www.feric.ca/pascal-en>. McCarthy, D. F., 2002: Essentials of soil mechanics and foundations. Basic geotechnics. Sixth Edition. Prentice Hall. 788 p. McDonald, T. P., Seixas, F., 1997: Effect of slash on forwarder soil compaction. Int. J. For. Eng. 8(2): 15–26. McMahon, S., Evanson, T., 1994: The effect of slash cover in reducing soil compaction resulting from vehicle passage. LIRO report. Rotorua, NZ. 19(1): 1–8. McNeel, J. F., Ballard, T. M., 1992: Analysis of site stand impacts from thinning with a harvester-forwarder system. Int. J. For. Eng. 4(1): 23–29. Minitab Inc., 2010: Meet Minitab 16. http://www.minitab. com/uploadedFiles/Shared_Resources/Documents/ MeetMinitab/EN16_MeetMinitab.pdf, 22 p. Natural Resources Canada, 2008: State of Canada’s Forest.2008 annual report. National Library of Canada. 44 p. Nugent, C., Kanali, C., Owende, P. M. O., Nieiwenhuis, M., Ward, S., 2003: Characteristic site disturbance due to harvesting and extraction machinery traffic on sensitive forest sites with peat soils. Forest Ecol. and Manage. 180: 85–98. Olsen, H. J., Wästerlund, I., 1989: Terrain and vehicle research with reference to forestry at the Swedish University of agriculture. The Swedish University of agricultural sciences. Department of operational efficiency.Garpenberg. No. 149, 60 p. Owende, P. M. O., Lyons J., Haarlaa, R., Peltola, A., Spinelli, R., Molano, J., Ward, S. M., 2002: Operations protocol for ecoefficient wood harvesting on sensitive sites. Project ECOWOOD. 74 p. Poltorak, B. J., 2011: Mitigating soil disturbance in forest operations. Master of Science in Forest Engineering Thesis. University of New Brunswick, 107 p. Power, W. E., 1974: Effects and observations of soil compaction in the Salem District. USDA BLM Tech. Note. 256: 1–11.

E. R. Labelle and D. Jaeger

Richardson, R., Makkonen, I., 1994: The performance of cutto-length systems in eastern Canada. For. Eng. Res. Inst. of Canada. Tech. Rep. No.TR-109. SPSS, 2007: SPSS Base 16.0 User’s guide. Chicago, IL. 551 p. Steinbrenner, W., 1936: A rational method for the determination of the vertical normal stresses under foundations. Proc. Int. Conf. Soil. Mech. Found. Eng. Cambridge, Massachusetts. Vol. 2. Syunev, V., Sokolov, A., Konovalov, A., Katarov, V., Seliverstov, A., Gerasimov, Y., Karvinen, Y., Valkky, E., 2009: Comparison of wood harvesting methods in the Republic of Karelia. Working Papers of the Finnish Forest Research Institute. 120: 117 p. Taiz, L., Zeiger, E., 1998: Plant physiology. Second edition. Sinauer Associates Inc. Publishers. Sutherland, Massachusetts. 792 p. Tsoumis, G., 1968: Wood as raw material. 1st ed. Pergamon Press Ltd., London, UK., 27 p. Tufts, R. A., Brinker, R. W., 1993: Valmet’s woodstar series harvesting system: a casestudy. Southern. J. Applied. For. 17(2): 69–74. Vepraskas, M. J., 1988: Bulk density values diagnostic of restricted root growth in coarse-textured soils. Soil Sci. Soc. Am. J. 52: 1117–1121. Vidrine, C. G., de Hoop, C., Lanford, B. L., 1999: Assessment of site and stand disturbance from cut-to-length harvesting. Tenth Biennial Southern Silvicultural Research Conference, Shreveport, LA. p. 16–18. Vishay Micro-Measurements, 2011: StrainSmart data acquisition system. Reference and technical guide. 5 p. Voorhees, W. B., Senst, C. G., Nelson, W. W., 1978: Compaction and soil structure modification by wheel traffic in the northern corn belt. Soil Sci. Soc. Am. J. 42: 344–349. Wood, M. J., Moffat, A. J., Carling, P. A., 2003: Improving the design of slash roads used to reduce soil disturbance during mechanised harvesting of coniferous forest plantations in the UK. Int. J. For. Eng. 14(1): 11–23. Wronski, E. B., Stodart, D. M., Humphreys, N., 1990: Trafficability assessment as an aid to planning logging operations. Appita. 43(1): 18–22.

Sažetak

Kvantificiranje uporabe zastora granjevine pri snižavanju vršnih opterećenja i površinskih dodirnih tlakova forvardera Šumska biomasa iz ostatka pri sječi i izradbi drva često se upotrebljava tijekom mehaniziranoga pridobivanja drva za poboljšanje prohodnosti traktorskih vlaka. Takav postupak poboljšanja prohodnosti traktorskih vlaka, ali i smanji vanja dubine kolotraga, osobito je značajan pri pridobivanju drva sustavom harvester – forvarder. Međutim, šumska Croat. j. for. eng. 33(2012)2

273

E. R. Labelle and D. Jaeger

Quantifying the Use of Brush Mats in Reducing Forwarder Peak Loads ... (249–274)

biomasa sve više postaje važan izvor obnovljive energije. Za očuvanje njezine ogrjevne vrijednosti kao biogoriva potrebno je da granjevina (grane, ovršine, ali i lišće) bude čista od mineralnih čestica tla, koje granjevinu onečišćuju pri njezinoj uporabi radi poboljšanja nosivosti tla. Zbog toga se u potpunosti otklanja uporaba zastora granjevine za pobolj šanje prohodnosti trakorskih vlaka jer izravno dodiruje tlo. Uporaba granjevine isključivo kao biogoriva ostavlja trak torske vlake nepokrivene (nezaštićene), što dovodi do značajnih oštećenja šumskoga tla. Za te dvije suprotstavljene na mjene granjevine trebalo bi utvrditi najmanju količinu granjevine potrebne za djelotvornu zaštitu tla, čime bi se omogućila uporaba preostaloga dijela granjevine kao biogoriva. Ovo istraživanje utvrđuje sposobnost zastora granjevine da preraspodijeli opterećenja na tlo smanjujući vršna opte rećenja pri izvoženju drva forvarderom. Kako se navedena sposobnost zastora granjevine povećava, tako raste i učinak zaštite tla. Ciljevi su istraživanja sljedeći: i) kvantificiranje dinamičkih opterećenja forvardera ispod zastora granjevine u usporedbi s opterećenjima bez uporabe zastora granjevine, ii) određivanje najmanje količine zastora granjevine potrebne za ograničavanje djelovanja dinamičkih površinskih dodirnih tlakova na tlo kao određeni prag, iii) kvantificiranje utjeca ja višekratnoga opterećivanja zastora granjevine na njezinu sposobnost raspodjele opterećenja na tlo ispod zastora. Za utvrđivanje razlika među naprezanjima tla zbog vertikalnih opterećenja nije se mogao upotrijebiti uobičajeni sustav vaga zbog dimenzija i mase vozila, ali i značajki grana korištenih za izradu zastora granjevine. Stoga je naprav ljena čelična testna naprava (ispitna platforma) po kojoj je omogućeno kretanje teških forvardera, gdje se utvrđivanje opterećenja tla vozilom mjeri pomoću ugrađenih dinamometara velikoga mjernoga raspona. Raspoređeni dinamometri neovisno mjere opterećenje vozilom s rezolucijom 30 cm × 30 cm. Ukupno je provedeno 15 inačica ispitivanja s osmo kotačnim forvarderom Timbco TF820-D za analizu razlika u vršnim opterećenjima, izmjerenim ispod površine zastora granjevine gustoće 5, 10, 15, 20, 25 i 30 kg m–2 (u svježem stanju) koja je smještena na površinu ispitne naprave. Po svakom zastoru granjevine obavljeno je 12 prolazaka forvardera, ukljućujući neopterećeno i opterećeno kretanje. Ključni rezultati ovoga istraživanja pokazuju da: 1) najmanja količina zastora granjevine od 10 kg m–2 potrebna je za statistički niža srednja vršna opterećenja izmjerena ispod zastora granjevine u usporedbi s opterećenjima bez uporabe zastora granjevine (kretanje forvardera po čeličnoj ispitnoj platformi), 2) gotovo pola od svih vršnih opterećenja nadilazi 20 kN kada se forvarder kretao po ispitnoj čeličnoj platformi, a samo 4 % pri kretanju vozila kada je platforma bila prekrivena zastorom granjevine gustoće 30 kg m–2, 3) kretanje opterećenoga forvardera (s teretom mase 6689 kg) po zas toru granjevine gustoće 30 kg m–2 smanjilo je vršna opterećenja za 30 % u odnosu na kretanje forvardera po čeličnoj ispitnoj platformi, 4) pri kretanju forvardera po ispitnoj platformi bez zastora granjevine (0 kg m–2) utvrđeno je da 97 % svih vršnih dodirnih tlakova nadilazi granični prag od 150 kPa, 5) najveće je smanjenje vršnih opterećenja uočeno tijekom prva dva prolaska forvardera nakon kojih zastor granjevine i dalje smanjuje vršna opterećenja, ali znatno manje. Ključne riječi: biomasa, zastor granjevine, šumska vozila, površinski dodirni tlak, zaštita tla

Authors’ address – Adresa autorâ: Eric R. Labelle, PhD. Candidate e-mail: e.r.labelle@unb.ca Faculty of Forestry and Environmental Management University of New Brunswick PO Box 4400 Fredericton, New Brunswick, E3B 5A3 CANADA Assoc. Prof. Dirk Jaeger, PhD. e-mail: jaeger@unb.ca e-mail: dirk.jaeger@fobawi.uni-freiburg.de Faculty of Forestry and Environmental Management University of New Brunswick CANADA

Received (Primljeno): June 15, 2012 Accepted (Prihvaćeno): August 5, 2012

274

and Institute of Forest Utilization and Work Science University of Freiburg GERMANY Croat. j. for. eng. 33(2012)2

Original scietific paper – Izvorni znanstveni rad

A GIS Approach to Analyzing Off-Road Transportation: a Case Study in Sweden Sima Mohtashami, Isabelle Bergkvist, Björn Löfgren, Staffan Berg Abstract – Nacrtak Off-road driving in logging operations began in Sweden in the 1960s. Those operations took place at final fellings during winter on prepared ice roads that protected the soil and miti gated possible soil damage. Today logging operations are fully mechanized and performed all year round. Thus forest strip roads may suffer severe impacts from off-road operations. Soil disturbances may have physical, chemical, biological, hydrological, and economic effects and affect the water quality. Similar problems are encountered in other regions, where driving occurs close to watercourses or vulnerable areas. The EU Water Directive has an impact on operations in forests, creating an incentive for improvements. Ongoing efforts in the design of vehicles and equipment are likely to improve operations. Soil damage can be avoided by applying GIS-based planning techniques, and by taking advantage of soil radar-scanned and ground laser-scanned data sets, which would facilitate safer off-road driving to a great extent. A case study in southern Sweden revealed that the use of digital planning for the improvement of strip roads in order to avoid vulnerable terrain made forwarding of timber more profitable. Using elevation, slope, aspect and soil type digital layers, a model has been created in ‘model builder’ environment of ArcGIS to build up a cost-index surface, which classifies the terrain suitability for driving into five different levels. Implementing distance analysis, the model designs the least costly roads connecting any desired destination to the landing point. The result of this study reveals that this kind of pre-planning tool can mitigate ecological dam ages to soil and water and at the same time it can also assist decision makers to evaluate dif ferent possible choices of road layouts regarding preserving sensitive regions in forest lands. Keywords: GIS-based decision support system; Digital Terrain Model (DTM); ground dam age; forest operations; forwarding; planning; rutting; soil

1. Introduction – Uvod The mechanization of logging operations in Sweden and other countries began in the 1960s, and it led to off-road driving in order to bring timber to the landing. Those operations took place at final fellings during winter on prepared ice roads, thus protecting the soil and mitigating soil damage. Today, all logging operations are mechanized and performed throughout the year due to the all-year-round fresh timber demands of the pulp industry. Off-road driving is extensive on all forest lands during all seasons, even when the ground is vulnerable to soil disturbance. Any professional forester has to consider the great variety in the Swedish landscape in terms of topography, soil types and surface water. Croat. j. for. eng. 33(2012)2

In the 1970s, a terrain classification system was developed to aid the planning of off-road driving in conjunction with forest operations (Forskning stiftelsen Skogsarbeten 1969 and Skogforsk 1992). Recent guidelines (Ring et al. 2008) addressed the impacts of offroad driving from a water-quality perspective. Offroad driving close to surface water increases the surface water sediment load risk. Such sediments might be harmful to aquatic organisms (Skogforsk 2008). The guidelines allocate a 5–10 m zone bordering on lakes and streams in order to mitigate sediment release into water, avoiding driving on streams and wet areas, and utilizing technical devices in order to reduce physical soil disturbance at crossings. Planning is another important tool for reducing such negative impacts.

275

S. Mohtashami et al.

A GIS Approach to Analyzing Off-Road Transportation: a Case Study in Sweden (275–284)

The aim of this paper is to; Þ elucidate the risky aspects of road driving in forest operations with respect to current forms of legislation. Þ investigate ways of mitigating soil damage by trying out a planning decision support system in a case study in South-Eastern Sweden. This decision support system facilitates routing in terrain considering soil, water and restricted areas. Proper route alignment for avoiding probable sliding of the loaded forwarders on steep slopes was considered as part of this method. Þ estimate possible financial gains by evaluating different route alternatives at a felling site.

1.1 Environmental and social significance Okolišne i socijalne značajke Primary soil damage in conjunction with off-road driving may have secondary effects that cause physical, chemical, biological, hydrological, economic or aesthetic impacts. Although not specifically regulated in the environmental standards for forestry, the event of forest certification manifested in schemes like PEFC (2012) or FSC (2012) has stressed the importance of disturbance control in connection with off-road driving. One reason for this is the less permissive attitude to soil disturbance in Central Europe (Hauk 2001; Hildebrand and Schack Kirchner 2002). Off-road operations increase soil density down to 50 cm, and decrease soil aeration and thereby reduce root penetration (Eliasson and Wästerlund 2007). This impact varies with moisture content (Ziesak 2003; Yavuzcan et al. 2005). Rutting may result in compaction (Jamshidi et al. 2008) and adverse driving conditions, leading to costly interruptions and breakdowns, which eventually increase the energy use and related emissions. Reduction of soil disturbances at off-road operations can be beneficial both for the environment and for operational cost reduction. Physical soil disturbances may affect ecosystem pools of C and N in the soil (Finér et al. 2003). European environmental policies stipulate that processes in agriculture and forestry (e.g. draining, offroad driving and harvesting) which in general reduce C storage, should be avoided (Anon. 2004 and 2006). Increased nitrate leaching is commonly found after clear cutting and soil scarification (Ring et al. 2008). Logging tracks often induce similar disturbances to the soil and thus there is a risk of elevated N-mineralization and denitrification in these areas. Final fellings might result in the discharge of Hg and its consequent accumulation in fish (Bishop et al. 2009). This might be attributed to anoxic conditions in the soil caused by the raised water level in tracks.

276

Swedish environmental objectives (Swedish EPA 2011) regulate several impacts likely to be caused by soil damage in conjunction with forest operations and off-road driving. Under the EU Water Framework Directive (Anon. 2000), there is also a legal responsibility to maintain the water status. The generally anticipated process leading toward global warming (Peters 1990) is likely to affect the frequency of rainfall, droughts and what is nowadays called extreme weather, especially in Europe (Bolte et al. 2009). The authors believe that similar legal and consequent political pressure will affect forest planning and forest operations.

1.2 Technical means for mitigating ground damage – Tehničke mjere za ublažavanje oštećivanja podloge Ground damage is caused when forces and pressures are exerted on the ground surface via the wheels or tracks of terrain vehicles. The resultant effect is compaction of the soil, skidding, and shearing of vegetation or soil layers. These effects can technically be avoided either by the design of the machine/vehicle or mitigated via operational skills, for example adjusting vehicle properties by reducing the impact of the load on the ground (Ziesak 2003; 2004) or the use of ancillary equipment (Staland & Larsson 2002) along planned routes to enable passages over brooks or other watercourses. Technical improvements in laser scanning have also provided quite precise data layers representing terrain surfaces in the form of Digital Terrain Models (DTM). These DTM layers, especially high resolution ones, have been quite attractive in supporting forestry operations in recent years since they provide thorough and detailed information about terrain topography (i.e. terrain elevation and steepness), which in turn are used to choose the best skidding system in complex forest fields (Lubello 2008; Vega et al. 2009). Krč and Košir (2008) have also used Digital Elevation Models (DEM) to develop a model for terrain classification based on the best predicted skidding direction on steep terrain. Benefiting from DEM along with other inventory information about the rockiness and stoniness of a terrain, Mihelič and Krč (2009) analyzed how to define new skidding systems or forwarding possibilities on different terrain classes in Slovenia. High resolution DTMs are also used to define various soil wetness indexes like the Depth to Water Index, DTW (Murphy et al. 2008), and Compound Topographic Index (CTI) (Goetz 2010), to predict vegetation terrain types and ground bearing capacity in forest lands which are of great help for planning silvicultural activities. Croat. j. for. eng. 33(2012)2

A GIS Approach to Analyzing Off-Road Transportation: a Case Study in Sweden (275–284)

1.3 Design of machines – Konstrukcija strojeva Via machinery design, basic properties of a machine impact can be altered and adjusted to actual conditions. This can be done by adjusting the wheel pressure exerted on the soil through changing the tire pressure (wheel width) or using wider tires (Jonsson 2011). Tracks might be added in order to distribute forces more evenly over the ground surface, which itself will enhance the risk of shearing at turns. By adjusting the air pressure in the tires to the soil and the load or vice versa, the operator has the means of reducing ground damage (rut depths). Tests made with a CTI-system (Löfgren 1994) on a forwarder showed that the rut depth of 600 mm – wide tires with low pressure is the same as that of 800 mm – wide tires and high pressure. Basic machine properties have an impact on the ability to negotiate terrain. Positioning of the wheels before driving over obstacles using hydrostatic driving may be beneficial. The reduction and damping of vibrations has a similar effect (Baez 2008). Geometric design is the determining factor for a beneficial distribution of pressure on the wheels.

S. Mohtashami et al.

evant knowledge, evaluate possible alternatives and, finally, make appropriate decisions. A number of researchers have assessed the trafficability of terrain types for different goals with the aid of GIS. Initially, GIS was used for military off-road planning (Lubello 2008), and gradually it was introduced in the fields of agriculture and forestry. In most cases, GIS has been used for choosing the optimal routes out of a number of already-existing possible networks: Rongzu and Mikkonen (2004) used GIS to provide an optimized wood logistics GIS model based on a combined cost surface created from road transport costs and off-road transport cost surfaces. Pentek et al. (2005) used GIS to analyze the quality and quantity of an existing forest road network to determine potentials for planning future routes. Suvinen (2006) used a GIS-based simulation model to assess terrain tractability regarding two sets of constant and dynamic factors as well as machine characteristics to suggest a proper route layout for different load/terrain conditions. However, lateral inclination, which is an essential factor for properly guiding a machine on steep terrain, was neglected in that study.

1.4 Supporting equipment – Pomoćna oprema With the aid of different sorts of technical equipment, it is possible to reduce the damage to virtually nil; however, there is a cost and the issue is to have the required equipment on the right spot at the right occasion. The means for doing this are fixed or temporary bridges along the hauling route, or carrying prefabricated bridges for immediate use, for example when crossing a ditch. Other solutions are to work with ground cover rigs made of timber or tire mats. Moreover, the use of harvest residues and downgraded wood logs constitute other possible means. Residues or straw can be spread out along the hauling route in order to mitigate the ground damage when passing over sensitive spots (Eliasson 2007; Saunders and Ireland 2005). Consequently, it is important to plan the harvesting of residues, considering where the residues are needed to improve the bearing capacity of the ground, and where they could be harvested as forest fuel.

1.5 Planning – Planiranje The issue of the right application of routes and ancillary equipment, as well as where the residues should be used on the forest road, is coordinated by planning. In order to plan the operations and utilize the ancillary equipment, updated maps are mandatory. Planning based on the Geographic Information System (GIS) is a great step forward compared to former methods as it is feasible to explore a variety of digital layers of information, extract the required relCroat. j. for. eng. 33(2012)2

2. Materials and methods – Materijal i metode 2.1 Study area and scenario definition – Područje istraživanja i definiranje scenarija The study area under consideration was the property of Selesjö in Östergötland, located in South-Eastern Sweden, and it was mainly dominated by Norway spruce and Scots pine, Fig. 1. In the close vicinity of the harvesting site, with an area of 6.72 ha, a possible landing point was selected, where the harvested timbers were to be stored for further operations. This landing point was to be reached from 4 arbitrary destination points at the harvesting site. However, there was a wetland between the landing point and part of the stands, which needed to be protected against driving damage. This area as well as other existing sensitive parts, like ditches and streams, were called ‘No Go’ areas and were totally excluded for the purpose of locating the routes. Possible route alignment to reach the destination points with minimum disturbances to the surrounding environment was evaluated under two different scenarios: in Scenario 1, the routes were expected to go beyond the wetland and reach the landing point, while in Scenario 2 the possibility of building a corduroy road to pass the wetland was analyzed to see how the route layout would have to be adjusted to the new conditions and how much it could reduce the cost of transportation by providing shorter route stretches.

277

S. Mohtashami et al.

A GIS Approach to Analyzing Off-Road Transportation: a Case Study in Sweden (275–284)

Fig. 1 Location of the study area on a map of the world (left) and in Sweden (right) Slika 1. Područje istraživanja na karti svijeta (lijevo) i u Švedskoj (desno)

2.2 Input data and software – Ulazni podaci i softver Data layers used in this model were as follows: a high resolution, 0.5 × 0.5 meter Digital Terrain Model (DTM), FORAN SingleTree® Laser Method, which are both the products of high density, 8–10 points/m2, and laser scanning of the area using Foran Remote Sensing AB. The former layer was in raster format and the latter was in point format, containing information about tree species, crown diameter, stem diameter and the gross volume of stands. Slope and Aspect grid layers were extracted from the DTM and were used to evaluate the topography of the study area. All the grid materials were reclassified to a coarser resolution of 4 × 4 to be sure that each pixel can support the width of forwarders. Soil types, in polygon format, provided by the Swedish Geological Research Institute (SGU), represented different textures of the soil in the area. Environmentally-sensitive spots such as nature reserves, key biotopes and habitat-protected regions as well as historical values, were prepared by the Swedish Forest Agency, or Skogsstyrelsen (2011); these areas were to be set aside as protected. Separate shape layers localizing the landing point (source) and the destinations were other inputs in the model. In this study, the 10th version of the ArcGIS software packages provided by the Environmental Systems Research Institute, Inc. (ESRI 2012), including ArcMap, ArcCatalog were used for data preparation, data processing, information exploration, evaluation and, ultimately, for viewing the final results. The Slope, Aspect, Path Distance and Cost Path tools available from the

278

Spatial Analysis and the 3D Analysis extensions were used to build up the desired model within the ‘Model Builder’ environment of ArcGIS.

2.3 Procedure of the analysis – Postupak analize Planning sustainable forestry operations requires simultaneous consideration of the economic and ecological values in the forest. These two aspects do not always introduce similar approaches for forest managers in practice and consequently there is always an essential need to reach a consensus among all the stakeholders regarding evaluating and integrating various criteria and making the best decision. Eastman et al. (1998) defined it so simply: »decision is a choice between alternatives«, and Multi-Criteria Decision Analysis (MCDA) is a procedure that can unify several attributes and/or objectives as part of the decisionmaking process (Malczewski 2006). Therefore, this procedure formed the basis of the analysis for finding the optimal routes in this study. Weighted Linear Combination (WLC) was the rule applied in this process as it was compatible with the ArcGIS software. Applying MCDA, elevation, slope and soil types were regarded as the most determining factors for estimating different levels of suitability of the area for driving. Since soil bearing capacity for supporting massive forest machinery has a direct relation with the degree of soil moisture, it has been assumed that the lower the elevation in an area, the higher the probability of having wetness in soil would be, and thus the worse the ground conditions for driving would be. Following this assumption, the elevation layer was used as Croat. j. for. eng. 33(2012)2

A GIS Approach to Analyzing Off-Road Transportation: a Case Study in Sweden (275–284)

Table 1 Summary of data reclassification for cost-index surface preparation Tablica 1. Zbirni prikaz razredbe podataka za pripremu vrijednosnih indeksa površine Factors

Factor classification – Razredba faktora Original values

Cost-index values

Izvorne vrijednosti

Vrijednosti indeksa

65–60

60–55

55–50

50–46

0–6

Slope

6–11

Nagib

11–18

(0–90 degrees)

18–27

27–90

Rocks-outcrop

Till

Silt

Faktori

Elevation Visina (46–65 metres)

Soil classes Tipovi tla

a simple model to identify the risky wet parts in the area. Slope layer was used to quantify the terrain steepness and to avoid driving on steep terrains (slope > 18 degrees). The soil type layer was used to find water courses, wetlands and similar sensitive parts in the area. These three layers were reclassified to a new scale of 1 to 5, called a cost index, in order to have a common scale for defining the suitability of the ground for terrain driving on all layers. The better the driving conditions, the lower the assigned cost index value to the corresponding class in each of the data layers was, Table 1. For example the elevation values in this area ranged between 46 and 65, and therefore it was reclassified so that the values between 46 and 50 got the cost index (5), elevations between 50 and 55 got the cost index (3), elevations between 55 and 60 got the cost index (2), and the highest part with elevations between 60 and 65 got the minimum cost index (1). Finally, in order to integrate all these input layers into a single cost-index layer with 5 levels of suitability, different weights of importance were assigned to them based on ideas from a panel of 7 experts at the Forestry Research Institute of Sweden; Skogforsk. Elevation resulted in a weight gain of 50%, since in this case it had much better precision compared to the soil type layer for predicting where the moist soil texture could be located. Flat and low elevated areas are assumed to be wet and unsuitable for driving. The soil type layer gained 20% in terms of the Croat. j. for. eng. 33(2012)2

S. Mohtashami et al.

weights and was used as a complementary layer to find the areas with the best bearing capacity, and finally Slope gained the remaining 30% in order to avoid technical problems under operational conditions. Later on, soil classes with unsuitable bearing conditions (such as wetlands, peat lands), as well as steep slopes (> 18 degrees) and ditches were regarded as constraints on the study area and were extracted from the cost-index surface by assigning »No Data« to their values and visualized with the darkest grey color in Fig. 2 and Figure 3. Afterwards, feeding the cost-index surface as a cost raster into the Path Distance tool together with the landing point layer, as the source, with DTM layer as the surface raster and Aspect as the horizontal factor into the Path Distance tool, the least accumulative cost of getting back to the landing point, raster distance, and also the proper direction of moving to the neighboring cell, backlink raster, was determined at this stage. A maximum inclination of 5 degrees with respect to the slope direction of the ground was meant to be achieved for the route layouts. This was implemented by applying the horizontal factor parameters in the Path Distance tool. The Aspect layer, defining the direction of the slope of the ground, was used as the input horizontal raster. The horizontal factor was set as a table type in ASCI format. This table consists of two columns; the first one is called the Horizontal Relative Moving Angle (HRMA) and defines the relationship of the moving direction with respect to the horizontal direction of the terrain, while the second column is called the Horizontal Factor (HF), and defines the difficulty of moving from one cell to another (ESRI 2011). In this case, all the HRMA between 5 and 175 degrees, which indicate uphill or downhill movement with too much tilting, were assigned very high HF (100), while for other HRMA (0 ≤ HRMA ≤ 5 degree or 175 ≤ HRMA ≤ 180) that would not cause too much tilting on the terrain, the HF varied linearly between a value of 1 to 5; the smaller the tilting, the lower the assigned HF was. The formula applied to calculate the values of the raster distance in Path Distance tool is (ESRI 2011): Þ For perpendicular movement: Cost_distance = Cost_Surface * Surface_distance * {[Friction(a)* Horizontal_factor(a) + Friction(b) * Horizontal_factor(b)] / 2} * Vertical_factor Þ For diagonal movement: Cost_distance = Cost_Surface * Surface_distance * 1.414214 * {[Friction(a) * Horizontal_factor(a) + Friction(b) * Horizontal_factor(b)] / 2} * Vertical_ factor The outcomes of this part aligned with the destination layer were inserted into the Cost Path tool to de-

279

S. Mohtashami et al.

A GIS Approach to Analyzing Off-Road Transportation: a Case Study in Sweden (275–284)

sign the route layout in the harvesting site. As explained earlier, a wetland was located in the way of connecting the destinations within the site to the landing point, outside the harvesting border. In Scenario 1, the model was supposed to plan the routes by going beyond this restricted part, while in Scenario 2, a strip with the minimum cost-index (1) was added to the cost-index surface, over the wetland to a corduroy road that could be built on the wetland and contribute to a different route layout in the field.

The model suggested the routes with the lowest cost index within the context of two different scenarios (Fig. 2 and Fig. 3). Both of these two route alignments are promising for reducing soil and water disturbances by suggesting the routes on the lowest cost-index values (i.e. the best driving conditions), while compensating for the surface distance and slope directions. Moreover, this model gives decision mak-

ers the opportunity of comparing the route alignments in two different scenarios, with or without a corduroy road, in order to find the most appropriate course of action. In this case, it was conceived that following the route design in Scenario 2 and building a corduroy road on the preserved wetland could contribute to a reduction of almost 700 m in the length of the routes to be passed from the destination points to the landing point. Using the FORAN SingleTree® Laser Method layer, the standing volume in the southern and central part of the area was measured as almost 2 526 m3. This is actually the timber volume on the part that would be connected to the harvesting point through the corduroy road. The following table describes the equivalent volume in solid over bark and in tons, Table 2. Thus, having almost 1 573 tons of timber at this site, and assuming the maximum possible load of large (PONSSE ElephantKing) forwarders to be 20 tons, the number of loaded forwarders required for collecting the timber is 79, which would result in 158 (79×2) passages of the mentioned forwarder over the terrain.

Fig. 2 Least costly routes suggested by the model for Scenario 1 Slika 2. Izvozni pravci predloženi modelom za inačicu 1

Fig. 3 Least costly routes suggested by the model for Scenario 2 Slika 3. Izvozni pravci predloženi modelom za inačicu 2

3. Results – Rezultati

280

Croat. j. for. eng. 33(2012)2

A GIS Approach to Analyzing Off-Road Transportation: a Case Study in Sweden (275–284)

Table 2 Timber stands described as m3 standing and volume solid over bark and mass as metric tons Tablica 2. Sastojinske značajke; drvo na panju u m3, obujam s korom i masa u tonama m3 standing

m3 solid over bark

Tons

Drvo na panju u m

Obujam s korom u m3

Masa u tonama

2 526

2 097*

1 573**

* Conversion factor: m3 solid over bark/m3 standing = 0.83 ** Conversion factor: tons/ m3 solid over bark to = 0.75

Based on the experts’ ideas at the Forestry Research Institute of Sweden, it has been assumed that the maximum velocity of a large forwarder is 0.8 m/s and its average operational cost is 85 Euro/hour. Thus, the second route layout over the corduroy road would result in a reduction of EUR 3,200 for the whole forwarding operation. The estimated cost of constructing the corduroy road was EUR 500, according to a panel of experts, which is less than the operational cost saving in Scenario 2 and therefore makes it more profitable.

4. Discussion and Conclusion – Rasprava i zaključci Anticipated changes due to global warming and international agreements require well-considered planning of forest operations. The negative chemical, biological and physical consequence of soil damage is proven (e.g. Finér et al. 2003; Bishop et al. 2009). As a result of legislation and forest certification, it is important to show that operation managers have identified this aspect, and actions have been undertaken in order to remedy or improve deviations from the standards, which means that the operations must be close to best practices. Means are available for mitigating damage, namely equipment for crossing water streams and wetlands, or tracks to mitigate rutting, but their successful use depends on access to relevant terrain information. Not just any equipment will be used when the right equipment is not available when needed. The advent of better pre-planning tools with the aid of GIS can facilitate that. High-resolution digital terrain models generated from laser scanning of the forest lands have improved the task of planning by providing comprehensive details about the terrain structure e.g. elevation, slope, etc. The digital maps in general use, with information about factors such as wetland areas or other objects of concern to be safeguarded, can ensure improved plans for achieving sustainable forestry in practice that are Croat. j. for. eng. 33(2012)2

S. Mohtashami et al.

probably more economically rewarding and are likely to be asked for by planners, decision-makers or auditors from any certification agency. The case investigated in this study demonstrated that a practical application of available digital information and models for the planning and construction of alternative shorter and better routes in fact resulted in improved profitability of timber forwarding. Applied in a wider context, such improvements might result in substantial monetary savings and less disturbance to soil and water. The impact any vehicle has on soils is influenced by its basic design, its wheels, and its load. The damage caused is a combination of the driving and planning applied. Improvements in machine properties will take a long time before they have an effect on the fleet of logging machines, and some will be more effective than others. Planning tools will have an immediate effect and will enable better allocation of logging machines to appropriate logging areas.

Acknowledgement – Zahvala This research was performed at the Swedish Forestry Research Institute, Skogforsk, and the authors would like to thank the program managers Johan Sonesson and Magnus Thor for their valuable support. Dr. Wayne McCallum is also acknowledged for his help in checking the authors’ English. The editor and reviewers are thanked for their comments and recommendations.

5. References – Literatura Anon. 2000: Directive 2000/60/EC of the European Parliament of the Council of 23 October 2000 establishing a framework for community action in the field of water policy. L 237/1 EN Official Journal of the European Union 22/12/2004. Anon. 2004: Directive 2004/35/EC of the European Parliament and of the Council of 21 April 2004 on environmental liability with regard to the prevention and remedying of environmental damage. L 143/56 EN Official Journal of the European Union 30/4/2004. Anon. 2006: Proposal for a Directive of the European Parliament and of the Council establishing a framework for the protection of soil and amending Directive 2004/35/EC.COM (2006) 232 final 2006/0086 (COD). Commission of the European Communities Brussels, 22/9/2006. Bishop, K., Allan, C., Bringmark, L., Garcia E., Hellsten, S., Högbom, L., Johansson, K., Lomander.A., Meill, M., Munthe, J., Nilsson, M., Porvari, P., Skyllberg, U., Sörensen, R., Zetterberg, T., Åkerblom, S., 2009: The Effects of Forestry on Hg Bioaccumulation in Nemoral/Boreal Waters and Recommendations for Good Silvicultural Practice. AMBIO 38(7): 373– 380.

281

S. Mohtashami et al.

A GIS Approach to Analyzing Off-Road Transportation: a Case Study in Sweden (275–284)

Bolte, N., Ammer, C., Löf, M., Nabuurs, G.-J., Schall, P., Spathelf, P., 2009: Adaptive Forest Management: A Prerequisite for Sustainable Forestry in the Face of Climate Change. Sustainable Forest Management in a Changing World. Managing Forestry Ecosystems 2009, Volume 19: 115–139. DOI: 10.1007/978-90-481:3301-7_8p. Dynamisk däcksmodellering och markinteraktion för skogsmaskiner (Dynamic tyre modelling and soil interaction regarding forestry machines) (2011), Editor: Jonsson, J., Arbetsrapport 745, Skogforsk (Forestry Research Institute of Sweden). Eastman, J. R., Jiang, H., Toledano, J., 1998: Multi-criteria and multi-objective decision making for land allocation using GIS. In: Multi-Criteria Analysis for Land Use Management, Beinat, E., Nijkamp, P., Kluwer Academic Publisher, Netherlands, p. 227–247. Eliasson, L., Wästerlund, I., 2007: Effects of slash reinforcement of strip roads on rutting and soil compaction on a moist fine-grained soil. Forest Ecology and Management, 252(1-3): 118–123. Esri, 2008: http://www.esri.com/software/arcgis/arcgisonline/whats-new/index.html, Accessed 2012-02-16. Finér, L., Mannerkoski, H., Piirainen, S., Starr, M., 2003: Carbon and nitrogen pools in an old-growth, Norway spruce mixed forest in eastern Finland and changes associated with clear-cutting. Forest Ecology and Management 174: 51–63. FSC, 2012: Forest Stewardship Council: http://www.fsc.org/ index.htm. Accessed 2012-06-27.

Löfgren, B., 1996: CTI för terrängtransporter i skogsbruket (CTI for terrain transport in forestry) (1996), Editor: Löfgren, B., Resultat nr. 25, Skogforsk (Forestry Research Institute of Sweden). Lubello, D., 2008: A rule-based SDSS for integrated forest harvesting planning. Doctoral Dissertation, Padua: Università degli Studi di Padova; 213 p. Malczewski, J., 2006: GIS-based multicriteria decision analysis: a survey of literature, International Journal of Geographic Information Science 20(7): 703–726. Mihelič, M.; Krč, J: 2009: Analysis of inclusion of wood forwarding into a skidding model, Croatian Journal of Forest Engineering, 30(2): 113–125. Murphy, P. N., Ogilvie, J., Meng, F. R., Arp, P., 2008: Stream network modelling using lidar and photogrammetric digital elevation models: a comparison and field verification. Hydrological Processes 22(12): 1747–1754. PEFC, 2012: Programme for the Endorsement of Forest Certification. http://www.pefc.org/. Accessed 201210627. Pentek, T., Pičman, D., Potočnik, I., Dvorščak, P., Nevečerel, H., 2005: Analysis of an existing forest road network. Croatian Journal of Forest Engineering 26(1): 39–50. Peters, R. L., 1990: Effects of Global Warming on Forests. Forest Ecology and Management, Volume 35 Issues 1-215 June 1990: 13–33. doi.10.1016/0378-1127(90)90229-5. Rongzu, Q., Mikkkonen, E., 2004: GIS-based decision support system for wood logistics. Forestry Studies in China. 6(4): 29–33.

Fredricsson, J., Dahlin, B., 1995: Computer–Assisted Forest Road Planning – A Proposed Interactive Model With Spatial Emphasis on Private Forest Land. Journal of Forest Engineering 6(2): 35–39.

Saunders, C., Ireland, D., 2005: Extraction Route Trials on Sensitive Sites. Technical Note. Forest Research December 2005.

Goetz, H., 2010: Using a topographic index to define terrain types. In: Mapping Forestry; First Edition, 4P. ESRI Press, California 2010.

Skogsbruk och vatten – En kunskapsöversikt (Forestry and Water: A Review) (2008), Editor: Ring, E., Löfgren, S., Sandin, L. Högbom, L., and Goedkoop, W, Redogörelse nr 3. Skogforsk (Forestry Research Institute of Sweden).

Hauk, B., 2011: Aspekte des Bodenschutzesbei der Entwicklung, beim Kauf und dem Einsatz von Forstmaschinen (Soil protection aspects in the development, purchase and use of forestry machines; author’s translation) Editor: Hauk, B., Forsttechnische Informationen. Fachzeitung für Waldarbeit und Forsttechnik 4 (2001). Hildebrand, K.-E., Schack Kirchner, H., 2002: The influence of compaction on soil structure and functions in forest sites. In: Modern Trends in Applied Terrestrial Ecology, Kluwer Academic/Plenum Publishers, United States of America, p. 1–11. Jamshidi, R., Jäger, D., Raafatnia, N., Tabari, M., 2008: Influence of two Ground-Based Skidding Systems on Soil Compaction under Different Slope and Gradient Conditions. International Journal of Forest Engineering 19(1): 9–16. Krč, J., Košir, B., 2008: Predicting wood skidding direction on steep terrain by DEM and forest road network extension, Croatian Journal of Forest Engineering 29(2): 177–188.

282

Skogsbruk med hänsyn till vatten (Forestry with consideration to water – guidelines; author’s translation) (2008), Handledning. Skogforsk (Forestry Research Institute of Sweden). Skogsstyrelsen (2011): http://www.skogsstyrelsen.se/Agaoch-bruka/Skogsbruk/Karttjanster/Skogens-Parlor-/.Accessed2011_12_23 23. Staland, F., Larsson, K., 2002: Bra planering och rätt teknik minskar risken för markskador (Proper planning using the right techniques reduces risks for ground damage), Editors: Staland F., Larsson K., Resultat nr 4. Skogforsk (Forestry Research Institute of Sweden). Suvinen, A., 2006: A GIS-based Simulation Model for Terrain Tractability. Journal of Terramechanics 43: 427–449. Swedish Environmental Protection Agency, Swedish EPA, 2011:<http://www.naturvardsverket.se/en/In-English/Start/ Environmental-objectives/Swedens-environmental-objecCroat. j. for. eng. 33(2012)2

A GIS Approach to Analyzing Off-Road Transportation: a Case Study in Sweden (275–284) tives/The-national-environmental-objectives/> (Accessed 18 September 2011). Terrain Classification System for Forestry Works (1992), Editor: Berg, S., Skogforsk skopg. Terrängtypsschema för Svensk Skogsbruk (Terrain Classification for Swedish Forestry) (1969), Redogörelse Nr. 9. Forskning stiftelsen Skogsarbeten (The Logging Research Foundation). Vega, D., Nevis, J., Murphy, P., Castonguay, M., Ogilvie, J. Arp, P., 2009: A Modular field model for daily variations in machine-specific forest soil trafficability. Canadian Journal of Soil Science 89(1): 93–109. Vibrationsdämpning av skotare (Damping forwarders vibration) (2008), Editor: Baez, J., Arbetsrapport 667, Skogforsk (Forestry Research Institute of Sweden). Yavuzcan, G., Matthies, D., Auernhammer, H., 2005: Vulnerability of Bavarian silty loam soil to compaction under heavy wheel traffic: impacts of tillage method and soil water content, Soil & Tillage Research 84(2): 200–215. Ziesak, M., 2003: Avoiding soil damage caused by forest machines. Proceedings Decision Support System/Tools, In: 2nd Forest Engineering Conference, Växjö Sweden, Skogforsk (Forestry Research Institute of Sweden).

S. Mohtashami et al.

Ziesak, M., 2004: Entwicklung eines Informationssystems zum bodenschonen den Forstmachineneinsatz, Doctoral Dissertation (Development of an information system for soil protective machine operations; author’s translation), Lehrstuhl für Forstliche Arbeitswissenschaft und Angewandte Informatik, Technische Universität München. Freising. Panel of consulted experts: Bergkvist, I., PhD candidate, The Forestry Research Institute of Sweden, Skogforsk, Uppsala, 2011. Högbom, L., Associate Professor, PhD, The Forestry Research Institute of Sweden, Skogforsk, Uppsala, 2011. Jönsson, P., MSc., The Forestry Research Institute of Sweden, Skogforsk, Uppsala, 2011. Mörk, A., Forestry Technician and Instructor, The Forestry Research Institute of Sweden, Skogforsk, Uppsala, 2011. Ring, E., PhD, The Forestry Research Institute of Sweden, Skogforsk, Uppsala, 2011. Sonesson, J., PhD, The Forestry Research Institute of Sweden, Skogforsk, Uppsala, 2011. Thor, M., Programme Manager, PhD, The Forestry Research Institute of Sweden, Skogforsk, Uppsala, 2011.

Sažetak

Raščlamba terenskoga transporta uz pomoć GIS-a: primjer iz Švedske Izvoženje drva pri njegovu pridobivanju započelo je u Švedskoj 1960-ih godina. Taj se oblik primarnoga trans porta obavljao nakon dovršnih sječa tijekom zime na pripremljenim zaleđenim sekundarnim prometnicama, što je štitilo šumsko tlo i ublažavalo njegovo moguće oštećivanje. Današnji šumski radovi u potpunosti su mehanizirani i provode se cijele godine. Zbog toga su sekundarne šumske prometnice (vlake i putovi) pod značajnim utjecajem ra dova koji se provode u šumskom bespuću. Oštećivanje tla izaziva mehaničke, kemijske, biološke (smanjen prirast), hidrološke i ekonomske promjene i narušava kakvoću podzemnih i površinskh voda zbog prekomjernoga otpuštanja zagađivača, kao što je metil-živa koja nastaje od anorganske žive u anaerobnim uvjetima u jezerima ili rijekama (http:// en.wikipedia.org/wiki/Anaerobic_organism). Slični problemi pojavljuju se i u drugim regijama, gdje se privlačenje drva izvodi u blizini vodotoka ili u područjima podložnim oštećenju tijekom toplijih godišnjih razdoblja. Direktiva Europske unije o vodama utječe na šumske radove tako što stvara poticajno okruženje za ublažavanje i otklanjanje tih problema. Trenutačni napori u konstrukciji vozila i pripadajuće opreme vjerojatno će poboljšati provedbu šumskih radova. Oštećivanje tla može se izbjeći primjenom privremenih prijelaza preko ugroženoga područja ili pomoćne opreme, ali i upotrebom nove generacije tehnika i tehnologija laserskoga skeniranja u šumama, s 8–10 lokacija po kvadratnom kilometru, koje pruža prilično precizne podatke o sastojinskim i terenskim značajkama. Ta je vrsta sustava za pomoć pri donošenju odluka o određivanju izvoznih pravaca s manjim posljedicama za okoliš bila izrađena i testirana u pokusnom području smještenom u jugoistočnoj Švedskoj. Različiti digitalizirani slojevi, na primjer nadmorska visina i nagib, izlučeni su iz digitalnoga modela reljefa visoke razlučivosti (0,5 m × 0,5 m) radi pronalaženja najpovoljnijih područja za vožnju izbjegavajući pri tome tehničke probleme koji se javljaju na strmim terenima. Ti su slojevi bili združeni sa slojevima tipova tla i zaštićenih područja da bi se dobila osnovna karta vrijed nosnoga indeksa područja. Ta karta dijeli područje u pet razina prikladnosti za vožnju primjenom razredbe koja se zove vrijednosni indeks. Niži indeks označuje pogodniji teren s obzirom na nosivost tla. U sljedećem koraku pomoću navedenoga indeksa vrijednosti površine model pronalazi najkraće putove najmanjega kumulativnoga vrijednosnoga indeksa spajajući bilo koje željeno odredište u sječini s odabranom lokacijom pomoćnoga stovarišta. Svakako, u tom Croat. j. for. eng. 33(2012)2

283

S. Mohtashami et al.

A GIS Approach to Analyzing Off-Road Transportation: a Case Study in Sweden (275–284)

je studijskom području močvara, koja je trebala biti izuzeta od vožnje, bila locirana uz pomoćno stovarište, što je zahtijevalo da se pri planiranju procijene dva različita pristupa izradi izvoznih pravaca za prikupljanje drva, 1) vožnjom iza močvare i dosegom odredišta u sječini za utovar drva, ili 2) izgrađivanjem prijelaza preko močvare radi dolaska do pomoćnoga stovarišta. Rezultati su pokazali smanjenje udaljenosti vožnje dobivene prelaskom preko izgrađenoga mosta uz smanjenje operativnih troškova, ali i povećanje troškova zbog izgradnje mosta, do čega se došlo uz pomoć opisanoga modela za planiranje izvoznih putova. Ključne riječi: sustav GIS za pomoć pri odlučivanju, digitalni model reljefa, oštećivanje tla, šumski radovi, izvoženje drva, planiranje, gaženje, tlo

Authors’ addresses – Adrese autorâ:

Received (Primljeno): February 17, 2012 Accepted (Prihvaćeno): April 30, 2012

284

Sima Mohtashami, MSc. sima.mohtashami@skogforsk.se Isabelle Bergkvist, MSc. Isabelle.bergkvist@skogforsk.se Björn Löfgren, PhD. bjorn.lofgren@skogforsk.se Staffan Berg, Doctor in forestry* staffan.berg@skogforsk.se The Forest Research Institute of Sweden SE-90183 Uppsala Science Park SWEDEN Croat. j. for. eng. 33(2012)2

Original scietific paper – Izvorni znanstveni rad

Performance of GPS Stochastic Modeling for Forest Environment R. Cüneyt Erenoğlu Abstract – Nacrtak The Global Positioning System (GPS) now makes it possible to define forest boundaries using double differenced carrier phase observables. They are mostly processed with algorithms based on the Least-Squares Estimation (LSE). Although GPS was completely developed for outdoor navigation, sometimes it can be used in near/under tree or building shading. In such a case, before applying the LSE, both the functional and stochastic models should be properly defined in order to obtain reliable positioning. While the functional model for precise GPS positioning is sufficiently well known, realistic stochastic modeling is still a difficult task to accomplish in the case of unfavorable conditions. This paper analyzes the achievable efficiency of the stochas tic modeling for the positioning near/under the forest. A static campaign was performed at two surveying sites that have been established near the effect of tree shading. The experiments show the efficiency of stochastic models depending on the forest. It is clear that sigma-ε and sigma-Δ models give optimum solutions for the sites located near the tree canopy. Moreover, weighting procedures based on the C/N0 values can successfully cope with the corruptive effects caused by the tree canopy. As a result, a proper stochastic model for carrier phase ob servables should be used as an important tool in parameter estimation for handling multipath effect and signal distortion caused by the forest canopy. Keywords: GPS; forest; efficiency; stochasticity; weighting

1. Introduction – Uvod NAVSTAR GPS (Navigation System with Time and Ranging Global Positioning System) was developed in 1973 by the United States Department of Defense, and it is the satellite-based positioning and time transfer system. The United States NAVSTAR GPS and the Russian GLObal NAvigation Satellite System (GLONASS) are fully globally operational systems, as of October 2011. China is in the process of expanding its regional Beidou navigation system into the global Compass navigation system by 2020. The European Union’s Galileo positioning system is in initial deployment phase, scheduled to be fully operational by 2020 at the earliest. Generally, the satellite navigation system with global coverage may be termed a Global Navigation Satellite System (GNSS). The first purpose of establishing such a system is to provide the requirement for 24 hours of military positioning in all weather condition. It is well known that the system found for military purposes is limited to the management authority for Croat. j. for. eng. 33(2012)2

civil users. Thanks to the GPS system, the high precise information of 3D positioning, velocity and time can be obtained in a global coordinate system. GPS system basically consists of three components: Space component, control component and a user component. The overview of the development of GPS system is given by Clarke (1994); Kaplan (1996); Rizos (1997); Kleusberg and Teunissen (1998); Hoffmann-Wellenhof et al. (2001); Leick (2004). GPS measurements are carried out using the electromagnetic waves from satellites to receivers. For this purpose, there are the two main frequencies for a GPS satellite, L1 and L2. So the two GPS carried signals are used in engineering, topographic, forestry and cadastral applications. The reason for dual frequency GPS system is to serve as the backup frequency in case of interruption of one of the frequencies, and of course to model the ionospheric effect by using the dual frequency (Teunissen and Kleusberg 1998). In the application of GPS, the baselines between the surveying sites are generally short. Since the de-

285

R. C. ErenoÄ&#x;lu

Performance of GPS Stochastic Modeling for Forest Environment (285â&#x20AC;&#x201C;301)

sired accuracy is high, GPS carrier phase observables are processed by solving integer cycle ambiguity. For the GPS application, the accuracy depends strongly on the following factors: (a) Modeling of atmospheric effects and clock errors, (b) effective fixing of ambiguity parameter, (c) site environment (Rizos 1997; Hoffmann-Wellenhof et al. 2001; Leick 2004; Li et al. 2008). Atmospheric propagation on the GPS observables can be handled using well known atmospheric models (Roberts and Rizos 2001, Ibrahim and El-Rabbany 2008). Clock errors are completely removed by double differencing. Carrier phase ambiguity determination has been an important field of investigation for 30 years. All in all, the residual errors of GPS observables are mainly due to specific effects of the site, e.g. multipath. It is clear that signal deterioration caused by the GPS site environment is still a main concern (Talbot 1988; Han 1997; Hartinger and Brunner 1998; Lau and Mok 1999; Pirti 2008). Precision and accuracy of GPS receivers for static surveying in forest areas has been studied in detail by Naesset (2001); Hasegawa and Yoshimura (2003); Yoshimura and Hasegawa (2003); Naesset and Gjevestad (2008). In forest environment, there may be some restrictive factors that cause corruptive effects on the receiver of GPS signal, such as heavy forest canopy or steep terrain model (Danskin et al. 2009; Pirti et al. 2010). In such a case, corruptive objects may affect GPS satellite signal and make it difficult to get reliable products. Practically, the GPS receivers have to be used only on high elevation satellites when they are restricted by the skyview in a forest environment. In forestry applications, high precision GPS positioning is based on both mathematical and stochastic models. The mathematical model relates the observed quantities to unknown parameters, i.e. the phase observables and the coordinates of the GPS sites, respectively. If signal distortion and multipath errors by objects are not taken into account, the model will be misspecified and yield biased estimates. There are three ways to eliminate them: Modeling the effects mathematically, modifying stochastically and rejecting outlying observables. The aim of this study is to investigate GPS stochastic models in the case of the GPS signal failures due to the forest environment, especially for short baselines. To do it, we used some stochastic models based on signal quality indicators and derivation of variance models that reflect the characteristics of GPS observables. Thereby, the achievable efficiency of the stochastic models can be assessed on the basis of the parameter estimation.

286

2. Problem Description â&#x20AC;&#x201C; Opis problema The processing of the GPS data is based on the Least Squares Estimation (LSE) method as well as the adjustment procedure for other geodetic networks. The establishment of a functional model and a stochastic model is required before using the LSE method in computing 3D position of GPS stations/sites. As it is known, the functional model contains the mathematical relationships between the GPS measurements (code and carrier phase measurements) and the unknown parameters (atmospheric delays, clock error, the carrier phase ambiguity and baseline components). The stochastic model defined by a variance-covariance matrix, reflects the basic statistical properties of GPS observations, and it is very important in terms of data quality. In order to achieve high accuracy, both models should be correctly identified. In the case of accurately establishing the functional and stochastic model, the residuals obtained by the LSE have normal distribution. Although the functional model is created strictly, the stochastic model is still a fundamental concern for processing GPS phase observables. Estimation of the unknown parameter can provide good solutions if and only if the stochastic model of the observables is formed realistically. The stochastic model is described by the Variance-Covariance Matrix (VCM) that accounts for the variances of the GPS observables and their correlations. As it is well known, there are two possible ways to obtain the VCM: With the estimation of variance and covariance values for the observables, using an a priori variance model. As with all known observing techniques, the GPS code measurements and carrier phase also include random errors (Satirapod et al. 2000; Tiberius and Kenselaar 2000). While running under ideal conditions and with a sufficient observing period, it is expected that the residuals from LSE method will be of the same character as the true errors. Up to present day, many statistical techniques based on LSE method have been successfully developed and used in order to achieve the desired high accuracy (Rao 1971; Ziqiang 1991). Satellite orbital errors, satellite and receiver clock errors have been completely eliminated in the step of creating a functional model since the principle of relative point positioning by GPS is a difference processing algorithm (Rizos 1997; Teunissen and Kleusberg 1998; Hoffmann-Wellenhof et al. 2001; Leick 2004). However, the residuals and parameters estimated by the LSE will be unfavorably affected due to the existing functionally unmodeled errors and noise, such as the tropospheric delay, multipath and signal dispersion. Several studies have been done related to such effects to develop the stochastic models for increasing Croat. j. for. eng. 33(2012)2

Performance of GPS Stochastic Modeling for Forest Environment (285–301)

the sensitivity of GPS technique (Han 1997; Barnes et al. 1998; Wang 1999; Satirapod 2002). The delay effect due to the refraction of troposphere layer can be eliminated using appropriate models based on the atmospheric profiles. Multipath is the major site dependent error because it depends on environment around the GPS antenna, especially for the application with short baselines. For example, the electromagnetic wave may be achieved from the satellite to receiver in more than one way due to the multipath reflectors around the GPS site, such as tree or building shading. This effect is modeled using the techniques developed to reduce the multipath effect. The effects of multipath and signal dispersion can be reduced using the stochastic models developed in the GPS data processing, and accuracy of the parameter estimation is thereby increased. A priori and a posteriori variance is not equal to each other because all systematic and random errors cannot be modeled by using the double differenced observable equations. Moreover, these errors cause the correlations between the observables in the differencing step. All in all, the stochastic models that describe the character of the GPS observables affected by these errors, must be used (Tiberius et al. 1999; Pachter and Nguyen 2007). For this purpose, the appropriate processing models should be used, as well as the definitions of observables. The elements of the weight matrix define stochastic model, and it provides information on how each GPS observation can contribute to the solution. A kind of weight modification can be recommended for a highaccurate result. For example, a weight reduction and accession for the observables can be affected more and less than the others (Teunissen et al. 1998). Because of the relationship between the weights and variancecovariance matrix, it is important to know the realistic stochastic properties of the observables. For stochastic modeling of GPS observables, some signal quality criteria are widely used (Tiberius et al. 1999). Many studies based on signal quality criteria have been done. For example, the comparison of different GPS data for modeled residuals (Satirapod 2006; Satirapod and Luansang 2008), the determination of minimum shifting using a weighting based on the ionosphere (Jong and Teunissen 2000), the implementation of artificial neural networks approach (Jwo 2007), internal and external reliability of the GPS method (Kuusniemi et al. 2004). There have been some investigations on the effects of the mathematical correlations that arise from differencing of GPS observables (Yang et al. 2002; Ding et al. 2004). In addition, using weighting models derived from C/N0 values, the effect of multipath reflecCroat. j. for. eng. 33(2012)2

R. C. Erenoğlu

tion has been modeled for the short baselines (Wieser and Brunner 2000, 2002; Özlüdemir and Ayan 2003). Some GPS receiver manufacturers have developed receivers integrated with pre-assessment software in order to eliminate these effects.

3. GPS Processing Models – Modeli obrade GPS-ovih podataka In the case that the baseline length is less than 20 km for GPS application, the geometry based mathematical models can be easily generated using the technique of double differencing of the GPS observations. The processing models for GPS observables have been studied most extensively in the literature, and many contributions on the subject are available, e.g. Teunissen et al. (1998); Teunissen et al. (2000); HoffmannWellenhof et al. (2001); Odijk et al. (2002); Odijk 2003; Leick (2004); Verhagen (2004). The functional and stochastic models for double differenced code and carrier phase observations can be established as follows: (1) (2) where yGPS is the vector of double differenced code and carrier phase observables, aGPS is the vector of carrier phase ambiguity, b is the vector of coordinate unknowns, AGPS is the coefficient matrix of the vector of a, BGPS is the coefficient matrix of the vector of b, S is the covariance matrix of the observables of y, Qy is the cofactor matrix of the observables of y and s20 is the variance of the observation with a priori weight. In some applications, GPS cannot provide accurate results due to undesired errors (Teunissen and Kleusberg 1998). Especially in the case of signal distortion and mutlipath effects, some criteria should be used in creating stochastic models. These are based on the power of a GPS signal. In the following subsections, we will give the models of variance that are widely used for weighting double differenced observations.

3.1 Equal Variance Weighting Model – Model ponderiranih jednakih varijanci In the step of processing GPS data, the simplest approach for weighting observables is to assume that all observables have equal variances. This general approach has been used as standard stochastic model, especially in cases of ideal environmental conditions. For example, there is a panel of the standard stochastic model in the Bernese v.5.0 software (Dach et al. 2007). However, under unfavorable observing conditions,

287

R. C. Erenoğlu

Performance of GPS Stochastic Modeling for Forest Environment (285–301)

70% of the data can be successfully modeled in stochastic terms at most. Especially in the case of signal distortion or multipath, this rate reduced even more. Although some improvements have arisen mainly with increasing of satellite elevation angle, the volume of data collected decreased. For these reasons, it is recommended to avoid using the models of equal variance weighting for high-precision GPS applications (Hartinger and Brunner 1999).

3.2 Variance Model based on Satellite Elevation Angle – Model varijanci zasnovan na kutu nagiba satelita As an alternative to equal variance weighting, a large number of stochastic models are proposed, defined by a function of the elevation angles (Wang et al. 1998; Collins and Langley 1999). An example for this approach is the cosine weighting based on a cosine function of elevation angle for related satellite (Rothacher et al. 1997). In the developed stochastic model, variance of carrier phase observable can be shown as: ,

(3)

(4)

2 where sf2 is the variance of carrier phase, s90 is the variance of carrier phase in zenith direction, z is zenith angle of satellite, h is satellite elevation angle. s290 can be estimated experimentally. In this way, covariance matrix is formed for double differenced GPS observables.

3.3 Variance Models based on Signal to Noise Ratio – Model varijanci zasnovanih na odnosu signala i šuma As mentioned above, various weight functions based on Signal to Noise Ratio (SNR) have been developed in order to model the effects of multipath and signal distortion, and to reduce their corruptive effects on parameter estimation (Lau and Mok 1999; Hartinger and Brunner 1998; Tiberius and Kenselaar 2000; Schön and Brunner 2008). An example of these studies is the approach that uses stochastic models depending on SNR value regarded as a quality indicator for the GPS observations (Spilker 1996). The relationship between the values of SNR and variance of the observables will be established as follows: ,

(5)

where SNR is Signal to Noise Ratio and s2f is variance of carrier phase.

288

3.4 Variance Models based on Carrier to Noise Power Density Ratio – Model varijanci zasnovan na odnosu prijamnika i snage gustoće šuma For establishing stochastic models reflecting the characteristics of the carrier phase observables, some weighting models are available as a function of the C/ N0 models. Contribution of these models to the solution is the use of more realistic models that reflect signal quality instead of standard models. In addition, corruptive effects can be reduced using C/N0 models in the case of low satellite elevation angles. One of the models developed as a function of the C/N0 values is sigma-ε (Hartinger and Brunner 1999; Özlüdemir and Ayan 2003). For the variance model proposed, the relationship between C/N0 of the carrier phase and Ci model parameter can be expressed as follows: (6) where s2f is variance of carrier phase. The parameter Ci is taken as 1.61 × 10–4 mm2 (Brunner et al. 1999). The variances are estimated for each observation epoch. As a result, covariance matrices are computed for double differenced (DD) phase observables using the rules of error propagation. (Brunner et al. 1999; Koch 1999). Thus DD observables can be statistically modeled as required by the characteristic structure of the signal. At each GPS site/station, C/N0 values may vary depending on the type of antenna and receiver as well as on environmental conditions. Since the sigma-ε model derives different values of variance for each station, the stochastic modeling may not always be effective against signal distortion. As is known, the C/N0 values of GPS signal corrupted by various reasons are lower than the ones of the others that have the same elevation angle. Considering this corruption effect, the equation (6) is transformed as follows: ,

(7) ,

(8)

where s2f is variance of carrier phase, Ci and γ are parameters of model, Δ is difference between existing C/ N0 value and C/N0 in template model. This model is called sigma-Δ model because of the term Δ. In determining of template model, the largest possible values of C/N0 for signals received by GPS sites were recorded. In case that a signal is not corrupted, the difference of Δ is small. Croat. j. for. eng. 33(2012)2

Performance of GPS Stochastic Modeling for Forest Environment (285–301)

4. Case Study Fieldwork – Terensko istraživanje The biases in the parameter estimation can be anticipated at the centimeter level for the GPS applications in multipath environment such as forest boundaries. If observables are contaminated by systematic errors due to environmental objects, there will be some troubles in ambiguity solution and baseline estimation. The purpose was to investigate the effectiveness of stochastic models in estimating the parameters in the case of GPS sites near the forest environment. In this study, we performed a carefully designed experiment that includes signal distortion and multipath effects due to tree shading. Thereby, the effect of signal distortion in using stochastic modeling in post processing of GPS observables can be investigated in detail. The experiments were carried out in the Samandira region of Istanbul, Turkey, see Fig. 1. A basic geodetic baseline consisting of two sites was surveyed using the GPS surveying method. The baseline length was 20,000 m with a height difference of 1.970 m. The aim of this study was to compare the performance of the GPS stochastic models under the effects of signal diffraction. In order to create the multipath environment, one of the static GPS sites, the site

R. C. Erenoğlu

T1, was situated far from the forest obstruction. Data with signal distortion effects was collected at the site T2 located at the forest boundary. In forest areas, GPS signal diffraction effect may occur due to diffuse obstacles such as bushes, branches and trees. Herein we designed the example including the multipath error, see Fig.2. The static observables were taken on May 27, 2006 (Day of Year, DOY: 147). Data were collected at these two sites for 6 hours between 04:30–10:30 UT with a sample rate of 30 seconds. Therefore, 720 observations were obtained during one day of field data collection. Satellite elevation angle was assumed to be 5 degrees. Magellan Thales Z-Max receivers with internal antennas were used at all sites. Table 1 shows the basic information of the dataset. As an obstacle, the trees at the forest boundary mainly caused shading of satellites during the experiment. Fig. 3 shows the skyplots prepared for the sites T1 and T2. It is clear from the satellite distribution that there are some differences between the satellite visibilities between the sites. From the skyplot, we can see that the receiver at the site T1 tracked more satellites continuously than the other, as some satellites were shaded by trees. At the site T2, the receiver recorded less data from some satellites, e.g. PRN 4, PRN 15, PRN 18, PRN 19 and PRN 22. It can be seen that the

Fig. 1 Study area in Istanbul Slika 1. Područje istraživanja u Istanbulu Croat. j. for. eng. 33(2012)2

289

R. C. Erenoğlu

Performance of GPS Stochastic Modeling for Forest Environment (285–301)

Fig. 2 Study area, (a): GPS sites and forest area, (b): Site T1 far from the trees, (c): Site T2 at the forest border Slika 2. Područje istraživanja, (a): pozicije GPS-a i šumsko područje, (b): mjesto T1 udaljeno od stabala, (c): mjesto T2 na granici šume Table 1 Basic information about the experiment Tablica 1. Osnovne informacije o pokusu Date/Hour – Datum/Vrijeme

27.05.2006 / 04:30–10:30 UTC

Session type – Način rada

Static

Receiver type – Vrsta prijamnika

Magellan Thales Z-Max Receiver

Antenna type – Vrsta antene

Internal Antenna

GPS observables – Opažanja GPS-om

C1, P2, L1, L2, S1, S2

Data sampling rate – Frekvencija prikupljanja podataka

30 seconds

Satellite elevation angle – Kut nagiba satelita

5˚

Baseline length – Duljina osnovne linije

20.000 m

Height difference – Visinska razlika

1.970 m

Site coordinates – Koordinate mjesta pokusa Satellite orbit file – Datoteka orbite satelita Explanation – Dodatna pojašnjenja

trees shaded PRN 15 above 20° at the site T2, see Fig. 3 (right). The strong signal interruption and distortion may therefore be expected since 50% of the sky is obstructed by the trees at the site T2 mounted at the forest

290

T1: 40˚ 58’ 28”.75 N, 29˚ 13’ 14”.87 E T2: 40˚ 58’ 29”.36 N, 29˚ 13’ 15”.16 E igs13766.eph T1 : At a distance of 20 m from forest boundary T2 : At the forest boundary

boundary. It is clear that the coordinates of the site T1 were less affected by forest environment than the other site. Therefore, the coordinates of the site T1 should be fixed in the static processing. Croat. j. for. eng. 33(2012)2

Performance of GPS Stochastic Modeling for Forest Environment (285–301)

R. C. Erenoğlu

Fig. 3 Skyplot of sites T1 (left) and T2 (right), 04:30–10:30 with heavy obstruction by trees Slika 3. Skica satelita na nebu na mjestima: T1 (lijevo) i T2 (desno) s velikim smetnjama od stabala (04:30–10:30) In this study, the data were processed using the BERNESE v. 5.0, MATLAB v. 7.0 and some complementary tools (URL1, URL2). The computation scheme is shown in Fig. 4. The UNAVCO TEQC Software converts the data from BINARY format into RINEX files (URL3, URL4). In addition, TEQC is employed to obtain Signal to Noise Ratio (SNR) to be used in stochastic models later on. The pole data and CODE orbits are processed in the BERNESE in order to reconstruct the satellite orbits. In processing scheme, we used broadcast orbits for the short baselines because they provide enough accuracy (Teunis-

sen and Kleusberg 1998). Preprocessing of data is performed by the BERNESE software. The functional and stochastic models for the data are also generated via the GPSEST module of the BERNESE, i.e. the double differenced observables, the design matrix, the cofactor matrix and so on. For the other variance models, a complementary tool is used in order to obtain C/N0 values. The corresponding programs make a transformation from SNR into C/N0 values. Finally, the MATLAB software imports all functional and stochastic models, and processes static GPS sessions.

Fig. 4 Computing scheme: Inputs, algorithms, software programs and outputs Slika 4. Plan načina izračuna: ulazni podaci, algoritmi, računalni programi i izlazni podaci Croat. j. for. eng. 33(2012)2

291

R. C. Erenoğlu

Performance of GPS Stochastic Modeling for Forest Environment (285–301)

Fig. 5 Double difference residuals of PRN 29 for the baseline T1–T2 Slika 5. Dvostruki diferencijalni ostaci od PRN 29 za osnovnu liniju T1–T2

Fig. 6 Double difference residuals of PRN 25 for the baseline T1–T2 Slika 6. Dvostruki diferencijalni ostaci od PRN 25 za osnovnu liniju T1–T2 The ITRF2005 coordinates of the site T1 were estimated using data from the IGS permanent station ISTA, at a distance of 20 km from the project area, and the observables recorded during the experiment. Then, the coordinates of the site T1 were taken as known in processing of GPS experiment data including both sites. Atmospheric effects are completely eliminated by double differencing technique because of the short baseline and also low height difference for

292

this study (Teunissen and Kleusberg 1998). Moreover, we used the same type of antenna with same orientation during the experiment in order to remove errors occurring by antenna effect. Fig. 5 gives the double difference residuals (DDR) of PRN 29 for the baseline T1 and T2. These residuals mean very high data quality due to the visibility of PRN 29 only at the high elevation angle; see the skyCroat. j. for. eng. 33(2012)2

Performance of GPS Stochastic Modeling for Forest Environment (285–301)

R. C. Erenoğlu

Fig. 7 Double difference residuals of PRN 27 for the baseline T1–T2 Slika 7. Dvostruki diferencijalni ostaci od PRN 27 za osnovnu liniju T1–T2 plot in Fig. 3. As a result, the double differenced observables are not badly affected by systematic effects due to the shading of the forest. Only some scattering effects were recorded at the first 98 epochs of observation. A possible reason can be mutual shading of PRN 18 and PRN 29. Fig. 6 gives the DDR of PRN 25 for the same baseline. PRN 25 was only visible for a short interval of observation about the 420th epoch, due to heavy shading of trees. Although there is small scattering for the

limited data of PRN 25, the average of the DDR values is about 20 mm. We can conclude that the data quality is low because of the forest canopy. From the DDR of PRN 27 given in Fig. 7, it can be clearly seen that the spoiling effects are caused by the trees. Especially, during the first 60 epochs, non-systematic scatters attract the attention. Moreover, there are significant changes in the DDRs between the 60th and 254th observation epochs. This is because of the heavy distorting effects on the GPS signals. The bias

Table 2 Estimated baseline components and standard deviations for baseline between the sites T1 and T2 Tablica 2. Predviđene osnovne sastavnice i standardne devijacije za osnovicu između mjest T1 i T2

Baseline Component – Osnovna sastavnica

Standard Deviation – Standardna devijacija

(m)

(mm)

Stochastic Model Stohastički model

North

East

North

East

Sjever

Istok

Visina

Sjever

Istok

Visina

Equal variance weighting model

18.7780

6.6375

–1.2915

8.6

13.1

20.3

2 s90 / cos2z

18.7683

6.6589

–1.2892

7.7

11.0

18.5

1 / SNR

18.7795

6.6439

–1.3087

7.3

8.4

10.9

sigma-e

18.7727

6.6515

–1.3147

4.8

5.0

7.8

sigma-D

18.7717

6.6559

–1.3045

4.5

4.9

6.7

sin2h / SNR

18.7762

6.6463

–1.3075

6.1

8.8

9.5

Croat. j. for. eng. 33(2012)2

293

R. C. Erenoğlu

Performance of GPS Stochastic Modeling for Forest Environment (285–301)

Fig. 8 C/N0 values of L1 signal for the site T1 Slika 8. C/N0 vrijednosti signala L1 za mjesto T1

Fig. 9 C/N0 values of L1 signal and template model (^) for the site T2 Slika 9. C/N0 vrijednosti signala L1 i predloženi model (^) za mjesto T1 values decrease by about 40 mm. After the 254th epoch, the track of PRN 27 shaded by trees and the scatters for the DDR are significantly reduced. Then we investigated the C/N0 values of the signals recorded at each site. As mentioned above, C/N0 values provide key information about the effects of reflection and distortion of the signal. As is known, there are

294

different C/N0 values for L1 and L2 phases, which are designed in terms of fixing ambiguity. For the site T1, the C/N0 values and corresponding satellite elevation angles are given together in Fig. 8. The data collected at the site T1 have higher frequencies of C/N0 because the receiving signals from GPS satellites are less affected by forest canopy compared to the site T2. The impact of the tree canopy on the data collected at the site T2 is clearly seen in Fig. 9. The lower C/N0 values are the result of the signal distortion at the site T2 caused by the tree canopy. Thus C/N0 values of the signal recorded at the GPS receiver were significantly lower. In this study, stochastic models based on C/N0 values were used in evaluating GPS data, as well as other criteria of signal quality. In this study, the stochastic modeling of GPS data is used in several approaches. These are equal variance weighting model, satellite elevation angles dependent weighting model, SNR dependent weighting model, sigma-ε model, sigma-Δ model and a proposed model by the function of s2f  sin2h/SNR. Using these stochastic models, baseline components between the sites T1 and T2 were estimated. To do it, the data were processed by geometry-based model derived by the technique of double differencing. To obtain high precision in positioning, reliable estimates of the integer double difference ambiguities are determined in an efficient manner. In this study, we used the Least-squares AMBiguity Decorrelation Adjustment (LAMBDA) method (Teunissen, 1993). For the baseline estimation of T1 and T2, the data is processed using six different stochastic models. Table 2 presents a number of conclusions that have been reached about the effects of stochastic models used for the parameter estimation. Multipath effects are expected to occur especially at the site T2. In this regard, multipath fading badly affects the performance of the 1st and 2nd stochastic model in the processing, see Table 2. For these models, standard deviations are up to the cm level. This is why the processing strategies based on the equal variance weighting and low satellite angles have resulted in poor final products. The effects of multipath and signal distortion caused by the forest canopy were not adequately modeled using the first two models, especially with the low satellite angles. Considering SNR values of the signal in stochastic modeling, the 3rd model improved the results, see Table 2. The algorithms called sigma-ε and sigma-Δ have been developed to take C/N0 values of GPS signal into account in stochastic modeling. When looking all coordinate components and their standard deviations from epoch by epoch processing presented in Table 2, it is clear Croat. j. for. eng. 33(2012)2

Performance of GPS Stochastic Modeling for Forest Environment (285–301)

R. C. Erenoğlu

Fig. 10 Standard deviations from epoch by epoch processing of baseline T1-T2 Slika 10. Standardne devijacije obrade podataka za promatrano razdoblje osnovne linije T1-T2 that the sigma-ε and sigma-Δ give optimum solutions. For example, the standard deviations are at the level of 4–5 mm for the components of north and east. This is why the weighting procedure based on a signal quality indicator has successfully reduced the corruptive effects caused by the tree canopy. As is known, the elevation above ground gives worse products according to the components of north and east due to the troposphere, poor satellite geometry and reflection effects. Finally, we proposed a new method for stochastic modeling by combining the satellite elevation angle h and SNR values. The proposed model of sin2h/SNR provides better results compared to the ones from the 1/SNR model. In addition, the standard deviations in baseline components from the epoch by epoch processing are presented in Fig. 10. As seen from the results, the 3rd, 4th, 5th and 6th models provide better estimates than the others. Moreover, the standard deviations of the 3rd and 6th models are almost identical for the components of North and East. For estimating the baseline T1–T2, the elevation was significantly improved using the model of sin2h / SNR. In addition, the standard deviations of sigma-ε and sigma-Δ models are close to each other. The sigma-Δ model provides a positive contribution to the elevation. We processed GPS data to determine epoch by epoch positions as well as baseline components much more efficiently and accurately. This is why the epoch Croat. j. for. eng. 33(2012)2

by epoch analysis becomes more and more important for studying the efficiency of the variance models discussed before. Geodetic time series were used to show the contribution of stochastic modeling by applying some functions of signal quality indicators to estimation of the baseline components between the sites T1 and T2. The models of equal variance, sigma-Δ and sin2h / SNR are used as the variance model for the reasons described above. All data are processed for epoch by epoch positioning. We obtained the components of North, East and Up for the baseline T1–T2. Then, the averages of baseline components were determined separately. Finally, we computed deviations between the averages and time series. The deviations computed from geodetic time series of epoch by epoch processing are shown in Figs. 11, 12 and 13 for the equal variance, sigma-Δ and sin2h / SNR models, respectively. The deviations computed from the coordinates confirm that the stochastic models behave differently against the multipath and signal distortion caused by the forest canopy. The main reason is that they are totally based on various signal quality criteria. Fig. 11 indicates that the model of equal variance modeling produce less accurate solutions since it does not take quality criteria into account for stochastic modeling. The coordinate deviations have therefore the biggest scattering values. Figs. 12 and 13 show that sigma-Δ and sin2h / SNR models yield higher standard deviations for all coor-

295

R. C. Erenoğlu

Performance of GPS Stochastic Modeling for Forest Environment (285–301)

Fig. 11 Deviations of the components from averages for the equal variance modeling Slika 11. Odstupanja sastavnicâ od prosjeka za model jednakih varijanci

Fig. 12 Deviations of the components from averages for the sigma-Δ modeling Slika 12. Odstupanja sastavnicâ od prosjeka za model sigma-Δ

296

Croat. j. for. eng. 33(2012)2

Performance of GPS Stochastic Modeling for Forest Environment (285–301)

R. C. Erenoğlu

Fig. 13 Deviations of the components from averages for the sin2h/SNR modeling Slika 13. Odstupanja sastavnicâ od prosjeka za model sin2h/SNR

dinate components than the model of equal variance modeling. This is why the basic signal quality indicator, like C/N0, SNR and h, provide a positive contribution to the coordinate component reducing bad effects caused by the forest canopy. So they allow for successful positioning of the data.

5. Conclusions – Zaključci Engineering, topographic, forestry and cadastral applications using data are commonly performed for short baseline, short observing time and also high accuracy demands. However, the site cannot be established to ideal conditions for accurate positioning. In other words, natural or artificial obstacles close to an antenna cannot be avoided. In such a case, the GPS observables will be spoiled by the effects of multipath and signal distortions. This paper concluded that forest canopy has a great effect on the stochastic modeling for processing the GPS observables. Many disturbing effects can be modeled functionally and/or removed by the technique of double differencing. In addition, the carrier phase observables should be modeled stochastically against Croat. j. for. eng. 33(2012)2

effects of multipath and diffraction caused by the forest, using appropriate variance functions that reflect quality characteristics of GPS signal. Thereby, the influence of signal distortion will be reduced on the parameter estimation, i.e. fixing of carrier phase ambiguity and estimation of baseline components. The experiments also showed that the models of sigma-ε and sigma-Δ give optimum solutions for the sites located near the tree canopy. The weighting procedure based on the C/N0 values can successfully cope with the corruptive effects caused by the tree canopy. In addition, we presented a new method sin2h/SNR by combining the satellite elevation angle h and SNR values. According to the results, the proposed model gives better results compared to the ones from the 1/ SNR model.

Acknowledgements – Zahvala We obtained IGS precise orbits and ITRF2005 coordinates of ISTA from SOPAC archives. Also Google Earth picture was used as Fig 2. The author is deeply grateful to Dr. Atinc Pirti and Dr. Serif Hekimoglu for their contributions to this study. The comments by two anonymous reviewers are gratefully acknowledged.

297

R. C. Erenoğlu

Performance of GPS Stochastic Modeling for Forest Environment (285–301)

6. References – Literatura Barnes, B. J., Ackroyd, N., Cross, P. A., 1998: Stochastic Modelling for Very High Precision Real–Time Kinematic GPS in an Engineering Environment, Proceedings of FIG XXI International Conference, 21–25 July, Brighton, U.K., Commission 6: 61–76. Brunner, F. K., Hartinger, H., Troyer, L., 1999: GPS Signal Diffraction Modelling: The Stochastic SIGMA–δ Model. Journal of Geodesy 73: 259–267. Collins, J. P., Langley, R. B., 1999: Possible Weighting Schemes for GPS Carrier Phase Observations in the Presence of Multipath. Contract Report DAAH04-96-C-0086/TCN 98151 for United States Army Corps of Engineers Topographic Engineering Center, 33 pp. Dach, R., Hugentobler, U., Fridez, P., Meindl, M., 2007: Bernese GPS Software Version 5.0. Astronomical Institute, University of Bern, Switzerland. Danskin, S., Bettinger, P., Jordan T., 2009: Multipath Mitigation under Forest Canopies: A Choke Ring Antenna Solution. Forest Science 55(2): 109–116. Ding, X. L., Liu, G. X., Li, Z. W., Li, Z. L., Chen, Y. Q., 2004: Ground Subsidence Monitoring in Hong Kong with Satellite Sar Interferometry. Photogrammetric Engineering & Remote Sensing 70(10): 1151–1156. Clarke, B., 1994: Aviator’s Guide to GPS. McGraw–Hill, Inc.: 235 pp. Han, S., 1997: Carrier Phase–Based Long–Range GPS Kinematic Positioning, PhD Dissertation, School of Geomatic Engineering, The University of New South Wales, Sydney, Australia, S–49, 185 pp. Hartinger, H., Brunner, F. K., 1998: Experimental Detection of Deformations using GPS. In: Kahmen, H., Brückl, E., Wunderlich, T. (eds) Geodesy for Geotechnical and Structural Engineering, Proceedings of IAG Special Commission 4 Symposium Eisenstadt, pp. 145–152.

Jwo, D. J., 2007: Outlier Resistance Estimator for GPS Positioning – The Neural Network Approach. Journal of Navigation 60(1): 129–145. Kaplan, E. D., 1996: Understanding GPS, Principles and Applications, Boston: Artech House, Inc.: 726 pp. Koch, K. R., 1999: Parameter Estimation and Hypothesis Testing in Linear Models, 2nd Edition, Springer, Berlin: 333 pp. Kuusniemi, H., Lachapelle, G., Takala, J. H., 2004: Positioning and Velocity Reliability Testing in Degraded GPS Signal Environments. GPS Solutions 8: 226–237. Lau, L., Mok, E., 1999: Improvement of GPS Relative Positioning Accuracy by Using SNR. Journal of Surveying Engineering 125(4): 185–202. Leick, A., 2004: GPS Satellite Surveying, 3rd Edition, Hoboken, New Jersey: John Wiley & Sons, Inc.: 435 pp. Li, B. F., Shen, Y. Z., Xu, P. L., 2008: Assessment of Stochastic Models for GPS Measurements with Different Types of Receivers”, Chinese Sci. Bull. 53(20):3219–3225. Naesset, E., 2001: Effects of Differential Single– and Dual– Frequency GPS and GLONASS Observations on Point Accuracy under Forest Canopies. Photogrammetric Engineering & Remote Sensing 67: 1021–1026. Naesset, E., Gjevestad, J. G., 2008: Performance of GPS Precise Point Positioning under Conifer Forest Canopies. Photogrammetric Engineering & Remote Sensing 74: 661–668. Odijk, D., 2003: Ionosphere–Free Phase Combinations for Modernized GPS. Journal of Surveying Engineering 129(4):165–173. Odijk, D., Teunissen, P. J. G., Tiberius, C. C. J. M., 2002: Triple–Frequency Ionosphere–Free Phase Combinations for Ambiguity Resolution, Proceedings of the ENC–GNSS 2002, The European Navigation Conference, Copenhagen, Denmark. Ozludemir, T., Ayan, T., 2003: Effects of Stochastic Modelling for GPS Positioning. Journal of ITU 2:45–64 (In Turkish).

Hartinger, H., Brunner, F. K., 1999: Variances GPS Phase Observations: the SIGMA–ε Model. GPS Solutions 2(4): 35–43.

Pachter, M., Nguyen, T. Q., 2007: Stochastic Modeling Based DGPS Estimation Algorithm. Navigation 54(2): 125–138.

Hasegawa, H., Yoshimura, T., 2003: Application of Dual–Frequency GPS Receivers for Static Surveying under Tree Canopies. Journal of Forest Research 8(2): 103–110.

Pirti, A., 2008: Accuracy Analysis of GPS Positioning Near the Forest Environment, Croatian Journal of Forest Engineering 29(2): 189–199.

Hofmann–Wellenhof, B., Lichtenegger, H., Collins, J., 2001: GPS Theory and Practice, 5th Edition. Springer– Verlag Wien–New York, 382 pp.

Pirti, A., Gumus, K., Erkaya, H., Hosbas, R. G., 2010: Evaluating Repeatability of RTK GPS/GLONASS Near/Under Forest Environment, Croatian Journal of Forest Engineering 31(1): 23–33.

Ibrahim, H., El–Rabbany, A., 2008: On Stochastic Modelling of NOAA–Based Residual Tropospheric Delay. Journal of Arab Academy for Science, Technology and Maritime Transport 34(66): 28–36.

Rao, C. R., 1971: Estimation Variance and Covariance Components – MINQUE. Journal of Multivariate Analysis 1:257– 275.

De Jong, K., Teunissen, P. J. G., 2000: Minimal Detectable Biases of GPS Observations for A Weighted Ionosphere. Earth, Planets and Space 52:857–862.

Rizos, C., 1997: Principle and Practice of GPS Surveying, Monograph 17, School of Geomatic Engineering, The University of New South Wales.

298

Croat. j. for. eng. 33(2012)2

Performance of GPS Stochastic Modeling for Forest Environment (285–301) Roberts, C. A., Rizos, C., 2001: Mitigating Differential Troposphere for GPS Based Volcano Monitoring, Proceedings of the 5th Int. Symp. on Satellite Navigation Technology & Applications, Canberra, Australia. Rothacher, M., Springer, T. A., Schaer, S., Beutler, G., 1997: Processing Strategies for Regional GPS Networks, Proceedings of the IAG General Assembly, Springer, Rio de Janeiro: 93–100. Satirapod, C., 2002: Improving the GPS Data Processing Algorithm for Precise Static Relative Positioning, PhD Dissertation, School of Surveying and Spatial Information Systems, The University of New South Wales, Sydney. Satirapod, C., 2006: Stochastic Models Used in Static GPS Relative Positioning, Survey Review 38(299): 379–386. Satirapod, C., Wang, J., 2000: Comparing the Quality Indicators of GPS Carrier Phase Observations. Geomatics Research Australasia 73: 75–92. Satirapod, C., Luansang, M., 2008: Comparing Stochastic Models Used in GPS Precise Point Positioning Technique, Survey Review 40(308): 188–194. Schön, S., Brunner, F. K., 2008: A Proposal for Modelling Physical Correlations of GPS Phase Observations, Journal of Geodesy 82: 601–612. Spilker, J. J., 1996: GPS Signal Structure and Theoretical Performance, Global Positioning System: Theory and Applications, Progress in Astronautics & Aeronautics, 163: 57–119. Talbot, N., 1988: Optimal Weighting of GPS Carrier Phase Observations Based on the Signal–to–Noise Ratio, Proceedings of the International Symposium on Global Positioning Systems, 17–19 October 1998, Gold Coast, Queensland, 4.1– 4.17. Teunissen, P. J. G., 1993: Least–Squares Estimation of the Integer GPS Ambiguities, Invited Lecture, Section IV Theory and Methodology, IAG General Meeting, Beijing. Teunissen, P. J. G., 1998: The Least–Squares Ambiguity Decorrelation Adjustment: A Method for Fast GPS Integer Ambiguity Estimation. Journal of Geodesy 70: 65–82.

R. C. Erenoğlu

Teunissen, P. J. G., Kleusberg, A., 1998: GPS for Geodesy, 2nd Enlarged Edition, Springer– Verlag Wien–New York: 650 pp. Teunissen, P. J. G., Joosten, P., Jong C. D., 2000: Frequency Selection and Ambiguity Resolution, Galileo’s World 2(2): 38–43. Tiberius, C. C. J. M., Kenselaar, F., 2000: Estimation of the Stochastic Model for Code and Phase Observations, Survey Review 35:441–454. Tiberius, C. C. J. M., Jonkman, N., Kenselaar, F., 1999: The Stochastic Model of GPS Observables, GPS World 10: 49–54. Verhagen, S., 2004: The GNSS Integer Ambiguities: Estimation and Validation, PhD Dissertation, The Department of Mathematical Geodesy and Positioning of the Delft University of Technology, Thijsseweg. Wang, J., 1999: Modelling and Quality Control for Precise GPS and GLONASS Satellite Positioning, PhD Dissertation, School of Spatial Sciences, Curtin University of Technology, Perth. Wang, J., Stewart, M. P., Tsakiri, M., 1998: Stochastic Modeling for Static GPS Baseline Data Processing, Journal of Surveying Engineering 124(4): 156–170. Wieser, A., Brunner, F. K., 2000: An Extended Weight Model for GPS Phase Observations, Earth Planet Space 52: 777–782. Wieser, A., Brunner, F. K., 2002: Short Static GPS Sessions: Robust Estimation Results, GPS Solutions 50: 536–543. Yang, Y., Song, L., Xu, T., 2002: Robust Estimator for Correlated Observations based on Bifactor Equivalent Weights, Journal of Geodesy 76: 353–358. Yoshimura, T., Hasegawa, H., 2003: Comparing the Precision and Accuracy of GPS Positioning in Forested Areas. Journal of Forest Research 8(3): 147–152. Ziqiang, O., 1991: Approximately Bayes Estimation for Variance Components, Manuscripta Geodaetica 16:168–172. URL1: http://www.aiub.unibe.ch/ URL2: http://www.mathworks.com/products/matlab/ URL3: http://www.unavco.org/ URL4: ftp://ftp.unibe.ch/aiub/rinex/

Sažetak

Svojstva stohastičkoga modeliranja GPS-ovih podataka u šumskim područjima NAVSTAR GPS (Navigation System with Time and Ranging Global Positioning System) u današnje vrijeme omogućuje utvrđivanje šumskih granica uz pomoć dvostruko diferencijalnoga prijenosnoga prijamnika. U primjeni GPS-a udaljenost je između mjestâ izmjere mala. Kako je željena visoka točnost, podaci opažanja GPS-om obrađivani su rješavanjem dvoznačnih cjelobrojnih ciklusa. U šumskom području mogu se pojaviti ograničavajući čimbenici koji uzrokuju negativne utjecaje na signal GPSa, na primjer guste krošnje stabala ili veliki nagibi terena (Danskin i dr. 2009; Pirti i dr. 2010). U takvu slučaju Croat. j. for. eng. 33(2012)2

299

R. C. Erenoğlu

Performance of GPS Stochastic Modeling for Forest Environment (285–301)

navedene prepreke mogu utjecati na slab signal GPS-ovih satelita te prouzročiti teškoće pri izradi krajnjih proizvoda ovoga procesa, odnosno pouzdanih prostornih karata. To je zato što se obrada podataka koje šalje GPS zasniva na metodi najmanjih kvadratnih odstupanja (metoda LSE). Cilj je ovoga rada istražiti stohastičke GPS-ove modele u slučajevima gubitka signala zbog šumskoga okoliša, posebice za kratke udaljenosti. Da bi se to postiglo, uporabljeni su statistički modeli zasnovani na kvalitetnim pokazateljima signala i derivacijama razlike modela koji ukazuju na prijam i prepoznavanje GPS-ova signala. Utjecaj višestrukih putova (višestazja) i raspršenosti signala može se smanjiti pomoću stohastičkih modela raz vijenih pri obradi podataka, pa je točnost proračunskih parametara time povećana. U ovom je istraživanju upotrije bljena varijanca modelâ koji su korišteni u dosadašnjim istraživanjima za utvrđivanje jačine dvostrukih diferencijal nih opažanja. To su model jednakih varijanci (Hartinger i Brunner 1999), modeli zasnovani na kutnom nagibu satelita (Wang i dr. 1998; Collins i Langley 1999), modeli odnosa signala i raspršenosti (Schön i Brunner 2008) i modeli varijanci temeljeni na odnosu prijamnika i snage raspršenosti gustoće signala (Brunner 1999). Primjenom određenih postupaka proveden je precizno dizajniran pokus koji uzima u obzir iskrivljavanje signala i učinke višestazja zbog sjene drveća. Pokusi su bili obavljeni u regiji Samandira, Istanbul (Turska). Temeljna je linija bila spoj dviju nepomičnih lokacija, T1 i T2, a mjerenja su provedena na svakoj od njih pomoću GPS-ovih pri jamnika. Cilj je pokusa bio usporediti svojstva stohastičkoga GPS-ova modela pod utjecajem raspršivanja signala. U šumskim je područjima očekivano bilo raspršivanje signala zbog prepreka, kao što su grmlje, grane drveća i sama stabla. Da bi se osigurala višestazna okolina, mjesto T1 nalazilo se daleko od zapreke koju radi šuma. Podaci sa sig nalnim iskrivljavanjem bili su skupljani na mjestu T2 lociranom na granici šume. Iz rasporeda satelita vidljivo je da postoje neke razlike između vidljivosti satelita između mjestâ istraživanja. Postojao je pažljivo dizajniran primjer uključujući pogrešku višestazja. Temeljna je linija bila 20 m duga s visinskom razlikom od 1,970 m. Podaci su bili skupljani na spomenute dvije lokacije tijekom 6 sati između 4:30 i 10:30 sati svakih 30 sekundi. Tako je prikupljeno 720 opažanja tijekom jednoga dana terenskoga rada. Tijekom pokusa korišten je isti tip antene na obje lokacije identične orijentacije da bi se uklonile pogreške uzrokovane eventualnim utjecajem antena. Podaci su bili obrađivani pomoću programa BERNESE v. 5.0 i drugih prikladnih alata. Za razliku između mod elâ korišten je komplementarni alat da bi se dobila C/N0 vrijednost. Odgovarajući programi preveli su vrijednost SNR-a u C/N0. Naposljetku, u program MATLAB v. 7.0 uneseni su svi funkcionalni i stohastički modeli te obrađeni podaci nepomičnoga GPS-a. Atmosferski učinci sasvim su uklonjeni dvostruko diferencijalnom tehnikom zbog kratke osnovne linije, ali i niske visinske razlike (Teunissen i Kleusberg 1998). Dvostruki diferencijalni ostatak (DDR) satelita jasno daje informacije o kakvoći podataka. Na primjer, DDR od PRN 29 ukazuje na vrlo visoku kvalitetu s obzirom na vidljivost PRN 29 samo kod visokoga kuta nagiba satelita. S druge strane, DDR od PRN 25 ima nižu razinu kvalitete podataka zbog sjene koju čini šuma. Vrijednosti C/N0 signala zabilježene su na istraživanim lokacijama. Kao što je rečeno, C/N0 vrijednosti pružaju ključne informacije o utjecajima odraza i iskrivljavanja signala. Rezultati pokazuju da podaci skupljani na mjestu T1 imaju višu frekvenciju C/N0 zato što primanje signala GPS-ovih satelita nije bilo pod utjecajem krošanja stabala, kao što je to bio slučaj na mjestu T2. Pomoću stohastičkih modela bila je utvrđena osnovna linija između mjesta T1 i T2. Da bi se to postiglo, podaci su obrađivani geometrijski zasnovanim modelom dobivenom tehnikom dvostrukoga diferenciranja. Za visoku točnost pozicioniranja učinjene su pouzdane procjene cjelobrojne dvostruke razlike dvoznačnosti. U ovoj je studiji primijen jena metoda najmanjih kvadrata dvoznačnosti dekorelacijske prilagodbe (LAMBDA Teunissen 1993). Podaci sa T1 i T2 analizirani su pomoću šest različitih stohastičkih modela. Rezultati upućuju na nekoliko zaključaka koji su done seni pod utjecajem stohastičkih modela korištenih za procjenu parametara. U ovom je istraživanju primijenjeno stohastičko modeliranje GPS-ovih podataka. To su model jednakih težinskih varijanci, model ovisnosti kuta nagiba satelita, model SNR, sigma-ε model, sigma-Δ model i model predložene funkcije. Po rezultatima je očito da modeli sigma-ε i sigma-Δ daju optimalna rješenja. Na primjer, standardne su devijacije na razini od 4 do 5 mm za sjeveru i istočnu sastavnicu. To je zato što je procedura ponderirane procjene zasnovane na kakvoći signala uspješno sman jila negativne utjecaje krošanja stabala. Naposljetku, predložena je nova metoda stohastičkoga modeliranja kom biniranjem satelitskoga elevacijskoga kuta h i SNR vrijednosti. Predloženi model sin2h/SNR pruža bolje rezultate usporedbom s bilo kojim modelom 1/SNR. Zaključak je ovoga rada da krošnje stabala imaju velik utjecaj pri stohastičkom modeliranju u procesiranju GPSovih podataka. Mnogi ograničavajući čimbenici mogu se funkcionalno oblikovati i/ili izbjeći tehnikom dvostrukoga diferenciranja. Nadalje, očitanja prijenosnika mogu se modelirati stohastički usprkos utjecaju višestazja i rapršivanja signala koji se pojavljuju u šumi primjenom prikladnih funkcija, što se ogleda u kakvoći prijma signala GPS-a. Utjecaj raspršenosti signala može biti smanjen parametarskom procjenom, odnosno podešavanjem dvoznačnosti prijamnika i procjenom temeljnih sastavnica. Eksperimenti su također pokazali da modeli sima-ε i sigma-Δ daju optimalna

300

Croat. j. for. eng. 33(2012)2

Performance of GPS Stochastic Modeling for Forest Environment (285–301)

R. C. Erenoğlu

rješenja za lokacije ispod zastora krošanja drveća. Postupak ponderiranja zasnovan na vrijednostima C/N0 može uspješno premostiti negativan utjecaj krošanja stabala. Sukladno tomu, predstavljena je nova metoda sin2h/SNR kombiniranjem nagiba kuta satelita h i SNR vrijednosti. Predloženi model daje bolje rezultate u usporedbi s mode lima 1/SNR. Ključne riječi: GPS, šuma, učinkovitost, stohastički, ponderiranje

Author’s address – Autorova adresa:

Received (Primljeno): March 23, 2012 Accepted (Prihvaćeno): July 21, 2012 Croat. j. for. eng. 33(2012)2

R. Cüneyt Erenoğlu, PhD. e-mail: ceren@comu.edu.tr Canakkale Onsekiz Mart University Department of Geomatics Engineering 17020 Canakkale TURKEY

301

Original scietific paper – Izvorni znanstveni rad

Wood Density Impact on Hand-Arm Vibration Christian Rottensteiner, Petros Tsioras, Karl Stampfer Abstract – Nacrtak Despite technological advancements in machinery for timber harvesting, chainsaws are fre quently used in forest operations. In Austria 85% of the wood volume harvested (15 million m³) are cut by chainsaws. The two most frequently documented ergonomic threats during motor manual felling include exposure to noise and vibration. This paper presents the results of exposure to hand-arm vibration with focus on the impact of different density of wood species. Vibration exposure during crosscutting of Black poplar (oven-dry density of 400 kg/m³), Norway spruce (oven-dry density of 360 kg/m³) and European beech (oven-dry density of 700 kg/m³) was measured on three Husqvarna chainsaw models differing in size and power output. Measure ment and analysis of vibration were carried out in accordance with guidelines of ISO 5349-1 and ISO 5349-2. The results show that total values of unweighted root mean square (rms) vibration acceleration do not differ between tree species. Anyway, frequency-response curve of beech differed from that of poplar and spruce. Applying the weighting filter of ISO 5349-1, the frequency-weighted vibration of beech had higher values than those of poplar and spruce. No significant differences were found between poplar and spruce. Vibration values (measured on chainsaw rear handle) ranged from 4.06 m/s² to 4.92 m/s² for poplar, 4.38 m/s² to 5.66 m/s² for spruce, and 5.84 m/s² to 7.38 m/s² for beech, respectively. Keywords: chainsaw, hand-arm vibration, wood density, vibration, poplar, spruce, beech

1. Introduction – Uvod Chainsaw represents a very important tool for forest operations. Despite technological advancements that resulted in a wide range of machinery specialized for wood harvesting, chainsaws are frequently used in forest operations. This is due to multiple reasons such as low level of mechanization (Sessions 2007; Tsioras 2010), difficulty in obtaining capital for investments in new machinery (Uusitalo and Markkola 2006), unwillingness to invest in small-scale forestry practices due to high operational costs (Blinn et al. 1986; Wang et al. 2004) or large diameters of timber (Wang et al. 2004). However, the predominant reason is the fact that in many cases, felling operations can only be done by means of chainsaws because the steepness of the terrain makes the trafficability by machinery impossible and hence also the use of more mechanized operations (Heinimann 1999; Hall and Han 2006; Hittenbeck Croat. j. for. eng. 33(2012)2

2007). In Austria 85% of the wood volume harvested (15 million m³) are cut by chainsaws (BMLFUW, 2011). The use of chainsaws has been connected to a large number of ergonomic threats. Forest workers are affected by exhaust gas (Alander et al. 2005; Magnusson and Nilsson 2011), wood dust (Horvat et al. 2005; Puntaric et al. 2005; Kauppinen et al. 2006) and poor body posture causing lower back injuries (Hagen 1990; Hagen et al. 1998). However, the two most frequently documented threats during motor manual felling include exposure to noise and vibration. Extensive research has been done with regard to the measurement of the vibration during working with chainsaw (Cristofolini et al. 1990; Neitzel and Yost 2002; Pitts 2004), the effect of technological developments on the mitigation of hand-arm vibration (HAV) (Starck 1984; Koskimies et al. 1992), the prevalence of vibration induced white finger (VWF) and

303

C. Rottensteiner et al.

vascular disorders (Hellstrom and Andersen 1972; Futatsuka and Ueno 1986; Bovenzi 2008; Bovenzi et al. 2008), and musculoskeletal disorders due to HAV (Bovenzi et al. 1991; Kaewboonchoo et al. 1998). However, no studies have examined so far the effect of wood density on the exposure to vibration during work with the chainsaw. The originality of this study lies in the investigation of wood density impact on HAV. Different wood species have different wood structure and this is expected to differentiate the results of vibration exposure. Keeping these factors in consideration, the present study was undertaken in Lower Austria with the objective to increase our knowledge about the impact of different density of wood species on HAV. For this reason, HAV exposure during crosscutting operations was measured on poplar, spruce and beech.

2. Material and Methods – Materijal i metode 2.1 Study layout – Raspored istraživanja The experiment was carried out for investigating a potential influence on the vibration magnitude arising from chainsaws cutting tree species of differing wood density (Fig. 1). The data set was comprised of ten samples for each of the three species. This procedure was repeated for each one of the three chainsaw models used in the present study, resulting in a total of 90 measurement samples. The tree species under study were Black poplar (Populus nigra), Norway spruce (Picea abies) and Euro-

Wood Density Impact on Hand-Arm Vibration (303–312)

pean beech (Fagus sylvatica). Oven-dry density was found to be 400 kg/m³, 360 kg/m³ and 700 kg/m³, respectively. The oven-dry mass was determined on samples, according to the oven dry method (CEN/TS 14774-1:2004). The corresponding volume was achieved by dip-coating using Archimedes’ principle, according to which the upward force of a body immersed in water is equal to the weight of the amount of water the body displaces. Considering the fact that the specific gravity of water is 10 N/dm³ one can directly get the volume by reading the measurement result on a scale. Finally, oven-dry density for each sample was calculated by dividing the sample mass by its volume. The saw-logs of the above mentioned species that were used in this study had a diameter that ranged from 30 to 43 cm at the middle of their length. They were placed on a sawbuck 50 cm above ground, so that the operator was able to cut slices of four centimeters, until the designated sample time of two minutes was reached. ISO standard 5349-2 (ISO 2001b) stipulates that at least three samples of minimum one minute duration should be taken. Vibration was measured on the front and rear handle simultaneously. Each saw was equipped with a chain originally sharpened by the manufacturer, in order to eliminate the possibility of higher vibration values due to blunt chain. One new chain was mounted on each chainsaw, and it was used for all measurement samples for the same tree species.

2.2 Characteristics of chainsaws and operator Značajke motorne pile i sjekača Three brand-new Husqvarna chainsaws differing in size and power output were operated during all measurements (Table 1). The operator was a profes-

Fig. 1 Study layout Slika 1. Svojstva

304

Croat. j. for. eng. 33(2012)2

Wood Density Impact on Hand-Arm Vibration (303–312)

C. Rottensteiner et al.

Table 1 Equipment characteristics Tablica 1. Mjerna oprema Cutting tooth profile

Chain pitch, ‘’

# drive links

Gauge, mm

Oblik zubaca na lancu lanca

Korak lanca, ‘’

Br. pogonskih članaka

Mjera, mm

Semi-chisel

0.325

1.3

3.2

Semi-chisel

0.325

1.3

3.9

Semi-chisel

0.375

1.5

Weight, kg

Cylinder displacement, cm3

Masa, kg

Obujam cilindra, cm3

346 XP

5.0

50.1

2.7

357 XP

5.5

56.5

372 XP

6.1

70.7

Chainsaw type Vrsta motorne pile

Power output, Bar length, cm kW Duljina Izlazna snaga, vodilice, cm kW

cordance with ISO standard 5348 (ISO 1998). The accelerometer weight was 4.4 g. The orientation of the measurement axes and the accelerometer mount was done as described in the ISO standards 5349-1 (ISO 2001a) and 5349-2 (ISO 2001b). A Brüel&Kjaer LAN-XI 6-channel input module of the type 3050 with a frequency range from 0 to 51.2 kHz and 160 dB dynamic range was used for data recording. This portable data logger was driven by an accumulator; data were stored on a SD (Secure Digital) memory card for post-processing. Calibration was done before each measurement with the help of a Brüel&Kjaer calibration exciter of the type 4294, which produces an acceleration signal of 10 m/s² at a frequency of 159.2 Hz.

2.4 Data analysis – Obrada podataka

Fig. 2 Orientation of biodynamic axes Slika 2 Osi mjerenja vibracija sionally trained forest worker with experience in timber harvesting of eight years, who was requested to work at a normal work pace.

2.3 Vibration measurement – Mjerenje vibracija Two Brüel&Kjaer cubic triaxial piezoelectric accelerometers of the type 4524-B were used for simultaneous vibration measurements in three perpendicular directions on the chainsaw front and rear handle (Fig. 2). These transducers were fixed by a mounting clip stuck with ceramic adhesive glue. That ensures that there is no damping between the chainsaw and transducers and vibrations are correctly measured in acCroat. j. for. eng. 33(2012)2

The collected vibration data were analyzed with the Brüel&Kjaer PULSE LabShop 14.1 software. More specifically, the root-mean-square (rms) acceleration, at one-third octave bands in the frequency range between 6.3 and 1250 Hz, was used for the calculation of the weighted rms accelerations for each axis (ahwx, ahwy, ahwz). The weighting factors according to ISO standard 5349-1 (ISO 2001a) were applied. The total vibration value ahv (1) was calculated from the frequency weighted acceleration of all three axes with the following equation: (1) Identification of potential differences between the examined factors (wood density, type of chainsaw) was conducted with the help of analysis of variance (ANOVA), which is included in the software package PASW Statistics 18. Post-hoc analysis for investigating potential differences between tree species and chainsaw type was done using Bonferroni’s test. Significance level for all tests was set at 5%.

305

C. Rottensteiner et al.

Wood Density Impact on Hand-Arm Vibration (303–312)

3. Results – Rezultati

ences between the tree species (Table 2). As seen in Tab le 3, vibration values of chainsaws for the rear handle exceeded those for the front handle. Differences between handles were statistically significant. Values mea sured on Husqvarna 346 XP were on the same le vel with those of Husqvarna 372 XP, while on Husqvarna 357 XP higher vibration acceleration was found.

3.1 Wood density impact on frequency-unweighted HAV – Utjecaj gustoće drva na frekvencije nevrednovanih ubrzanja u sustavu šaka–ruka ANOVA and post-hoc analysis of frequency-unweighted vibration acceleration revealed no differTable 2 Bonferroni’s test for frequency-unweighted vibration Tablica 2. Bonferronijev test za frekvencije nevrednovanih ubrzanja Std. error

Glavna razlika, I–J

Standardna pogreska

Značajno

Spruce – Smreka

0.3214

3.38344

Beech – Bukva

–0.3796

Poplar – Topola

Tree species, I

Tree species, J

Vrsta stabala, I

Poplar – Topola Spruce – Smreka Beech – Bukva

95% Confidence Interval

Mean Difference, I–J

95 % interval pouzdanosti

Sig.

Lower bound

Upper bound

Donja granica

Gornja granica

1.000

–7.8640

8.5068

3.35667

1.000

–8.5003

7.7410

–0.3214

3.38344

1.000

–8.5068

7.8640

Beech – Bukva

–0.7010

3.26234

1.000

–8.5935

7.1914

Poplar – Topola

0.3796

3.35667

1.000

–7.7410

8.5003

Spruce – Smreka

0.7010

3.26234

1.000

–7.1914

8.5935

357 XP

–28.3683(*)

3.39245

0.000

–36.5755

–20.1611

372 XP

0.2283

3.23457

1.000

–7.5969

8.0536

346 XP

28.3683(*)

3.39245

0.000

20.1611

36.5755

372 XP

28.5967(*)

3.39245

0.000

20.3895

36.8039

346 XP

–0.2283

3.23457

1.000

–8.0536

7.5969

357 XP

–28.5967(*)

3.39245

0.000

–36.8039

–20.3895

Chainsaw, I

Chainsaw, J

Motorna pila, I

Motorna pila, J

346 XP 357 XP 372 XP

Table 3 Frequency-unweighted vibration acceleration (rms) in m/s2 Tablica 3. Frekvencije nevrednovanih ubrzanja (m/s2) Husqvarna 346 XP

Husqvarna 357 XP Mean

Arit. sred.

Stand. dev.

Min

Max

Maks.

Min.

Maks.

–

52.08

3.29

47.70

56.10

6.41

99.40

117.00

72.54

5.55

64.90

78.90

65.74

2.92

61.90

68.50

49.08

2.27

46.10

51.60

82.70

121.60

9.96

110.00

137.00

71.68

9.53

63.50

87.40

52.40

56.90

57.28

3.81

53.20

61.30

55.36

3.86

49.00

59.20

75.30

83.30

109.30

8.42

98.50

117.00

81.60

8.17

71.30

89.30

Mean

Maks.

Arit. sred.

47.60

56.50

5.18

70.70

49.10

2.04

Rear – Stražnja

74.45

Beech

Front – Prednja

Bukva

Rear – Stražnja

Mean

Ručka

Arit. sred.

Poplar

Front – Prednja

Topola

Stand. dev.

Min

Max

Min.

–

83.60

107.68

46.40

51.00

6.40

67.20

53.78

1.79

78.10

3.43

Stand. dev.

Min

Max

Min.

51.98

4.05

Rear – Stražnja

76.10

Spruce

Front – Prednja

Smreka

Tree species

Handle

Vrsta stabala

Husqvarna 372 XP

Values were excluded because of damaged mounting clip – Izostavljene vrijednosti zbog oštećenja ugrađenih spojnica

306

Croat. j. for. eng. 33(2012)2

Wood Density Impact on Hand-Arm Vibration (303–312)

C. Rottensteiner et al.

Table 4 Bonferroni’s test for frequency-weighted vibration Tablica 4. Bonferronijev test za frekvencije vrednovanih ubrzanja

Tree species, I

Tree species, J

Mean Difference, I–J

Vrsta stabala, I

Glavna razlika, I–J

Poplar – Topola Spruce – Smreka Beech – Bukva

Spruce – Smreka

0.0886

95% Confidence Interval

Std. error

95 % interval pouzdanosti

Sig.

Standardna pogreska

Značajno

0.15478

1.000

Lower Bound

Upper Bound

Donja granica

Gornja granica

–0.2858

0.4630

Beech – Bukva

–1.4969(*)

0.15355

0.000

–1.8684

–1.1254

Poplar – Topola

–0.0886

0.15478

1.000

–0.4630

0.2858

Beech – Bukva

–1.5855(*)

0.14924

0.000

–1.9466

–1.2245

Poplar – Topola

1.4969(*)

0.15355

0.000

1.1254

1.8684

Spruce – Smreka

1.5855(*)

0.14924

0.000

1.2245

1.9466

357 XP

–0.3283

0.15519

0.108

–0.7038

0.0471

Chainsaw, I

Chainsaw, J

Motorna pila, I

Motorna pila, J

346 XP 357 XP 372 XP

372 XP

0.9550(*)

0.14797

0.000

0.5970

1.3130

346 XP

0.3283

0.15519

0.108

–0.0471

0.7038

372 XP

1.2833(*)

0.15519

0.000

0.9079

1.6588

346 XP

–0.9550(*)

0.14797

0.000

–1.3130

–0.5970

357 XP

–1.2833(*)

0.15519

0.000

–1.6588

–0.9079

Table 5 Frequency-weighted vibration acceleration (rms) in m/s2 Tablica 5. Frekvencije vrednovanih ubrzanja (m/s2) Husqvarna 346 XP

Mean

Ručka

Arit. sred.

Rear – Stražnja

4.02

Tree species

Handle

Vrsta stabala Poplar

Husqvarna 357 XP Mean

Maks.

Arit. sred.

4.40

Stand. dev.

Min

Max

Min.

0.29

3.70

Husqvarna 372 XP Mean

Sd Stand. dev.

Min

Max

Maks.

Arit. sred.

Min.

Maks.

–

3.46

0.21

3.20

3.70

Stand. dev.

Min

Max

Min.

–

Topola

Front – Prednja

4.92

0.62

4.40

5.90

5.88

0.34

5.50

6.40

4.06

0.32

3.60

4.40

Spruce

Rear – Stražnja

3.78

0.31

3.50

4.20

4.28

0.30

4.10

4.80

3.42

0.51

2.80

4.20

Smreka

Front – Prednja

4.95

0.79

4.30

6.10

5.66

0.81

5.20

7.10

4.38

1.26

3.50

6.60

Beech

Rear – Stražnja

6.34

0.57

5.50

7.10

5.24

0.37

4.70

5.70

4.38

0.49

4.00

5.20

Bukva

Front – Prednja

7.38

0.65

6.80

8.40

6.64

0.42

6.10

7.20

5.84

1.21

4.90

7.90

Values were excluded because of damaged mounting clip – Izostavljene vrijednosti zbog oštećenja ugrađenih spojnica

3.2 Wood density impact on frequency-weighted HAV – Utjecaj gustoće drva na frekvencije vrednovanih ubrzanja u sustavu šaka–ruka Differences in frequency-weighted vibration acceleration between beech and poplar, as well as, between beech and spruce have been verified by Bonferroni’s test (Table 4) and are also shown in the measurement Croat. j. for. eng. 33(2012)2

results (Table 5). These findings might be explained by differences in wood density and wood structure, respectively. However, vibration values of poplar and spruce were not found statistically different, despite the fact of comparing a coniferous with a deciduous species. These investigations supported the assumption that wood density alone and independent of the species, has an impact on HAV. Spruce and poplar had

307

C. Rottensteiner et al.

Wood Density Impact on Hand-Arm Vibration (303–312)

Fig. 3 Frequency-unweighted vibration acceleration for Husqvarna 346 XP for different tree species Slika 3. Frekvencije nevrednovanih ubrzanja za motornu pilu Husqvarna 346 XP i različite vrste drva

Fig. 4 Frequency-weighted vibration acceleration for Husqvarna 346 XP and weighting factor for HAV Slika 4. Frekvencije vrednovanih ubrzanja za motornu pilu Husqvarna 346 XP i vrednovani faktor u sustavu šaka–ruka comparable wood density (360 kg/m3 and 400 kg/m3, respectively), while that of beech was almost twice as high (700 kg/m3). The obvious discrepancy between weighted and unweighted vibration values appeared at first glance, as presented for the chainsaw Husqvarna 346 XP in

308

Fig. 3. Vibration in the beech sawlogs had higher values in the lower frequency range (6.3–100 Hz) than the other two species. Nevertheless, these were just minor differences in the absolute values. The absolute values are equal to the area below the graphs, and this applied for all three species of comparable level. HowCroat. j. for. eng. 33(2012)2

Wood Density Impact on Hand-Arm Vibration (303–312)

ever, major differences were evident between the observed tree species with regard to frequency-weighted vibration, due to the fact that the application of the weighting-filter Wh rates the lower frequencies as more important than the higher ones (Fig. 4). The weighted vibration mean values regardless of tree species were found to be 5.21 m/s², 5.54 m/s², and 4.26 m/s², respectively, for Husqvarna 346 XP, Husqvarna 357 XP and Husqvarna 372 XP. Post-hoc analysis showed the difference only between Husqvarna 372 XP and the other two chainsaws. The reason probably lies in the fact that Husqvarna 346 XP had the highest frequency-weighted vibration values for beech. For instance, if one just compares the mean values of measurements on spruce (4.27 m/s² for Husqvarna 346 XP, 4.97 m/s² for Husqvarna 357 XP and 3.90 m/s² for Husqvarna 372 XP), the same results appear as for frequency-unweighted vibration (no differences between Husqvarna 346 XP and Husqvarna 372 XP). Anyway, the main focus of this study was to reveal potential differences between tree species and density, respectively, which was found for all observed chainsaws.

4. Discussion – Rasprava The experiment revealed that wood density influences the magnitude of weighted vibration, while no differences in total values for unweighted HAV were found. This is due to differences in the frequency range where vibration appeared. Lower frequencies are weighted by higher factor values because of resonance characteristics of the human body. Vibration acceleration in spruce and poplar in the range of 6.3 to 100 Hz (one-third-octave band center frequencies) were lower than that of beech. On the contrary, in the frequency range greater than 100 Hz, vibration acceleration of spruce and poplar had slightly higher magnitudes than that of beech. This is assumed to be caused by differences in wood density and wood structure, respectively. Modulus of elasticity might explain these differences along the frequency spectrum. It is 8.9 kN/mm² for poplar, 11 kN/mm² for spruce and 14.4 kN/mm² for beech (Sell 1997). Sound velocity in different wood species can be derived by dividing the species modulus of elasticity by its density, and by taking the square root of that quotient (Burmester 1965). The radial sound velocity is increasing, as wood density increases, a fact that is directly correlated to higher resonance frequencies. Buksnowitz (2006) measured the resonance frequency in radial direction of spruce (Picea abies) with an average of 63.25 Hz (minimum 39.60 Hz, Croat. j. for. eng. 33(2012)2

C. Rottensteiner et al.

maximum 79.74 Hz). These facts might explain the existence of local maximum values for spruce, at a onethird-octave band with center frequency of 50 Hz. In beech and poplar these peak values appeared at 63 Hz, probably due to higher wood density. The higher vibration amplitude of beech was probably due to inferior damping characteristics of higher wood density (Holz 1973). However, as wood resonance is affected by various parameters other than density and modulus of elasticity (Buksnowitz 2006), it could be assumed that differences in frequency-response curves were due to structural differences in wood species. Wood density impact on weighted HAV was found statistically significant for all three chainsaws. The frequency spectra were pretty much the same for all saws according to the chainsaw engine characteristics. Differences were found with regard to vibration magnitude. Peak values and harmonic components appeared at 160 Hz (engine speed 9.600 rotations per minute, i.e. the maximum power speed for all three saws according to manufacturer’s declaration), 320, 640 and 1280 Hz, respectively. Statistical analysis for frequencyunweighted vibration showed no differences between Husqvarna 346 XP and Husqvarna 372 XP. Medium sized Husqvarna 357 XP showed the highest vibration values. For frequency weighted-vibration that relationship could only be measured for spruce. Using all measurements regardless of the tree species only Husqvarna 372 shows differences compared to the other two chainsaws. With respect to the manufacturer’s vibration emission declaration in accordance with ISO standard 7505 (ISO 1986), Husqvarna 357 XP has higher values than Husqvarna 346 XP and 372 XP, respectively. Anyway, this is just a comparison of the published mean values without any statistical analysis. Considering this drawback, it cannot be stated if there are any real differences between the published values. Idling speed is the same for all three chainsaws with 2700 rotations per minute according manufacturer’s declaration. Rottensteiner and Stampfer (2012) found that frequency-weighted vibrations during idling for all three saws were of comparable level; therefore idling was not included in the statistical model. According to the Directive 2002/44/EC, employers have to assess the risk of mechanical vibration to which workers are exposed. The assessment of the level of exposure to hand-arm-vibration is based on calculation of the daily vibration exposure value A(8). Daily exposure values are normalized to an eight-hour reference period, which represents the total value of the eight-hour energy equivalent vibration for a worker including all hand-arm vibration exposures during

309

C. Rottensteiner et al.

the day. In this study only the influence of wood density on vibration during cutting activities was examined. Nevertheless, this working element counts for around 25% of the total working time (Laier 2011). Therefore, the wood density is also fundamental when calculating the A(8) value. If one reaches the limit value working in a spruce or poplar stand for eight hours, this implicates that vibration exposure for the same worker was above the limit value in a beech stand. Such a case entails a reduction of time performing chainsaw work. In further research, it would be interesting to repeat the experiment in other tree species, work phases (e.g. delimbing), chainsaw types (different manufacturers) as well as different operators.

5. References – Literatura Alander, T., Antikaine, E., Raunemaa, T., Elonen, E., Rautiola, A., Torkkell, K., 2005: Particle emissions from a small twostroke engine: Effects of fuel, lubricating oil, and exhaust aftertreatment on particle characteristics. Aerosol Science and Technology 39(2): 151–161. Blinn, C. R., Sinclair, S. A. Hassler, C. C, Mattson, J. A., 1986: Comparison of productivity, capital, and labor efficiency of five timber harvesting systems for northern hardwoods. Forest Products Journal 36(10): 63–69. BMLFUW 2011: Holzeinschlagsmeldung über das Kalenderjahr 2010 (Timber Harvesting Report 2010). Bundesministerium für Land- und Forstwirtschaft, Umwelt und Wasserwirtschaft. (Ministry of Agriculture and Forestry) (Ed.), Wien. Bovenzi, M., Zadini, A., Franzinelli, A., Borogogni, F., 1991: Occupational musculoskeletal disorders in the neck and upper limbs of forestry workers exposed to hand-arm vibration. Ergonomics 34(5): 547–562. Bovenzi, M., 2008: A follow up study of vascular disorders in vibration-exposed forestry workers. International Archives of Occupational and Environmental Health 81(4): 401–408. Bovenzi, M., D’Agostin, F., Rui, F., Negro, C., 2008: A longitudinal study of finger systolic blood pressure and exposure to hand-transmitted vibration. International Archives of Occupational and Environmental Health 81(5): 613–623. Buksnowitz, C., 2006: Resonance wood of Picea Abies. A review-based interrelation model supported by an index-like compilation of scientific research performed on resonance wood quality affecting parameters. Doctoral thesis at the Institute of Wood Sciences and Technology. University of Natural Resources and Life Sciences Vienna. 211 p. Burmester, A., 1965: Zusammenhang zwischen Schallgeschwindigkeit und morphologischen und mechanischen Eigenschaften von Holz (Relationship between sound veloc-

310

Wood Density Impact on Hand-Arm Vibration (303–312) ity and morphological and mechanical parameters of wood). Holz als Roh- und Werkstoff 23(6): 227–236. CEN/TS 17774-1:2004: Solid biofuels- Methods for determination of moisture content – Oven dry method-Part 1: Total moisture – Reference method. European Committee for Standardization, Brussels. 10 p. Cristofolini, A., Pollini, C., Maggi, B., Costa, G., Colombini, D., Occhipinti, E., Bovenzi, M., Perettu, S., 1990: Organizational and ergonomic analysis of forest work in the Italian alps. International Journal of Industrial Ergonomics 5(3): 197–209. Directive 2002/44/EC. Council Directive 2002/44/EC of 25 June 2002 on the minimum health and safety requirements regarding the exposure of workers to the risks arising from physical agents (vibration). Futatsuka, M., Ueno, T., 1986: A follow-up study of vibration-induced white finger due to chain-saw operation. Scandinavian Journal of Work, Environment and Health 12(4): 304–306. Hagen, K. B., 1990: Biomechanical Analysis of Spinal Load in Motor-Manual Cutting. International Journal of Forest Engineering 2(1): 39–41. Hagen, K. B., Magnus, P., Vetlesen, K., 1998: Neck/shoulder and low-back disorders in the forestry industry: relationship to work tasks and perceived psychosocial job stress. Ergonomics 41(10): 1510–1518. Hall, R., Han, H. S., 2006: Improvements in value recovery through low stump heights: Mechanized versus manual felling. Western Journal of Applied Forestry 21(1): 33–38. Heinimann, H., 1999: Ground-based Harvesting Technologies for Steep Slopes. In: Proceedings of the International Mountain Logging and 10th Pacific Northwest Skyline Symposium. Sessions and Chung (ed.). March 28 – April 1, Corvallis, Oregon 19 p. Hellstrom, B., Andersen, K. L., 1972: Vibration injuries in Norwegian forest workers. British Journal of Industrial Medicine 29(3): 255–263. Hittenbeck, J., 2007: Limits of Wheelbased Timber Harvesting in Inclined Areas. In: Institute of Forest Engineering, University of Natural Resources and Applied Life Sciences Vienna (ed.): Austro 2007/FORMEC 2007. Meeting the Needs of Tomorrows` Forests: New Developments in Forest Engineering. Proceedings CD-Rom. Vienna. Holz, D., 1973: Untersuchungen an Resonanzholz. 5. Mitteilung: Über bedeutsame Eigenschaften nativer Nadel- und Laubhölzer im Hinblick auf mechanische und akustische Parameter von Piano-Resonanzböden (Analysis of resonance wood. 5th Report: Characteristics of native soft – and hardwood species with focus on mechanical and acoustical parameters of piano-sound boards). Holztechnologie 14(4): 195–202. Horvat, D., Čavlović, A., Zečić, Ž., Šušnjar, M., Bešlić, I., Madunić-Zečić, V., 2005: Research of fir-wood dust concenCroat. j. for. eng. 33(2012)2

Wood Density Impact on Hand-Arm Vibration (303–312) tration in the working environment of cutters. Croatian Journal of Forest Engineering 26 (2): 85–90. ISO 1986: ISO standard 7505. Forestry Machinery – Chain saws – Measurement of hand-transmitted vibration. International Organization for Standardization, Geneva. 8 p. ISO 1998: ISO standard 5348. Mechanical vibration and shock – Mechanical mounting of accelerometers. International Organization for Standardization, Geneva. 12 p. ISO 2001a: ISO standard 5349-1. Mechanical vibration – Measurement and evaluation of human exposure to hand-transmitted vibration – Part 1: General requirements. International Organization for Standardization, Geneva. 24 p. ISO 2001b: ISO standard 5349-2. Mechanical Vibration – Measurement and evaluation of human exposure to handtransmitted vibration. Part – 2: Practical guidance for measurement at the workplace. International Organization for Standardization, Geneva. 39 p. Kauppinen, T., Vincent, R., Liukkonen, T., Grzebyk, M., Kauppinen, A., Welling, I., Arezes, P., Black, N., Bochmann, F., Campelo, F., Costa, M., Elsigan, G., Goerens, R., Kikemenis, A., Dromhout, H., Miguel, S., Mirabelli, D., McEneany, R., Pesch, B., Plato, N., Schlänssen, V., Schulze, J., Sonntag, R., Verougstraete, V., Vicente, M., Wolf, J., Zimmermann, M., Husgafvel-Pursiainen, K., SaVolainen, K., 2006: Ocupational Exposure to Inhalable Wood Dust in the Member States of the European Union, Ann. Occup., Hyg. 50(6): 549–561. Koskimies, K., Pyykko, I., Starck, J., Inaba, R., 1992: Vibration Syndrome among Finnish Forest Workers between 1972 and 1990. International Archives of Occupational and Environmental Health 64(4): 251–256. Kaewboonchoo, O., Yamamoto, H., Miyai, N., Mirbod, S. M., Morioka, I., Miyashita, K., 1998: The standardized nordic questionnaire applied to workers exposed to hand-arm vibration. Journal of Occupational Health 40(3): 218–222. Laier, J., 2011: Oral information of a professional forest worker (6.6.2011).

C. Rottensteiner et al. Magnusson, R., Nilsson, C., 2011: The influence of oxygenated fuels on emissions of aldehydes and ketones from a two-stroke spark ignition engine. Fuel 90(3): 1145–1154. Neitzel, R., Yost, M., 2002: Tasks-based assessment of occupational vibration and noise exposures in forestry workers. American Industrial Hygiene Association Journal 63(5): 617–627. Pitts, P., 2004: Hand-arm vibration emission of chainsaws – comparison with vibration exposure. Health and Safety Laboratory UK, Derbyshire. 53 p. Puntarić, D., Kos, A., Smit, Z., Zečić, Ž., Šega, K., Beljo-Lučić, R., Horvat, D., Bosnir, J., 2005: Wood dust exposure in wood industry and forestry. Collegium Antropologicum 29(1): 207–211. Rottensteiner, C., Stampfer, K., 2012: Evaluation of operator exposure to chainsaws equipped with a Kesper safety bar. Scandinavian Journal of Forest Research iFirst article: 1–8. Sell, J., 1997: Eigenschaften und Kenngrößen von Holzarten. Zürich (Characteristics and parameters of wood species). Baufachverlag 88 p. Sessions, J., 2007: Appropriate Harvesting Technology. In: Sessions, J. (Ed) Harvesting Operations in the Tropics. Springer, Berlin- Heidelberg. Starck, J., 1984: High impulse acceleration levels in handheld vibratory tools. An additional factor in the hazards associated with the hand-arm vibration syndrome. Scandinavian Journal of Work, Environment and Health 10(3): 171–178. Tsioras, P. A., 2010: Perspectives of the forest workers in Greece. iForest 3: 118–123. Uusitalo, J., Markkola, J.M., 2006: Entrepreneurship in Forestry – Is it worth activating? Forestry Studies 45: 67–73. Wang, J. X., Long, C., McNeel, J., Baumgras, J., 2004: Productivity and cost of manual felling and cable skidding in central Appalachian hardwood forests. Forest Products Journal 54(1): 45–51.

Sažetak

Utjecaj gustoće drva na vibracije u sustavu šaka-ruka Unatoč tehnološkomu napretku i razvoju potpuno strojne sječe, motorne pile lančanice i dalje su najčešći alat pri sječi šuma. U Austriji se 85 % drvnoga obujma (15 milijuna m³) posiječe motornim pilama lančanicama. Dvije najčešće ergonomske prijetnje tijekom sječe motornim pilama lančanicama jest izloženost buci i vibracijama. Prove dena su opsežna istraživanja o mjerenju vibracija nastalih tijekom rada s motornom pilom lančanicom, učinku tehnoloških dostignuća na ublažavanje nastalih vibracija na sustav šaka–ruka (HAV), o bolesti bijelih prstiju (VWF), vaskularnih bolesti te mišićno-koštanih poremećaja nastalih zbog izloženosti vibracijama prilikom rada na motornoj pili lančanici. Međutim, nema istraživanja o povezanosti gustoće drva i izloženosti vibracijama tijekom rada s motornom pilom lančanicom. Različite vrste drva imaju različitu unutarnju građu pa se i očekuju razlike u vrijednostima izmjerenih Croat. j. for. eng. 33(2012)2

311

C. Rottensteiner et al.

Wood Density Impact on Hand-Arm Vibration (303–312)

vibracija. Izloženost vibracijama mjerena je tijekom sječe crne topole (gustoća suhe tvari 400 kg/m³), obične smreke (gustoća suhe tvari 360 kg/m³) i obične bukve (gustoća suhe tvari 700 kg/m³) na tri modela motornih pila proizvođača Husqvarna, koji su se razlikovali u veličini i snazi. Za simultano mjerenje vibracija u tri međusobno okomita smjera na prednjim i stražnjim ručkama motornih pila lančanica korišten je vibrometar Brüel & Kjaer (troosni akcelerom etar 4524-B i LAN-XI 6-kanalni ulaz 3050). Mjerenje i analiza podataka provedeni su u skladu s normama ISO 5349-1 i ISO 5349-2. Obrada podataka učinjena je pomoću računalnoga paketa Brüel & Kjaer Pulse LabShop 14,1. Istraživanje mogućih razlika među čimbenicima (gustoća drva, vrsta motorne pile) provedena je uz pomoć anal ize varijance (ANOVA). Post-hoc analiza za istraživanje moguće razlike između pojedinih vrsta drveća i vrsta motorne pile lančanice učinjena je pomoću Bonferronijeva testa. Rezultati pokazuju da ne postoji razlika ukupne vrijednosti usrednjenih kvadrata (RMS) nevrednovanoga ubrzanja između različitih vrsta drveća. U svakom slučaju, frekvencijska krivulja bukve razlikuje se od frekvencijskih krivulja topole i smreke. Primjena propusnika iz norme ISO 5349-1 donijela je veće vrijednosti vrednovanoga ubrzanja kod bukve od vrijednosti izm jerenih na topoli i smreci. Nema značajne razlike između topole i smreke. Ovo istraživanje podupire pretpostavku da gustoća drva utječe na vibracije u sustavu šaka–ruka. Izmjerene vibracije bile su više na stražnjoj nego na prednjoj ručki motornih pila lančanica. Razlike između mjerenja na obje ručke (za vrednovano i nevrednovano ubrzanje) statistički su značajne. Nevrednovano ubrzanje izmjereno na motornoj pili lančanici Husqvarna 346 XP na istoj je razini s modelom 372 XP, dok je model 357 XP razvio više vrijednosti ubrzanja. Gustoća drva također treba biti jedan od temelja pri izračunu dnevne izloženosti radnika vibracijama – A(8). Za procjenu rizika od vibracija nastalih pri radu motornom pilom lančanicom mora se uzeti veća vrijednost (mjeren je na stražnjoj ručki). Vrijednosti su se kretale, ovisno o vrsti motorne pile, od 4,06 m/s² do 4,92 m/s² kod topole, 4,38 m/s² do 5,66 m/s² kod smreke i 5,84 m/s² do 7,38 m/s² kod bukve. Razlike u vibracijama između različitih vrsta drveća koje imaju i različitu gustoću drva nastaju zbog povezanosti značajki rezonatnosti drva i vlage u drvu s gustoćom drva i modulom elastičnosti. Ključne riječi: motorna pila lančanica, gustoća drva, vibracije, topola, smreka, bukva

Authors’ address – Adresa autorâ: Christian Rottensteiner, MSc. e-mail: christian.rottensteiner@boku.ac.at Prof. Karl Stampfer, PhD. e-mail: karl.stampfer@boku.ac.at University of Natural Resources and Life Sciences Vienna Peter-Jordan-Straße 82 1190 Vienna AUSTRIA

Received (Primljeno): June 20, 2012 Accepted (Prihvaćeno): August 18, 2012

312

Petros Tsioras, PhD. e-mail: ptsioras@for.auth.gr Aristotle University (POB 227) GR-541 24 Thessaloniki GREECE Croat. j. for. eng. 33(2012)2

Original scietific paper – Izvorni znanstveni rad

An Adaptive Network-based Fuzzy Inference System for Rock Share Estimation in Forest Road Construction Ismael Ghajar, Akbar Najafi, Seyed Ali Torabi, Mashalah Khamehchiyan, Kevin Boston Abstract – Nacrtak This paper presents a new Rock Share Estimation (RSE) procedure that can estimate the cost of forest road construction. One of the key elements of the total cost in road construction is the cost of embankment. The proportion of the rock directly influences the price of this activity. Hence, a reliable estimation of rock proportion should be made within the entire project area, especially in rocky areas. The objective of the study is to introduce a practical expert system to estimate the share of rock as a function of terrain slope and geological formations using the Adaptive Network – based Fuzzy Inference System (ANFIS) and Analytic Hierarchy Process (AHP). This approach can be very useful first to show the variability of rock proportion and second to model the excavation costs in an area, which are essential for planning forest roads. This study treats geological composition as a decision variable that is solved by AHP method and applies the ANFIS to model and predict the share of rock in different physiographic and geological conditions. In order to investigate the impact of change in membership functions (MF), four types of MFs were adopted to generate the hybrid RSE–ANFIS models. Further more, to show the applicability of the proposed approach, the optimum model was applied to a mountainous forest, where additional forest road network should be constructed in the future periods. Keywords: Rock Proportion, ANFIS, AHP, Forest Road Cost, Membership Function

1. Introduction – Uvod Estimation of rock proportion of subsoil as a real life problem is an important element in both the design and construction stages of road projects. All road planners and contractors prefer to construct the road in a soft terrain with the least rock excavation as this is the most cost-effective construction. However, it is very hard to exactly calculate the rock ratio of subsoil before the excavation of a project begins. In this regard, a number of methods have been suggested for assessment of slope stability and excavation quantities (Hoek and Bray 1981; Goodman 1989; Pettifer and Fookes 1994). Additionally, rock mass classification concept, which provides quantitative data and guidelines for engineering purposes, has been applied extensively for tunneling and underground excavation. Croat. j. for. eng. 33(2012)2

The main shortcoming of the existing traditional rock mass classification systems such as Rock Structure Rating, RSR (Wickham et al. 1972), the Rock Mass Rating, RMR (Bieniawski 1975), and the Q-system (Barton et al. 1974) is that they ignored the regional and local geological features and rock properties, and they were developed with the fixed weight for each rating factor (Liu and Chen 2007). On the other hand, other researchers that have developed tools to estimate the rock share related to forest topics are so limited (e.g. Inaba et al. 2001; Stuckelberger et al. 2006). In most prior studies, the share of rock is introduced as a function of terrain slope and geology information. Inaba et al. (2001) developed a numerical model to estimate the share of rock excavation volume. They assigned a coefficient for each geological unit and used this coefficient along with the slope and road crown

313

I. Ghajar et al.

An Adaptive Network-based Fuzzy Inference System for Rock Share Estimation ... (313–328)

width as inputs of the model to estimate rock portion for each geological unit. Stuckelberger et al. (2006) have adopted the current model and the above mentioned coefficients to estimate rock excavation volume and cost for each geological unit. They calculated the rock excavation cost as a function of earth embankment costs to be used as one of the cost elements of forest road construction cost model. However, considering a fixed coefficient for each type of geological unit regardless of local conditions could lead to inaccurately estimating the rock share and consequently inaccurately estimating the construction costs. The approach performed in practice in Iran is as follows: after visiting the proposed alignment of forest road, three general classes of rock, soft, medium, and hard, are assigned to several parts of the road project by experts. In fact, these linguistic values are used to describe the difficulty levels of the earthwork and to calculate the base price of project appraisal. As a result, providing a framework that could be practical and at the same time have a justifiable analytic foundation is essential for forest managers. From another perspective, developing a model to estimate the share of rock is problematic due to the uncertainty associated with geological information and environmental variables. The numerical modeling and optimization approaches have long been employed in forest researches worldwide. The traditional simulation and optimization tools are appropriate when data are known well enough, while in many real-world problems, there are many uncertain variables, and/or vague and ambiguous input data that should be handled for modeling. A computing system that has the ability to analyze these kinds of data should be more flexible and adaptive than the traditional approaches. In other words, a Real-World Computing (RWC) system should be capable of distributed representation of information, massively parallel processing, learning and self-organization to achieve enough flexibility in information processing (Sreekanth et al. 2010). In this relation, soft computing techniques as the open, robust, and real-time processing systems can be adopted efficiently to cope with the RWC systems. Hereafter, a brief introduction of relevant techniques and literature are provided. Fuzzy logic (Zadeh 1965) is increasingly used in various fields of science and technology for prediction purposes (Gail et al. 2002). Fuzzy systems are stable, easily tunable and could be validated conventionally. One of the significant advantages of linguistic methodology in fuzzy rule based systems is that a welldefined physical relationship is not required to systematically convert an input to an output (ASCE Task

314

Committee on Applications of Artificial Neural Networks in Hydrology 2000). These properties make them sufficiently flexible for solving the real time problems especially when the input or output data are defined by several linguistic values. Fuzzy sets theory has been applied in forest management (Mendoza and Sprouce 1989; Mendoza et al. 1993; Zandik 2006), water runoff, sediment yield, and recreation (Tecle et al. 1994), resource allocation (Ghajar et al. 2010), and forestry planning problems (Kangas et al. 2006). Artificial Neural Networks (ANNs) are another type of soft computing and data driven techniques that, because of their heuristic problem-solving capabilities, have been applied successfully in many fields of geological engineering problems (Shahin et al. 2008). Among them, ANNs have been employed in predicting the settlement and bearing capacity of shallow foundations (Shahin et al. 2005; Padmini et al. 2008), applications concerning earth retaining structures (Kung et al. 2007), site characterization (Najjar and Basheer 1996), mining (Rankine and Sivakugan 2005), groutability of soils (Tekin and Akbas 2010) and many other problems. A comprehensive overview of ANN application in geological engineering problems can be found in Shahin et al. (2008). To gain more efficiency from fuzzy logic and ANNs, a combined approach called neuro-fuzzy was developed by Jang (1993). Neuro-fuzzy systems are fuzzy systems that use ANN theory in order to determine their properties (i.e. the respective fuzzy sets and fuzzy rules) by processing of available data set (Andrews et al. 1995). In this regard, »the adaptive network-based fuzzy inference system (ANFIS), which has shown significant results in modeling nonlinear functions« (Wang et al. 2010) is of particular interest. The important contribution of Jang (1993) to ANFIS development was the establishment of the universal approximant nature of ANFIS, and the functional equivalency of the Sugeno fuzzy inference systems with radial neural networks, providing the essential theoretical support for the practical application of ANFIS to nonlinear system modeling (Roger and Sun 1993). The membership function parameters in the ANFIS representing the system behavior are extracted from input data patterns. ANFIS learns features in data patterns and then adjusts the consequent parameters according to a given error criterion. There is a lack of literature in the application of ANFIS in forestry but, successful implementations of ANFIS in geological engineering have been reported recently (for example for strength prediction, Yilmaz and Yuksek 2009). Determination of compressive strength of a rock material is time consuming, expensive and involves destructive tests. If Croat. j. for. eng. 33(2012)2

An Adaptive Network-based Fuzzy Inference System for Rock Share Estimation ... (313–328)

I. Ghajar et al.

Fig. 1 Geological units of study area Slika 1. Geološke jedinice na području istraživanja reliable predictive models could be obtained to correlate unconfined compressive strength to quick, cheap, and non-destructive test results, they would be very valuable for at least the preliminary stage of designing a structure (Yilmaz and Yuksek 2009). In the present paper, the Analytic Hierarchy Process (AHP) method was used to prioritize geological units of the study area with respect to strength factor. AHP originally developed by Saaty (1977; 1980), is one of the most popular Multi-Criteria Decision Making (MCDM) methods, which has been widely used in many fields, including natural resource management. AHP, as a comprehensive framework for modeling the real-world problems has the ability to incorporate both tangible and intangible criteria into the decision making process. Murray and von Gadow (1991); Kangas (1992); Vacik and Lexer (2001) and; Kangas and Kangas (2005) have used AHP in forestry applications. Furthermore, the number of AHP applications in othCroat. j. for. eng. 33(2012)2

er fields, such as geology, is continuously increasing. As a successful application of fuzzy logic and AHP in Engineering geology problems, Liu and Chen (2007) presented a systemic procedure by combining the AHP and the Fuzzy Delphi method (FDM, Kaufmann and Gupta 1988) for assessing the quality of slope rock mass, and classifying the stability of rocks using Linear Discriminant Analysis (LDA) model. This paper proposes an approach that applies the ANFIS technique along with AHP to develop a novel knowledge-based model for rock share estimation. The model is developed based on two input variables of terrain slope and geological units. Furthermore, the impacts of different membership functions on the static parameters of rock share estimation, which are used in the ANFIS model, were investigated. As a case study, the best resulted model is implemented in the 7938 (ha) of Educational and Research Forest of Tarbiat Modares University (TMU) in the northern Iran,

315

I. Ghajar et al.

An Adaptive Network-based Fuzzy Inference System for Rock Share Estimation ... (313–328)

Table 1 Description of geological units of study area Tablica 1. Opis geoloških jedinica Era (groups)

Period

Abr.

Geološke ere

Geološki periodi

Oznaka

Cretaceous Mesozoic

Kreda

Mezozoik Triassic Trijas Paleozoic

Permian

Paleozoik

Perm

Cenozoic

Quaternary

Cenozoik

Kvartar

Description – Opis

K11

Orbitolina limestone and calcareous shale – Vapnenci s orbitolinama i vapneni škriljevci

K21

Conglomerate in lower part, limestone, marly limestone and sandy limestone – Konglomerati u donjem dijelu, vapnenac, laporoviti vapnenac i pjeskoviti vapnenac

K2ml

Limestone, marl, limy marl and silty marl – Vapnenac, lapor, vapnenasti lapor i pjeskoviti lapor

sh,1 3

Shale, sandstone and limestone – Škriljevac, pješčenjak i vapnenac

R2dl

Thick bedded to massive dolomitic limestone, dolomite and limestone – Debeli do masivni dolomitni vapnenac, dolomiti i vapnenci

Cherty limestone, calcareous and sandy shale – Čertni vapnenac, vapneni i pjeskoviti škriljevci

Gray, thick bedded to massive limestone and dolomite (Fuzulinid limestone) – Sivi, debeli do masivni vapneci i dolomiti (fuzulinidni vapnenac)

Qal

Recent alluvium in river beds – Rani aluvij u koritima rijeka

where a 24 (km) forest road has already been constructed and additional forest road alignment should be planned and constructed in the future periods.

2. Materials and Methods – Materijal i metode 2.1 Study area and data collection – Područje istraživanja i prikupljanje podataka The research was carried out in districts 2, 3, 4, and 5 of Educational and Research Forest of TMU situated between longitudes 51°40’37’’E–51°51’36’’E and latitudes 36°29’08’’N–36°34’33’’N. The topographic elevation is about 2 to 2 206 m above mean sea level. The slope classes and their area percentage ranged from 0 to 10% (6.3%); 10–25% (21.6%); 25–45% (31.6); 45–70% (22.3%), and more than 70% (18.2%). In order to make adaptable data that could be applied in the study area, similar geological units to the project area and their constructed road network were investigated to find where the sample data should be collected. Fig. 1 shows the map of geological units in the study area. The cut slope of the constructed road in each geological unit was divided into 10 (m) non-overlapping intervals (Fig. 2). In each sample the type of geological unit and slope degree of terrain were recorded as inputs and the amount of rock share was estimated by an expert in the form of linguistic values as the observed output. A total of 130 samples, including all combination of

316

Fig. 2 Input and output data collection Slika 2. Ulazni i izlazni podaci input variables, were recorded. In the current study, we determined the share of rock for cut-slope areas in three groups of geological units; Mesozoic sediment formations, Paleozoic geological units and Quaternary formation of alluvial deposit (Table 1). To quantify the geological information, Analytical Hierarchy Process (AHP) was used to prioritize the geological units regarding strength against the earth embankment. Since the allowable terrain slope for road construction is up to 70%, to collect a complete set of data, additional samples were collected from higher terCroat. j. for. eng. 33(2012)2

An Adaptive Network-based Fuzzy Inference System for Rock Share Estimation ... (313–328)

I. Ghajar et al.

Table 2 Linguistic values of input and output variables Tablica 2. Vrijednosti ulaznih i izlaznih podataka Variables – Varijable

Rating values – Ocjene vrijednosti

Terrain Slope, % – Nagib terena, % Inputs Ulazni podaci

Output Izlazni podaci

Flat

Gentle

Moderate

Steep

Very Steep

Ravnica

Blago

Umjereno

Strmo

Vrlo strmo

0–10

10–25

25–45

45–70

70<

Strength of geological units (normalized AHP priorities)

Very Low

Low

Medium

High

Very High

Izrazito nisko

Nisko

Srednje

Visoko

Izrazito visoko

Ocjena geološkoh jedinica (normalizirani AHP)

0.25

0.5

0.75

Share of Rock – Udio stijena

rain slopes so that the minimum and maximum slope of the entire area were included in the samples. The observed rock share, as the output of the model, was recorded as three linguistic values »soft«, »medium« and »hard« according to the practical expert-based method of rock share estimation. The linguistic values of input and output variables specified by the fuzzy sets as well as their ranges are shown in Table 2.

2.2 Data division – Razdioba podataka The purpose of the training process in ANNs and fuzzy systems is to interpolate (generalize) the data used for calibration in high dimensional space. Having a large number of model parameters (connection weights), ANNs and neuro-fuzzy systems can overfit the training data when the data are little or noisy. In other words, if the number of degrees of freedom of the model is large compared with the number of data points used for calibration, the model might no longer fit the general trend, as desired, but might learn the idiosyncrasies of the particular data points used for calibration leading to »memorization«, rather than »gen eralization« (Shahin et al. 2008). To prevent overfitting and evaluate the generalization power of the model, a separate validation and test sets, respectively, are needed. In this research, in order to develop the ANFIS, we used a modified data division method, i.e. cross-validation (Stone 1974), in which the data are divided to three sets: Þ Training set used to adjust the connection weights, membership functions and model parameters, Þ Validation set that checks the performance of the model through the training process and stops the training to avoid overfitting, Þ Testing set used to evaluate the trained ANFIS performance and generalization power. Croat. j. for. eng. 33(2012)2

Soft – Mekano

Medium – Srednje

Hard – Tvrdo

≤30%

50%

70%≤

About 20% of the data were used for testing and the remaining data were divided to training – 80% and validation – 20%.

2.3 Analytic Hierarchy Process – Analitički hijerarhijski proces (AHP) 2.3.1 Fundamentals – Osnove i temelji AHP is a mathematical method for analyzing complex decisions with multiple attributes (Saaty 1977; Saaty 1980). AHP can consider the objective information, expert knowledge, and subjective preference at the same time. It aggregates separate performance indicators into an integrated performance indicator (Bouma et al. 2000). In addition, both qualitative and quantitative criteria can be included in the judgments and comparisons of alternatives. By decomposing the decision problem into its elements, a hierarchical decision structure is constructed in the AHP that helps decision makers to view the problem. The preferences for the attributes (or alternatives) are compared in a pairwise manner and numerical techniques are used to derive quantitative values from these comparisons (Kurttila et al. 2000). Unlike other related methods that require quantitative values of criteria, which are measured in ratio or interval scales in their analysis; AHP can transfer the qualitatively expressed measures into a ratio scale through the pairwise comparison. The intensity of preference between alternatives can be expressed on a nine-point scale. If two alternatives are of equal importance, a value of 1 is given in the comparison, while a 9 indicates the absolute importance of one criterion over the other (Saaty 1980). Pairwise comparison data can be analyzed using either regression methods or the eigenvalue technique (Ananda and Herath 2007). The Eigenvalue is calculated for every pair of compo-

317

I. Ghajar et al.

An Adaptive Network-based Fuzzy Inference System for Rock Share Estimation ... (313–328)

nents. If there are n components to be computed the matrix I, is defined as follow:

the ranking of geological units was performed in Super Decision software ver.1.6.0. 2.3.3 Normalization technique – Normalizacija

After computing all pairwise comparisons the priority weight vector (w) is computed as the unique solution of equation 1. Iw = lmax w

(1)

Where: lmax is the largest Eigenvalue of matrix I. The priority vector w is often normalized by a = Sni = 1wi. This ensures the uniqueness of w and provides that a becomes unity (Saaty 1980). The Consistency Index (CI) of derived weights is a parameter that measures the consistency of pairwise comparisons and could be calculated by:

Due to different scales of input and output variables, and in order to increase the speed and accuracy of data processing, input and output data were normalized in a boundary of (0.1) before using them in the ANFIS. As a result, the synthesized priorities of geological units in AHP and terrain slope, as the inputs, and also the percent of rock share (estimated by expert knowledge), as the output of the ANFIS, were normalized using the equation 3. As shown in the equation, the normal forms of each input and output were calculated by dividing each value minus the minimum value by the maximum value minus the minimum value so that the largest individual had a priority of 1.0. The normalized values are also used to create the membership function of input and output variables. (3)

(2) As a rule of thumb, a CR value of 10% or less is considered acceptable (Saaty 1977).

2.4 The Adaptive Network-based Fuzzy Inference System – Prilagodljivi mrežno-fazni sustav 2.4.1 Architecture – Arhitektura sustava

Fig. 3 The Structure of AHP for prioritization of geological units with respect to their strength variable Slika 3. Ustrojstvo AHP-a i ocjene geoloških jedinica 2.3.2 Applying AHP to rank the geological units Primjena AHP-a za ocjenjivanje geoloških jedinica AHP approach was adopted to calculate the degree of preference of geological units from the viewpoint of rock strength. The structure of AHP model is shown in Fig. 3. Otherwise AHP was applied to quantify the preference of each of the eight units with respect to the strength against the earth embankment. The less strength against the earthwork, the more value of computed preference. The weighting process to synthesize

318

ANFIS proposed by Jang (1993) is one of the most applied fuzzy inference systems especially in modeling of the real-world physical objects. ANFIS is a combination of Fuzzy Logic (FL) and Artificial Neural Network (ANN), which applies the learning process developed in ANN approaches to a fuzzy inference system (FIS). The selection of the FIS is the main concern in designing the ANFIS. Several FIS were developed in the literature, each of which was based on the type of fuzzy reasoning and the employed fuzzy ifthen rules (e.g. Mamdani and Assilian 1974; Tsukamoto 1979; Takagi and Sugeno 1983). In the current study, ANFIS uses a Sugeno inference system (Sugeno and Kang 1988). The consequent part of the linear equations and the parameters can be estimated by a simple least squares method (Farokhnia et al. 2010). ANFIS structured by a five-layered network and a hybrid algorithm is used to tune this system based on the structure of input and output data. A mathematicalschematic representation of Takagi-Sugeno’s type of ANFIS with two inputs x and y, one output z, two MFs for each input and two rules is shown in Fig. 4. It is a simple example of a fuzzy inference system, known as the first-order Sugeno FIS. The fuzzy ruleCroat. j. for. eng. 33(2012)2

An Adaptive Network-based Fuzzy Inference System for Rock Share Estimation ... (313–328)

I. Ghajar et al.

Fig. 4 Schematic-Mathematical representation of Takagi-Sugeno’s type of ANFIS for two rules Slika 4. Shematsko-matematički prikaz modela Takagi-Sugeno ANFIS s dva pravila

base of this example including two if-then rules can be presented as follows:

Where: x or y

are the input variables to node i,

Ai, Bi-2 are the linguistic label (such as low or high) associated with this node, characterized by the appropriate MFs in this node. (4) Where: A1, A2 and B1, B2

are the membership functions of inputs X and Y, respectively,

p1, q1, r1 and p2, q2, r2 are the parameters of output function. The modeling process of ANFIS is described layer by layer: Input variables are uncertain in the first hidden layer. Every node i in this layer is an adaptive node with a node function: (5)

Croat. j. for. eng. 33(2012)2

To investigate the impact of different MFs on the result of the ANFIS model in this research, the Gaussian, generalized bell-shaped, trapezoidal-shaped, and triangular-shaped functions were applied. Fig. 5 illustrates an overview of development of the RSE-ANFIS models with different MFs based on input variables applying first-order Sugeno reasoning method. There are a total of 25 rules for each model. The first layer is one of the two adaptive layers of this ANFIS architecture, because three modifiable parameters {ai, bi, ci} are related to MFs present in this layer. These parameters are so-called premise (antecedent) parameters used to calculate the fuzzy output of each node function. The shape of MFs varies with any change of the above mentioned parameters at various stages of training, thus exhibiting various forms of membership functions on linguistic label Ai,

319

I. Ghajar et al.

An Adaptive Network-based Fuzzy Inference System for Rock Share Estimation ... (313–328)

Fig. 5 Architecture of adopted approach for modeling of rock share using ANFIS Slika 5. Sustav preuzetoga modela ANFIS za procjenu udjela stijena u tlu

with a maximum equal to 1 and minimum equal to 0 (Jang 1993), and consequently, results in more fuzzy values of input variables for each type of MF. The antecedent parts of rules are computed in the second hidden layer using T-norm operators. This layer consists of the nodes labeled ∏, which multiplies the incoming signals and sends the product out. For instance: (6) The output (wi) represents the firing strength of a rule. The third hidden layer is used for normalization of the rules’ firing strength (FS). Every node in this layer labeled as N calculates the ratio of ith rules’ FS to the sum of all rules’ FS. (7) The fourth hidden layer determines the consequent part of the rules. Node i computes the contribution of ith rule toward the model output using the following function:

320

(8) The fourth layer is the second adaptive layer of ANFIS architecture. Like with the first layer, there are three modifiable parameters (e.g. pi, qi, ri), the so-called consequent parameters (Jang 1992). Finally the single node in fifth layer labeled with ∑, computes the overall output as the sum of all incoming signals. The corresponding function can be as follows: (9) 2.4.2 Learning algorithm – Algoritam učenja Since there are two adaptive layers in the ANFIS, the task of learning algorithm for this architecture is to tune all the modifiable parameters, namely {ai, bi, ci} (premise parameters) and {pi, qi, ri} (consequent parameters), to make the ANFIS output match the training data (Polat et al. 2008). Similar to conventional statistical models, the model parameters are adjusted in the model calibration phase (training) using a hybrid learning algorithm, so as to minimize the error between model outputs and the corresponding measured values for a particular data set, e.g. the training Croat. j. for. eng. 33(2012)2

An Adaptive Network-based Fuzzy Inference System for Rock Share Estimation ... (313–328)

I. Ghajar et al.

Fig. 6 ANFIS learning process by Hybrid algorithm Slika 6. Postupak učenja modela ANFIS pomoću hibridnoga algoritma set. The learning process of ANFIS by this Hybrid algorithm is shown in Fig. 6. ANFIS implements a hybrid algorithm for the learning process. This algorithm combines the gradient descent method used to learn and modify the premise parameters, and least squares method (LSM) which determines the consequent part. The reason for this combination is that when the premise parameters are not fixed, the search space becomes larger and the convergence of the training becomes slower. Thus, the training process that results in learning has two steps in each iteration. Step 1 In the first epoch, the input patterns are propagated and the optimal consequent parameters are identified by the LSM, while the premise parameters are assumed to be fixed for the current cycle through the training set. Step 2 In the second epoch, the patterns are propagated again; the error signals propagate backward to modify or update the premise parameters, by gradient descent (A back propagation gradient descent method). 2.4.3 Evaluation of model performance – Ocjena modela The ANFIS models were developed in MATLAB ver.7.6 environment. In order to evaluate the predictability, performance and validity of models as well as consistency of results, three well-known statistical criteria, including coefficient of determination (R2), Root Mean Square Error (RMSE), and Mean Absolute Error (MAE), were used (equations 10, 11 and 12). (10)

(11)

Croat. j. for. eng. 33(2012)2

(12) Where: n denotes the number of data, Xio and Xip are the observed and predicted output of pattern number I, respectively. The optimal performance of the model will tend to the R2 of 1 and 0 for other criteria. The purpose of the model validation phase is to ensure that the model has the ability to generalize within the limits set by the training data in a robust fashion, rather than simply having memorized the input-output relationships that are contained in the training data (Shahin et al. 2008). To achieve this, the performance of trained ANFIS is tested on an independent test data set, which has not been used as a part of the model building process. If such evaluation is adequate, the model is considered to be able to generalize and is deemed to be robust. 2.4.4 Model implementation – Primjena modela One of the goals of this research was to create a reliable framework for forest managers or stakeholders to estimate the volume of rock in the ground before road construction begins. The assignment of a given area element (i.e. pixel) to any rock share classes was encountered with the problems due to the variation of physiographic and geological properties within the area and matching these properties with rock share of the underground layers, which would affect the cost of earth embankments. After developing an appropriate ANFIS model, it was implemented to estimate the rock ratio of the ground in the study area. Geographic information system (ArcGIS 9.3), as the best suited tool for handling the spatial data, was used to extract the input data (e.g. slope and geological information) of the ANFIS model from each pixel of the study area rasterized slope and geology maps. By entering these data as the input data to the optimum developed ANFIS, the output value of each record was calculated. The obtained values were

321

I. Ghajar et al.

An Adaptive Network-based Fuzzy Inference System for Rock Share Estimation ... (313–328)

transferred to the attribute table of original map to produce the final rock share map, which illustrates the spatial variability of rock ratio in the study area. The resulted map was classified to three classes based on the primary expert-based classes, which were considered at the beginning of the research.

3. Results – Rezultati In the present research, the AHP method was applied to obtain the priority of existing geological units with respect to the least strength against the earthwork in forest road construction. There were eight Table 3 Final priorities of geological units with respect to the least strength against the earthwork Tablica 3. Ocjene geoloških jedinica s obzirom na količinu zemljanih radova (iskop) Geological units

AHP priorities

Geološke jedinice

ocjena AHP-a

Ocjena za iskapanje

Qal

0.315

0.234

K2ml

0.168

R3sh,1

0.106

K21

0.071

0.047

K11

0.032

R2dl

0.023

Total Ranking for excavation

Strength Težina Low Nisko

High Visko

CI=0.0266 < 0.1 (OK)

types of geological units in the study area that had to be prioritized. The final results synthesized from the AHP model are presented in Table 3. As shown in this table, the structure Qal took the highest priority and it is the least strength unit and R2dl is the most resistant structure to the earthwork from the viewpoint of geology experts. The judgments and comparisons between alternatives in this AHP model were based on the question as to which geological unit showed less strength against the earthwork. The results of normalization of input and output field data, calculated by equation (3), are shown in Table 4. The prepared normalized values were, then, transferred into the ANFIS models. Table 5 shows a part of normalized data that were fed to the model as the training, validation, and testing data. All models were developed using the first-order Sugeno FIS and Hybrid optimization method. The adopted strategy for obtaining the model with the best performance was the incorporation of four different membership functions in designing the ANFIS models. The best result obtained from this strategy was related to triangular-shaped MF (Table 6). The main preference criteria of the models were the coefficient of determination (R2) and Root Mean Square Error (RMSE) of validation data set. R2 and RMSE for triangular-shaped MF were 94.89% and 9.23%, respectively, obtained after training epoch 12. The step size adaptation for the parameters of optimum ANFIS is shown in Fig. 7. The process of each epoch presented in this figure is illustrated in Fig. 6. Fig. 7 shows that the training process stopped at epoch 12 and the model could not improve its own performance after this stage. These results showed that the ANFIS model developed using the triangular MFs had the highest power of generaliza-

Table 4 Maximum and minimum values of input and output variables before and after normalization Tablica 4. Najveće i najmanje vrijednosti varijabli prije i poslije normalizacije Inputs – Ulazni podaci Slope, % – Nagib, %

322

Output – Izlazni podaci

Geology formation priorities

Rock share

Ocjena geoloških tvorbi

Udio stijena

Before normalization

Min.

0.023

Low – Nizak (≤ 30%)

Prije normalizacije

Max.

140

0.315

High – Visok (≥ 70%)

After normalization

Min.

Poslije normalizacije

Max.

Croat. j. for. eng. 33(2012)2

An Adaptive Network-based Fuzzy Inference System for Rock Share Estimation ... (313–328)

I. Ghajar et al.

Table 5 A part of data used as training, validation and testing data Tablica 5. Dio podataka korištenih za vježbu i ispitivanje valjanosti modela Geology

Slope

Observed rock share

Geologija

Nagib

Procijenjeni udio stijena

0.753

0.185

0.753

0.269

0.5

0.030

0.704

0.481

0.164

0.667

0.284

0.444

0.753

0.444

0.753

0.222

0.5

tion and performance in rock share estimation. Consequently, this model was selected as the optimum ANFIS. The results of incorporation of Gaussian MF were very close to triangular ones. The obtained R2 and RMSE of Gaussian MF were 93.74% and 10.03%, respectively. These results of this study also showed a robust perfor-

Fig. 7 Step size diagram of training process of optimum ANFIS (stopped at epoch 12) Slika 7. Razvijanje modela ANFIS »korak po korak« (zaustavljeno u 12. stavki)

Table 6 Result of application of different MFs in rock share ANFIS model Tablica 6. Primjena različitih funkcija članstva (MFs) u modelu ANFIS Type of FIS – Vrsta modela sustava neizrazitoga zaključivanja

Adaptive network based Fuzzy Inference System (Takagi-Sugeno) Prilagodljivi mrežno-fazni sustav za modeliranje (Takagi-Sugeno)

Learning algorithm – Algoritam učenja Type of MF – Vrsta funkcije članstva

Hybrid – Hibridni sustavi G.Bell

Gaussian

Trapezoidal

Triangular

Gauss

Trapezoidna

Trokutasta

R2 (training data) % – Koeficijent determinacije (podaci za vježbu modela), %

96.87

96.40

93.31

96.90

R2 (validation data) % Koeficijent determinacije (podaci za ispitivanje valjanosti modela), %

81.50

64.00

88.80

85.60

R2 (test data) % – Koeficijent determinacije (podaci za ispitivanje modela), %

82.50

93.74

78.05

94.89

RMSE (training data) % Pogreška korijena usrednjenih kvadrata (podaci za vježbu modela), %

6.77

7.18

7.28

6.37

RMSE (validation data) % Pogreška korijena usrednjenih kvadrata (podaci za ispitivanje modela), %

18.25

30.58

14.39

16.33

RMSE (test data) % Pogreška korijena usrednjenih kvadrata (podaci za ispitivanje modela), %

21.25

10.03

19.28

9.23

MAE (training data) % Srednja apsolutna pogreška (podaci za vježbu modela), %

2.09

2.31

2.14

1.94

MAE (validation data) % Srednja apsolutna pogreška (podaci za ispitivanje valjanosti modela), %

8.37

8.24

3.63

4.61

MAE(test data) % Srednja apsolutna pogreška (podaci za ispitivanje modela), %

9.89

4.23

6.63

4.06

Epochs – Stavke Croat. j. for. eng. 33(2012)2

323

I. Ghajar et al.

An Adaptive Network-based Fuzzy Inference System for Rock Share Estimation ... (313–328)

Table 7 Estimated areas for rock share classes Tablica 7. Procjena površine ovisno o udjelu stijena u tlu Rock share classes

Area, ha

Relative area, %

Udio i vrsta tla

Površina, ha

Udio površine, %

Soft – Mekano

2137.44

26.92

Medium – Srednje

3240.11

40.83

Hard – Tvrdo

2560.26

32.25

assigned to each class of rock share are presented in Table 7. The best coefficient of determination, related to the ANFIS model with triangular MF, is shown in Fig. 8. The predicted results driven from this optimum ANFIS have been plotted with the results of field data (i.e. real data). The total set of field data were used to calculate the result of the optimum ANFIS.

4. Discussion – Rasprava mance in modeling the ratio of rocks in the ground. Application of other types of MFs resulted in either less R2 or more RMSE than triangular and Gaussian ones. The result of application of the developed ANFIS in predicting the rock share in the study area is shown in Fig. 8. The results of final estimated areas that were

The ultimate purpose of this modeling was to provide a practical approach for estimating the proportion of rock for the purpose of estimating the cost of forest roads. In other words, by estimating the rock proportion in various conditions of a mountainous forest before planning the forest road network, a planner

Fig. 8 Spatial variability of rock share based on ANFIS model with triangular-shaped MFs Slika 8. Prostorna promjenjivost udjela stijena u tlu prema modelu ANFIS s trokutastom funkcijom članstva

324

Croat. j. for. eng. 33(2012)2

An Adaptive Network-based Fuzzy Inference System for Rock Share Estimation ... (313–328)

Fig. 9 The optimum ANFIS predicted values of rock share estimation versus field data values Slika 9. Najtočnije predviđeni podaci modelom ANFIS can effectively decide where to place the road to decrease the time and cost of earthwork. In this research, AHP and ANFIS methods were applied to model the share of rock in different geological and physiographic conditions. In addition, four different types of membership function were adopted for the analysis in ANFIS training to compare their differences regarding statistical parameters. Although the presented ANFIS approach is an experimental method in which just two main input variables (e.g. geological structures and terrain slope) were considered, the acceptable ranges of statistical parameters of R2 and RMSE were obtained from all four developed models. The result of this study showed that an ANFIS can obtain a higher level of accuracy and generalization power for rock share estimation when triangular membership function is used to conduct system training. The coefficient of determination for ANFIS with triangular MF was 0.94. This result means that 94.89% of changes of rock proportion are related to the changes of the two considered variables, i.e. geological information and terrain slope. The RMSE is the most popular measure of error and has the advantage that large errors receive much greater attention than small errors (Hecht-Nielsen 1990). The result of RMSE for applied MFs indicated that the use of triangular MF produced the least RMSE compared to other three MFs. In contrast with RMSE, MAE eliminates the emphasis given to large errors. Both RMSE and MAE are desirable when the evaluated output data are smooth or continuous (Twomey and Smith 1997). Since Croat. j. for. eng. 33(2012)2

I. Ghajar et al.

the defined output linguistic variables, i.e. Low, Medium, and High, as well their corresponding normalized values, i.e. 0, 0.5, and 1 created a relatively discrete space (Fig. 9), MAE was not a determinant criteria for preference of a model to another. Nevertheless, a conclusion can still be drawn from the MAE results. According to the results of MAE shown in Table 6, it can be seen that the ANFIS developed by triangular MF still has the least error compared to other models. Thus it could be concluded that the ANFIS developed by triangular MF can be selected as the optimum model for rock share estimation. After triangular MF-based ANFIS, which produced the best results, the adaptation of Gaussian MF was determined as the second robust model from the viewpoint of statistical criteria. The application of Gaussian MF resulted in a R2 of 93.74%, an RMSE of 10.03%, and a MAE of 4.23%. Hence, it could be expected that the use of Gaussian MF in RSE-ANFIS model will generate robust performance and high generalization power. The g.bell and trapezoidal MFs results were less favorable than those of triangular and Gaussian MFs and therefore they were not proposed as acceptable models in the present research. In order to produce a zoning map of rock proportion, the ANFIS value calculated for each pixel was transferred to the corresponding point on the map of the study area. The current map included a continuous set of data that needed to be changed into two or more categories. A number of classifiers are available and namely natural breaks, quantile, equal intervals, Kmeans, etc; each of them may lead to different clustering results because of their different statement about the method of dividing. The final classification of rock share map in this study was based on the practical procedure used by the experts to determine the price of embankments for forest road projects. Thus, the normalized scores on the map were grouped into three categories of rock ratio: »Soft« (0–0.3), »Medium« (0.3–0.7), and »Hard« (0.7–1). This classification may differ from a country to another but it is a common principle in forest management that the least rock ratio areas should be traversed in forest road construction to minimize the total time and cost of construction. The weathering condition of geological structure is one of the factors in determining the cost of excavating the rocks; however this factor was neglected since the investigation of the nature of rocks was not the purpose of this study and the focus was on the proportion of rocks in the underground layers. Although it is more expensive to use geotechnical rock testing facilities than experts’ opinions, the investigation of mechanical and physical properties of

325

I. Ghajar et al.

An Adaptive Network-based Fuzzy Inference System for Rock Share Estimation ... (313–328)

near–surface rocks, and performing the strength and deformation tests for the evaluation of the possibility of excavation, could give more accurate data for such models. In the current study, the AHP technique was used to quantify the geology formations with respect to their strength against earthwork to persuade forest managers to apply practical fuzzy models for various construction purposes in forest management. The advantages of the presented procedure are as follows: Easy sampling and applied methodology that allow other researchers to repeat the study with lower costs in other regions with specific local conditions, Ability of handling several types of data (numerical, ordinal, or nominal) in fuzzy inference systems, which makes it flexible for the use in modeling natural resources, Acceptable statistics with the emphasis on effective input factors and reliability of the results in the present study. A comparison between the applied soft computing method and other traditional statistical methods such as multiple regressions could indicate the degree of robustness or fault tolerance of presented models. It appears that there is a possibility of estimating rock share of subsoil by using the proposed soft computing models. The number of the analyzed data is relatively limited in this study. Therefore, the practical outcome of the proposed model could be used, with acceptable accuracy, for the estimation of earthwork cost at the preliminary stage of planning the forest roads in the study area.

5. References – Literatura Ananda, J., Herath, G., 2008: Multi-attribute preference modeling and regional land-use planning. Ecological Economics 65(2): 325–335. Andrews, R., Diederich J., Tickle, A., 1995: A survey and critique of techniques for extracting rules from trained artificial neural networks. Knowledge-Base Systems 8(6): 373–389. ASCE Task Committee on Applications of Artificial Neural Networks in Hydrology 2000: Artificial neural networks in hydrology: I: preliminary concepts; II: hydrologic applications. Journal of Hydrological Engineering 5(2): 115–137. Barton, N., Lien, R., Lunde, J., 1974: Engineering classification of rock masses for the design of tunnel support. Rock Mechanics 6(4): 189–236. Bieniawski, Z. T., 1975: Case studies: prediction of rock mass behavior by the geomechanics classification. Proc. 2nd Australia–New Zealand Conference Geomechanics, Brisbane, 36–41 pp. Bouma, J., Brouwer, R., Van Ek, R., 2000: The use of integrated assessment methods in Dutch water management: a com-

326

parison of cost-benefit and multi-criteria analysis. Third International Conference of the European Society for Ecological Economics, Vienna, 3–6 pp. Farokhnia, A., Morid, S., Byun, H. R., 2010: Application of global SST and SLP data for drought forecasting on Tehran plain using data mining and ANFIS techniques. Theoretical and Applied Climatology 10(1–2): 71–81. Gail, M., Brion, T. R., Neelakantan, S. L., 2002: A neuralnetwork-based classification scheme for sorting sources and ages of fecal contamination in water. Water Research 36(15): 3765–3774. Ghajar, I., Najafi, A., Ezzati, S., 2010: Skidding Machines Allocation (SMA) using fuzzy set theory. Croatian Journal of Forest Engineering 31(2): 99–110. Goodman, R. E., 1989: Introduction to Rock Mechanics. (2nd ed.). Wiley, 562 pp, New York. Hecht-Nielsen, R., 1990: Neurocomputing, Addison-Wesely Publishing Company, Reading, MA. Hoek, E., Bray, J. W., 1981: Rock Slope Engineering. (3rd ed.). Institute of Mining and Metallurgy, 358 pp, London 1981. Inaba, S., Heinimann, H. R., Shiba, M., 2001: A Model to estimate rock excavation volume of forest roads in steep terrain conditions. In: Anonymous (Ed.): Proceedings of the 112th Meeting of the Japanese Forestry Society. April 2–4, Japan. Jang, J.S.R., 1992: Self-learning fuzzy controllers based on temporal backpropagation. IEEE Transactions on Neural Networks 3(5): 714–723. Jang, J.S.R., 1993: ANFIS: Adaptive-Network-based Fuzzy Inference Systems. IEEE Transactions on Systems, Man, and Cybernetics 23(3): 665–685. Kangas, J., 1992: Multiple-use planning of forest resources by using the analytic hierarchy process. Scandinavian Journal of Forest Research 7(1–4): 259–268. Kangas, J., Kangas A., 2005: Multiple criteria decision support in forest management – the approach, methods applied, and experiences gained. Forest Ecology and Management 207(1– 2): 133-143. Kangas, A., Kangas, J., Laukkanen, S., 2006: Fuzzy multicriteria approval method and its application to two forest planning problems. Forest Science 52(3): 232–242. Kung, G. T., Hsiao, E. C., Schuster, M., Juang, C. H., 2007: A neural network approach to estimating deflection of diaphram walls caused by excavation in clays. Computers and Geotechnics 34(5): 385–396. Kurttila, M., Pesonen, M., Kangas, J., Kajanus, M., 2000: Utilizing the analytic hierarchy process (AHP) in SWOT analysis – a hybrid method and its applications to a forest certification case. Forest Policy and Economics 1(1): 41–52. Liu, Y. C., Chen, C. S., 2007: A new approach for application of rock mass classification on rock slope stability assessment. Engineering Geology 89(1–2): 129–143 Mamdani, E. H., Assilian, S., 1974: An experiment in linguistic synthesis with a fuzzy logic controller. International Journal of Man-Machine Studies 7(1): 1–13. Croat. j. for. eng. 33(2012)2

An Adaptive Network-based Fuzzy Inference System for Rock Share Estimation ... (313–328)

I. Ghajar et al.

Mendoza, G. A., Sprouse, W., 1989: Forest planning and decision making under fuzzy environments: an overview and illustration. Forest Science 35(2): 481–502.

Stone, M., 1974: Cross-validatory choice and assessment of statistical predictions. Journal of Royal Statistical Society 36(2): 111-147.

Mendoza, G. A., Bare, B. B., Zhou, Z., 1993: A fuzzy multiple objective linear programming approach to forest planning under uncertainty. Agricultural Systems 41(3): 257–274.

Stuckelberger, J. A, Heinimann, H. R., Burlet, E. C., 2006: Modeling spatial variability in the life-cycle costs of low-volume forest roads. European Journal of Forest Research 125(4): 377–390.

Murray, D. M., Von Gadow, K., 1991: Prioritizing mountain catchment areas. Journal of Environmental Management 32(4): 357–366.

Sugeno, M., Kang, G. T., 1988: Structure identification of fuzzy model. Fuzzy Sets and Systems 28(1): 15–33.

Najjar, Y. M., Basheer, I. A., 1996: Neural network approach for site characterization and uncertainty prediction. ASCE Geological Special Publication 58(1): 134–148.

Takagi, T., Sugeno, M., 1983: Derivation of fuzzy control rules from human operator’s control actions. In: Proc IFAC Symp Fuzzy Inf, 55–60 pp.

Padmini, D., Ilamparuthi K., Sudheer, K. P., 2008: Ultimate bearing capacity prediction of shallow foundations on cohesionless soils using neurofuzzy models. Computers Geotechnics 35(1): 33–46.

Tecle, A., Duckstein, L., Korhonen, P., 1994: Interactive, multiobjective programming for forest resources management. Applied Mathematics and Computations 63(1): 75–93.

Pettifer, G. S., Fookes, P. G., 1994: A revision of the graphical method for assessing the excavatability of rock. Journal of Engineering Geology 27(2): 145–164. Polat, K., Yosunkaya, S., Güneş, S., 2008: Pairwise ANFIS Approach to Determining the Disorder Degree of Obstructive Sleep Apnea Syndrome. Journal of Medical Systems 32(5): 379–387. Rankine, R., Sivakugan, N., 2005: Prediction of paste backfill performance using artificial neural networks. Proceedings of 16th international society for soil mechanics and foundation engineering, Osaka, 1107–1110 pp. Roger, J. S., Sun, C., 1993: Functional equivalence between radial basis function networks and fuzzy inference systems. IEEE Trans Neural Netw 4:156–159. doi:10.1109/72.182710 Saaty, T. L., 1977: A scaling method for priorities in hierarchical structures. Journal of Mathematical Psychology 15(3): 234–281. Saaty, T. L., 1980: The Analytic Hierarchy Process. McGrawHill, New York. Shahin, M. A., Jaksa, M. B., Maier, H. R., 2005: Neural network based stochastic design charts for settlement prediction. Canadian Geotechnical Journal 42(1): 110–120. Shahin, M. A., Jaksa, M. B., Maier, H. R., 2008: State of the art of artificial neural networks in geotechical engineering. Electronic Journal of Geotechnical Engineering 8: 1–26. Sreekanth, P. D., Sreedevi, P. D., Ahmed, S., Geethanjali, N., 2010: Comparison of FFNN and ANFIS models for estimating groundwater level. Environmental Earth Sciences, DOI 10.1007/s12665-010-0617-0.

Tekin, E., Akbas, S. O., 2011: Artificial neural networks approach for estimating the groutability of granular soils with cement-based grouts. Bulletin of Engineering Geology and the Environment 70(1): 153–161. Tsukamoto, Y., 1979: An approach to fuzzy reasoning method. In: Gupta MM, Ragade RK, Yager RR (eds) Advances in fuzzy set theory and applications. Elsevier, Amsterdam, 137–149 pp. Twomey, J. M., Smith, A. E., 1997: »Validation and verification«. Artificial neural networks for civil engineers: Fundamentals and applications, N. Kartam, I. Flood, and J. H. Garrett, eds., ASCE, New York, 44-64. Vacik, H., Lexer, M. J., 2001: Application of a spatial decision support system in managing the protection forests of Vienna for sustained yield of water resources. Forest Ecology and Management 143(1–3): 65–76. Wang, C. H., Liu, B. J., Wu, L. S. H., 2010: The Association Forecasting of 13 Variants Within Seven Asthma Susceptibility Genes on 3 Serum IgE Groups in Taiwanese Population by Integrating of Adaptive Neuro-fuzzy Inference System (ANFIS) and Classification Analysis Methods. Journal of Medical Systems 36(1): 175–185. Wickham, G. E., Tiedemann, H. R., Skinner, E. H., 1972: Support determination based on geologic predictions. Proc. Conf. Rapid Excavation and Tunneling, 43–64 pp. Yilmaz, I., Yuksek, G., 2009: Prediction of the strength and elasticity modulus of gypsum using multiple regression, ANN, and ANFIS models. International Journal of Rock Mechanics & Mining Sciences 46(4): 803–810. Zadeh, L. A., 1965: Fuzzy sets. Inf Control. 8: 338–353.

Sažetak

Prilagodljivi mrežno-fazni sustav za procjenu udjela stijena pri izgradnji šumskih prometnica Ovo je istraživanje novi model procjene udjela stijena u tlu (RSE). Postupak je koristan za izračun troškova zemljanih radova koji se ubrajaju među glavne troškove izgradnje šumskih prometnica. Udio stijena u tlu izravno Croat. j. for. eng. 33(2012)2

327

I. Ghajar et al.

An Adaptive Network-based Fuzzy Inference System for Rock Share Estimation ... (313–328)

utječe na cijenu izgradnje prometnica. Dakle, postoji potreba za pouzdanom procjenom udjela stijena unutar cijeloga područja izgradnje prometnice, osobito u stjenovitim područjima gdje je udio stijena visok. Nije poznato numeričko svojstvo ni pouzdan numerički parametar za mjerenje udjela stijena u tlu, a nestalnost udjela stijena u tlu bio je i dodatan razlog ovoga istraživanja. Cilj je istraživanja bio uvesti stručni sustav za procjenu udjela stijena u tlu u različitim uvjetima pomoću prila godljivoga mrežno-faznoga sustava za modeliranje (ANFIS) i analitičkih hijerarhijskih procesa (AHP). Nagib terena i vrste geoloških tvorbi smatrani su ulaznim varijablama za stvaranje modela ANFIS. Kako bi se smanjili troškovi uzorkovanja udjela stijena u tlu, razvijen je praktičan pristup koji osam postojećih geoloških jedinica obrađuje kao nositelje problema u odlučivanju. AHP-i na temelju znanja stručnjaka korišteni su za rješavanje problema procjene udjela stijena u tlu. Rezultat modela ANFIS jest razradba na tri vrste tla, ovisno o udjelu stijena: meko, srednje i tvrdo. Nakon uzorkovanja, normalizacije podataka te podjele Sugeno sustav neizrazitoga zaključivanja prvoga reda i hibridna metoda optimizacije usvojeni su kako bi se stvorio model ANFIS. Izlazni podaci modela predstavljaju funkciju prvoga reda čiji su parametri prilagođeni svakoj stavki sustava optimizacije. Trokutaste funkcije članstva (MF) dale su najbolje rezultate. Sustav je primijenjen u planinskim šumama u Iranu gdje će u skoroj budućnosti biti izgrađena mreža šumskih prometnica. Predviđene vrijednosti zatim su opisane prostorno odnosno u okruženju GIS-a. Procijenjeni udjeli stijena u tlu po razredima: meko, srednje i tvrdo iznosili su 6,92 %, 40,83 % i 32,25 %. Ovaj je pristup koristan kao prvo za pokazivanje nestalnosti udjela stijena u tlu te drugo kao model za izračun troškova izgradnje šumskih prometnica koji će uz pomoć ostalih matematičkih modela još točnije prikazati izračun troškova šumskih prometnica i potom omogućiti odabir zamjenskih trasa u slučaju pre visokih troškova izgradnje. Ključne riječi: udio stijena u tlu, ANFIS, AHP, troškovi izgradnje šumskih prometnica, funkcije članstva

Authors’ address – Adresa autorâ:

Received (Primljeno): February 08, 2012 Accepted (Prihvaćeno): August 18, 2012

328

Ismael Ghajar, MSc. e-mail: ismael.ghajar@modares.ac.ir Assist. Prof. Akbar Najafi, PhD. e-mail: a.najafi@modares.ac.ir Tarbiat Modares University Faculty of Natural Resources Department of Forestry Noor, PO Box: 64414–356 IRAN Assoc. Prof. Seyed Ali Torabi, PhD. e-mail: satorabi@ut.ac.ir University of Tehran College of Engineering Department of Industrial Engineering Tehran IRAN Assoc. Prof. Mashalah Khamehchiyan, PhD. e-mail: khamechm@modares.ac.ir Tarbiat Modares University Department of Engineering Geology Tehran IRAN Assoc. Prof. Kevin Boston, PhD. e-mail: kevin.boston@oregonstate.edu Oregon State University Department of of Forest Engineering Corvallis, OR USA Croat. j. for. eng. 33(2012)2

Preliminary note – Prethodno priopćenje

Utilizing Airborne Laser Scanning Technology in Predicting Bearing Capacity of Peatland Forest Jori Uusitalo, Tomi Kaakkurivaara, Maarit Haavisto Abstract – Nacrtak Airborne laser scanning (ALS) technology is receiving increasing attention in forestry. So far, ALS is mainly utilized at a rather abstract level to derive better estimations on the composition of forest stands (e.g. volumes of wood assortments per hectare). Recent rather expensive ALS inventory projects have raised interest in whether other purposes can also benefit from ALS data. The paper demonstrates a case study where the ALS data was utilized in predicting the bearing capacity of peatland forest. The study stand was divided into 16 × 16 m2 grids within which the volume of trees in each cell was estimated. In addition, the height of the terrain (elevation) at each point was estimated. The stand was harvested with a harvester and a for warder during the summertime, after which the ruts of the trails were measured. The results indicate that ALS may be a promising technology in making predictions on the bearing capac ity of peatland forests prior to harvest. More research in this subject is however needed to develop the most effective methods. Both the volume of trees and the relative elevation of the surface of the peatland may be regarded as beneficial information in the prediction of bearing capacity. Keywords: tree harvesting, ALS, rut depths

1. Introduction – Uvod Airborne laser scanning (ALS) data has been shown to provide reliable estimates on growing stock. Hence, ALS-based forest inventories are gradually replacing field inventories in many countries. The estimation methods can be divided into two approaches. The aim of the individual tree delineation (ITD) approach is to discern individual trees or groups of trees based on 3D ALS data. The aim of the area-based statistical approach (ABSA) is to estimate mean forest stand variables for a fixed area using statistical correlations between explanatory variables derived from the ALS data and forest stand variables. So far, ALS has mainly been utilized in a rather abstract level to derive better estimations of the composition of forest stands. Research around ALS technology has been very active recently, but these projects have mainly concentrated on making better predictions as to the quality attributes of trees (Maltamo et Croat. j. for. eng. 33(2012)2

al. 2009; Bollandsås et al. 2010; Holopainen et al. 2010), improved predictions of tree species (Korpela et al. 2010) or estimation of saw log recoveries (Peuhkurinen et al. 2008). Recent rather expensive ALS inventory projects have also attracted interest in the benefits that ALS data can bring to other purposes. Peatlands are problematic from a logging operations point of view. The mean tree size and stand density are generally lower than on mineral soils leading to low harvesting removal. The ditch network weakens machine mobility and the average primary transportation distance is generally double that of mineral soils. The low bearing capacity (the strength of soil to prevent vehicles from sinking into it) is, however, the most severe factor affecting timber harvesting. The mobility of forwarders on peatland has been studied extensively. Comprehensive in situ driving tests have been arranged to compare the mobility of vehicles (Sirén et. al. 1987; Ala-Ilomäki et al. 2011), to

329

J. Uusitalo et al. Utilizing Airborne Laser Scanning Technology in Predicting Bearing Capacity of Peatland Forest (329–337)

predict the mobility of vehicles in various types of peatland (Nugent et al. 2003; Zeleke et al. 2007), to compare the effects of wheels and bogie tracks on rut formation (Bygden et al. 2004; Suvinen 2006; AlaIlomäki et al. 2011), or to utilize the data recorded by the harvester to assist in planning the routes of the forwarder (Suvinen and Saarilahti 2006). These studies have resulted in recommendations on suitable vehicle sizes and the equipment needed to carry out logging in unfrozen peatland forests. Peatland usually consists of a top layer with living and slightly decomposed plants, followed by a layer of decomposed peat and, finally, mineral soil. From the bearing capacity point of view, the top layer with considerable tensile strength provided by roots of trees and shrubs is essential, whereas the supporting function of the decomposed peat is of secondary importance. The strength of the top layer is subject to the

variation of density and species of the vegetation, resulting in extreme spatial variation in trafficability (Ala-Ilomäki 2006). Due to the low bearing capacity in peatland forest, special attention should be given to planning logging operations. Main logging trails should be placed on spots with the highest volumes of trees. Logging trail network should be planned in a way that minimizes the number of passes along the main logging trails. Stands where a main logging trail cannot be located on terrain having sufficient bearing capacity should only be harvested during the coldest period of winter. ALS might offer useful information when carrying out analyses on the trafficability of peatland forests prior to harvest. The aim of this study was to test whether it is possible to utilize DEM and ALS-based estimates on volumes of trees calculated using the ABSA method in predicting the bearing capacity of peatland forest.

Fig. 1 Digital elevation model derived from the ALS data together with the logging trails. The darker the color of the point, the higher the elevation Slika 1. Digitalni model terena nastao iz podataka dobivenih laserskim snimanjem, zajedno s izvoznim putovima (područja obojena tamnije više su nadmorske visine)

330

Croat. j. for. eng. 33(2012)2

Utilizing Airborne Laser Scanning Technology in Predicting Bearing Capacity of Peatland Forest (329–337) J. Uusitalo et al.

2. Material and methods – Materijal i metode The study was conducted in a Pine mire in southern Finland, close to the Hyytiälä Forest Station, in Juupajoki. The area of the study site was 4.5 ha, the mean size of trees was 15 cm, basal area 24 m2 and tree density 1500 trees/ha. The study site had been laserscanned for study purposes several times, in 2004, 2006, 2008 and 2010. During those years the groundtruth, field inventory data on exact diameters and heights of trees, had been collected from more than 20 circular plots, each with a 9 meter radius. The laser scanner surveys provided a point cloud, at which the points x, y and z were known. A digital elevation model (DEM) was produced from the last pulse data, whereas the first pulse data was used to produce the canopy height model (CHM). Since the last pulses include both ground and non-ground hits,

the data was first classified with the Multiscale Curvature Classification (MCC) system proposed by Evans and Hudak (2007). Parameters –s 1.0 and –t 0.2 were used in the classification. The data was then interpolated using ArcGis software with the Inverse distance weighted (IDW) tool. The three closest observations were used in the interpolation (Fig. 1). Canopy heights (z values) were calculated as the difference between the orthometric heights of laser hits and the estimated ground elevation values at the corresponding locations. The stand was covered by a 16 × 16 m grid. The volume of trees and the basal area were estimated for each grid cell with the aid of z values, field inventory data and regression models suggested by Holopainen et al. (2008). The stand had been ditched a couple of times, the last ditching operation having been carried out roughly ten years prior to our tests. Hence, some ditches had

Fig. 2 Volume of trees for each grid cell and locations of the study plots. The darker the color of the cell, the higher the volume Slika 2. Obujam stabala u svakom kvadratu rasterske mreže i lokacije ploha na kojima je provedeno istraživanje (područja obojena tamnije imaju veći obujam stabala) Croat. j. for. eng. 33(2012)2

331

J. Uusitalo et al. Utilizing Airborne Laser Scanning Technology in Predicting Bearing Capacity of Peatland Forest (329–337)

dug soil at the edges of the ditch, giving higher elevation hits (altitude above sea level) than the surrounding area. In addition, the stand included two mineral soil spots being at a clearly higher level than surrounding peatland. The lowest points occurred at the bottoms of ditches. The maximum difference in altitude above sea level (ASL) within the stand was 1.5 m. Each 16 × 16 m grid was given an ASL value (the mean of ground elevation values within the grid). The planning of the logging trail network for thinning operations was carried out prior to harvest. Following that, 20 study plots were systematically placed all over the stand along the logging trail network. The exact location of the logging trails and the study plots can be examined in Figs. 1 and 2. The study stand was harvested using a John Deere 1070D harvester in early August 2010. The harvester driver was given a printed map of the pre-planned logging trail network, which he was asked to follow while executing the thinning operation. The logging trail network actually accomplished was tracked by the GPS recorder in the machine (Fig. 1). The final location of the central point of the study plot was recorded using the Trimble ProXH GPS recorder. A wooden pole was hammered into the ground beside the logging trail to ensure that the measurement, after cutting and forwarding operations, was carried out at the same points. A ten-meter-long vector was placed along the centerline of the logging trail, starting from the central point of the study plot. The depth of both ruts caused by the harvester was measured at one-meter intervals along the vector. The basic level of the ground was defined by pushing a 3.5 m long and 3 cm thick metal pole against the ground perpendicular to the centerline of the trail. The mean depth of rut caused by harvester (Rutharv) for the sample plot is the mean of all twenty measurements taken within the plot. Subsequently, ten-meter-long sample plot vectors were drawn using ArcGIS software. The mean value for the ASL along the vector was

calculated by weighting each ASL class value with the length of that class within the vector. The forwarding operation was carried out the week after the cutting, using a John Deere 810E forwarder with band tracks in both the front and rear bogies. Each route from the storage point to the storage point of the forwarder was recorded using John Deere TimberNavi software. The bearing capacity of the soil was very low, so the operation was forced to stop before all the wood was forwarded to the roadside. Rut depths after forwarding (Rutforw) were measured similarly to the measurements carried out after harvesting. Ruts after forwarding were measured from only 16 study plots. Based on the tracking reports of each route of the forwarder, the total over-driven mass for each sample plot was estimated. The total over-driven mass for the sample plot is calculated by summing up the mass of the harvester and the estimate of mass of the forwarder (mass of the machine + mass of load) for each time it passes the sample plot.

3. Results – Rezultati 3.1 Correlations between rut depth of harvester and the parameters estimating bearing capacity – Povezanost dubine kolotraga harvestera i pokazatelja procjene nosivosti podloge The mean rut depth of the study plots varied between 2.0–17.1 cm. No clear spatial correlation was found. Small and high mean depths were found all over the stand. The correlations between rut depths and the most important parameters estimating bearing capacity are presented in Table 1. The correlation between the rut depth of the harvester and volume is weak, and in this case it is also illogical, since the volume and the rut depth of the harvester should have negative correlation (i.e. the

Table 1 Correlation coefficients between volume, basal area, ASL and rut depth after harvesting (Rutharv) Tablica 1. Koeficijenti korelacije između obujma stabala, temeljnice, nadmorske visine i dubine kolotraga harvestera Basal area

Volume

ASL

Rutharv

Temeljnica

Obujam

Nadmorska visina

Dubina kolotraga harvestera

Basal area – Temeljnica

0.926(**)

–0.086

–0.059

Volume – Obujam

–

–0.199

0.125

ASL – Nadmorska visina

–

–0.378 (*)

Rutharv – Dubina kolotraga harvestera

–

Significance levels – Razine značajnosti: (**) = 0.01; (*) = 0.1; N = 20

332

Croat. j. for. eng. 33(2012)2

Utilizing Airborne Laser Scanning Technology in Predicting Bearing Capacity of Peatland Forest (329–337) J. Uusitalo et al.

Fig. 3 Covariation between rut depth of harvester and volume Slika 3. Ovisnost dubine kolotraga harvestera i obujma stabala

Fig. 5 Covariation between rut depth of harvester and ASL Slika 5. Ovisnost dubine kolotraga harvestera i nadmorske visine correlation between the volume and the rut depth of the harvester would be roughly –0.2. The correlation between the rut depth of harvester and the basal area is close to zero but it is neither negative nor illogical. A moderate and statistically significant negative correlation was found between the rut depth of the harvester and ASL. The study plot located at the edge of the mineral soil spot clearly differs from the other data. The rut depth of the harvester of this plot is only 2 cm (Fig. 5).

3.2 Correlations between the rut depth of the forwarder and the parameters estimating bearing capacity – Povezanost dubine kolotraga forvardera i pokazatelja procjene nosivosti podloge

Fig. 4 Covariation between rut depth of harvester and basal area Slika 4. Ovisnost dubine kolotraga harvestera i temeljnice smaller volume, the deeper the ruts). Close examination of the results (Figs. 3 and 4) revealed that the most illogical observations, study plots 10, 15 and 17) are located in the north-east corner of the stand. In this area there are relatively high volumes and deep ruts. If these observations were removed from the data, the Croat. j. for. eng. 33(2012)2

Forwarding caused very deep ruts for all trails and was highest for those main (arterial) trails that lead to the storages. Rut depths caused by the forwarder were measured in 16 plots. The forwarder passed through study plot 5 unloaded only once. Correspondingly, in study plot 6, forwarder passed through six times; three times unloaded, once half-loaded and twice fully loaded. The total over-driven mass for the sample plot varied between 17,200–123,100 kg. The mean rut depth after forwarding varied 2.0–27.5 cm. The correlations between rut depths after forwarding and the most important parameters estimating bearing capacity are presented in Table 2. Vol-

333

J. Uusitalo et al. Utilizing Airborne Laser Scanning Technology in Predicting Bearing Capacity of Peatland Forest (329–337)

Table 2 Correlation coefficients between volume, basal area, ASL, over-driven mass and rut depth after forwarding Tablica 2. Koeficijenti korelacije između obujma stabala, temeljnice, nadmorske visine, privučenoga drva i dubine kolotraga forvardera Rutforw Dubina kolotraga forvardera Basal area – Temeljnica

–0.681(**)

Volume – Obujam

–0.576(*)

ASL – Nadmorska visina

–0.205

Over-driven mass – Privučeno drvo

0.386

Significance levels – Razine značajnosti: (**) = 0.01; (*) = 0.1; N = 16

Table 3 Partial correlation coefficients between basal area, volume, ASL and rut depth of forwarder while controlling the over-driven mass Tablica 3. Parcijalni koeficijenti korelacije između obujma stabala, temeljnice, nadmorske visine i dubine kolotraga uz nadzor količine privučenoga drva

Fig. 7 Covariation between rut depth of forwarder and basal area Slika 7. Ovisnost dubine kolotraga forvardera i temeljnice

Rutforw Dubina kolotraga forvardera Basal area – Temeljnica

–0.622

Volume – Obujam

–0.470

ASL – Nadmorska visina

–0.234

N = 16

Fig. 8 Covariation between rut depth of forwarder and ASL Slika 8. Ovisnost dubine kolotraga forvardera i nadmorske visine

Fig. 6 Covariation between rut depth of forwarder and volume Slika 6. Ovisnost dubine kolotraga forvardera i obujma stabala

334

ume, basal area and ASL have a distinct, negative correlation with the rut depth of the forwarder. This correlation is illustrated in Figs. 6, 7 and 8 to get a better insight into the covariation of these parameters. It is quite obvious that over-driven mass has a positive correlation with the rut depth of the forwarder and this factor has a distinct effect on the correlations givCroat. j. for. eng. 33(2012)2

Utilizing Airborne Laser Scanning Technology in Predicting Bearing Capacity of Peatland Forest (329–337) J. Uusitalo et al.

en. Therefore, the relationship between the rut depth of the forwarder and other parameters should be examined by excluding the effect caused by the overdriven mass. The partial correlations between rut depth, basal area, volume and ASL while controlling the over-driven mass are given in Table 3.

4. Discussion – Rasprava This field study was carried out in only one stand, so one cannot make strong judgement on the results of the study. No negative correlation was found between tree volume and rut depth caused by the harvester, which was rather surprising. One corner of the study stand gave the most illogical observations. The removal of these observations would have changed the correlation coefficient to negative (i.e. the less volume, the deeper the ruts). It might be that the structure of the peat layers within this area were different compared to the rest of the stand. There are also many other factors affecting the rut depth caused by the harvester that are irrelevant in terms of the relationships investigated here. These factors include the effect of the driver, the curve of the trail, individual tree stumps, logging residual, and so on. A logical, negative correlation was found between tree volumes and rut depths measured after forwarding. Basal area gave a slightly higher correlation with rut depths after forwarding than volume. This is in accordance with earlier findings which state that in peatlands, a greater relative proportion of biomass is allocated into the root system compared to mineral soil (Laiho and Finer 1996; Finer and Laine 1998). This means that from a bearing capacity point of view, it is better to have two small trees than one big tree. It is not clear why the correlation after tree volume and rut depths were illogical after harvesting but logical after forwarding. Other factors such as moisture content of peat, peat type, shear strength of the roots or the amount of cutting depris that were not measured in this study may have an effect on these results. It might also be so that the differences in bearing capacity appear distinctly only after forwarding because in many plots harvester alone do not break the root mat that have most important supportive function. The relative elevation of the surface of the peatland seems to have an influence both on the rut depths measured after harvesting and forwarding. Peatland might include spots or areas where mineral soil is at a higher level than the surrounding area. There might even be areas where no peat layer exists. In areas where ditching operations have been carried out, the edges of the Croat. j. for. eng. 33(2012)2

ditches where dug soil have been placed have higher elevation. These edges of the ditches have a higher bearing capacity than peatland far away from the ditch. DEM and CHM are basic elements provided in typical ALS projects aimed at estimating tree volumes in forest stands. It seems to be very promising that the information produced can also be utilised in making predictions on bearing capacity within peatland stands. However, mean stand volume predicted for 16 x 16 m2 grid cells may not be the most useful way of utilising information on trees detected by ALS. Future projects should concentrate on analysing the most cost-effective way to make predictions on the bearing capacity on peatlands. Information on the multitude of trees within the stand can be expressed in many alternative ways. Future projects should concentrate on identifying whether is worth calculating the volume of trees within a certain area or whether the CHM itself can provide sufficient information on the bearing capacity of peatland forests. It is also important that the most important factors affecting bearing capacity such as moisture content of peat, peat type, shear strength of the roots and the amount of cutting depris are measured in order to clarify the role of each factor.

5. References – Literatura Ala-Ilomäki, J., 2006: The effect of weather conditions on the trafficability of unfrozen peatlands. Forestry Studies – Metsanduslikud Uurimused 45: 57–66. Ala-Ilomäki, J., Högnäs, T., Lamminen, S., Sirén, M., 2011: Equipping a Conventional Wheeled Forwarder for Peatland Operations. International Journal of Forest Engineering 22(1): 7–13. Bollandsås, O. M., Maltamo, M., Gobakken, T., Næsset, E., 2011: Prediction of Timber Quality Parameters of Forest Stands by Means of Small Footprint Airborne Laser Scanner Data. International Journal of Forest Engineering 22(1): 14– 23. Bygdén, G., Eliasson, L., Wästerlund, I., 2004: Rut depth, soil compaction and rolling resistance when using bogie tracks. Journal of Terramechanics 40(3): 179–190. Evans, J. S., Hudak, A. T., 2007: A Multiscale Curvature Algorithm for Classifying Discrete Return LiDAR in Forested Environments. IEEE Transactions on Geoscience and Remote Sensing 45(4): 1029–1038. Finer, L., Laine, J., 1998: Root dynamics at drained peatland sites of different fertility in southern Finland. Plant Soil 201(1): 27–36. Holopainen, M., Haapanen, R., Tuominen, S., Viitala, R., 2008: Performance of airborne laser scanning- and aerial

335

J. Uusitalo et al. Utilizing Airborne Laser Scanning Technology in Predicting Bearing Capacity of Peatland Forest (329–337) photograph-based statistical and textural features in forest variable estimation. SilviLaser 2008, Sept. 17-19. Edinburgh, UK. 105–112 s.

ing and extraction machinery traffic on sensitive forest sites with peat soils. Forest Ecology and Management 180(1–3): 85–98.

Holopainen, M., Vastaranta, M., Rasinmäki, J., Kalliovirta, J., Mäkinen, A., Haapanen, R., Melkas, T., Yu, X., Hyyppä, J., 2010: Uncertainty in timber assortment predicted from forest inventory data. European Journal of Forest Research 129(6):1131–1142.

Peuhkurinen, J., Maltamo, M., Malinen, J., 2008: Estimating species-specific diameter distributions and saw log recoveries of boreal forests from airborne laser scanning data and aerial photographs: a distribution-based approach. Silva Fennica 42(4): 625–641.

Korpela, I., Ørka, H. O., Maltamo, M., Tokola, T., Hyyppä, J., 2010: Tree Species Classification Using Airborne LiDAR – Effects of Stand and Tree Parameters, Downsizing of Training Set, Intensity Normalization, and Sensor Type. Silva Fennica 44(2): 319–339.

Sirén, M., Ala-Ilomäki, J., Högnäs, T., 1987: Harvennuksiin soveltuvan metsäkuljetuskaluston maastokelpoisuus. Summary: Mobility of forwarding vehicles used in thinnings. Folia Forestalia 692. ISBN 951-40-0784-0. 60 p.

Laiho, R., Finer, L., 1996: Changes is root biomass after water-level drawdown on pine mires in Southern Finland. Scandinavian Journal of Forest Research 11(1–4): 251–260. Maltamo, M., Peuhkurinen, J., Malinen, J., Vauhkonen, J., Packalen, P., Tokola, T., 2009: Predicting tree attributes and quality characteristics of Scots Pine using airborne laser scanning data. Silva Fennica 43(3): 507–521. Nugent, C., Kanali, C., Owende, P. M. O., Nieuwnhuis, M., Ward, S., 2003: Characteristic site disturbance due to harvest-

Suvinen, A., 2006: Economic Comparison of the Use of Tyres, Wheel Chains and Bogie Tracks for Timber Extraction. Croatian Journal of Forest Engineering 27(2): 81–102. Suvinen, A., Saarilahti, M., 2006: Measuring the Mobility Parameters of Forwarders using GPS and CAN Bus Techniques. Journal of Terramechanics 43(2): 237–252. Zeleke, G., Owende, P. M. O., Kanali, C. L., Ward, S., 2007: Predicting the pressure-sinkage characteristics of two forest sites in Ireland using in situ soil mechanical properties. Biosystems Engineering 97(2): 267–281.

Sažetak

Primjena tehnologije laserskoga snimanja iz zraka za procjenu nosivosti tresetišta Lasersko snimanje iz zraka (Airbourne laser scanning – ASL) postaje sve zanimljivije u šumarstvu. Do sada je ASL primjenjivan većinom za dobivanje procjene stanja sastojine (npr. procjena drvne zalihe po hektaru). No neda vni projekti ALS-a potaknuli su zanimanje za primjenu ALS-a i u drugim područjima šumarstva. S gledišta prido bivanja drva tresetišta su vrlo problematična staništa zbog svoje osjetljivosti. Na tresetnim staništima srednje kubno stablo i sječna gustoća većinom su manji nego na mineralnim tlima, što utječe na proizvodnost harvestera i forvar dera. Mreža odvodnih kanala na tresetištima, koja je umjetno napravljena radi olakšanja odvodnje, smanjuje kretnost vozila te u prosjeku dva puta povećava srednju udaljenost privlačenja. Ograničena nosivost tla, karakteristična za tresetišta, značajan je čimbenik koji utječe na proizvodnost prilikom pridobivanja drva. Zbog ograničene nosivosti tla na tresetnim staništima posebna se pozornost treba posvetiti planiranju pridobi vanja drva. Tako bi se glavni izvozni putovi trebali postaviti da prolaze preko mjesta na kojima je najveći obujam stabala, a mrežu izvoznih putova treba planirati da se što više smanji broj prolazaka vozila po glavnim izvoznim putovima. Sve radove pridobivanja drva u sastojinama u kojima je nemoguće postaviti glavne izvozne putove na mjesta dobre nosivosti tla trebalo bi provoditi zimi dok je tlo smrznuto. Pomoću ALS-a možemo dobiti korisne infor macije koje će pomoći pri analizi nosivosti tla i prohodnosti tresetnih staništa prije početka sječe. Cilj je ovoga rada testirati mogućnost primjene digitalnoga modela terena i ALS-a, temeljenoga na procjeni obujma stabala pomoću metode ABSA (metoda koja se koristi podacima izvedenima na osnovi laserskih snimaka i podacima prikupljenim na terenu) za procjenu nosivosti tla na tresetnim staništima. Istraživanje je provedeno u poplavnoj borovoj šumi u južnoj Finskoj, blizu šumarije Hyytiälä, u Juupajoki. Područje obuhvaćeno istraživanjem protezalo se na 4,5 ha, srednji je prsni promjer iznosio 15 cm, temeljnica je bila 24 cm2/ha, a gustoća stabala 1500 stabala/ha. Istraživano je područje laserski snimano za potrebe ove studije nekoli ko puta 2004, 2006, 2008. i 2010. godine. Tijekom tih godina prikupljani su i terenski podaci, točni prsni promjeri stabala i visine su skupljene na više od 20 kružnih ploha, promjera 9 metara. Laserskim snimanjem dobiven je oblak podataka, gdje je svaka točka bila definirana koordinatom x, y i z. Iz dobivenih podataka zadnja je serija podataka

336

Croat. j. for. eng. 33(2012)2

Utilizing Airborne Laser Scanning Technology in Predicting Bearing Capacity of Peatland Forest (329–337) J. Uusitalo et al.

korištena za izradu digitalnoga modela terena, dok su podaci iz prve serije snimanja upotrijebljeni za izradu digital noga modela visine stabala. Visina stabala (koordinata z) izračunata je kao razlika između ortometrijske visine udara laserske zrake i procijenjene nadmorske visine za određenu lokaciju. Istraživano je područje pokriveno rasterskom mrežom 16 × 16 metara. Obujam stabala i temeljnica procjenjivani su za svaku rastersku plohu uz pomoć vrijednosti koordinate z, terenskih podataka i regresijskih modela. Sječa, izrada i izvoženje obavljeno je ljeti, nakon čega su mje rene dubine kolotraga. Istraživanjem odnosa obujma stabala i dubine kolotraga harvestera nije ustanovljena negativna korelacija, što je bilo iznenađujuće, dok je kao rezultat istraživanja odnosa obujma stabala i dubine kolotraga forvardera dobivena očekivana negativna korelacija. Nije utvrđeno zašto odnos obujma stabala i dubine kolotraga poprima neočekivanu vrijednost nakon sječe i izrade harvesterom, dok nakon izvoženja forvarderom poprima očekivanu vrijednost. U ovom istraživanju nije utvrđeno imaju li ostali čimbenici, kao što su vlažnost treseta, vrsta treseta, posmična čvrstoća ko rijenskoga sustava i količina šumskoga ostatka, utjecaj na dobivene rezultate. Moguće je da se izrazita smanjena nosivost pojavljuje samo nakon prolaska forvardera jer na mnogim plohama, naknon prolaska harvestera, nije utvrđen slom korijenskoga sustava koja značajno pridonosi povećanju nosivosti tla. Temeljnica, u odnosu na obujam stabala, u jačoj je ovisnosti o dubini kolotraga, nastalim nakon prolaska forvar dera, što je u skladu s prijašnjim istraživanjima, koja kazuju da se na tresetištima relativno veća količina biomase nalazi u korijenskom sustavu nego kod mineralnih tala. Iz toga slijedi zaključak, s aspekta nosivosti tla, da je bolje imati dva mala stabla nego jedno veliko. Na dubinu kolotragâ, nastalih nakon prolazka harvestera i nakon prolaska forvardera, relativna nadmorska visi na ima značajan utjecaj, što dolazi do izražaja jer u samim tresetištima ponekad postoje dijelovi na kojima je mine ralno tlo izdignuto od ostatka tresetišta, čak se znaju pojavljivati i dijelovi bez tresetnoga sloja. Nadalje, u tresetišti ma u kojima su prije kopani odvodni kanali iskopana je zemlja odlagana uz rubove odvodnih kanala te su zbog toga sad ti dijelovi nešto viši od okolnoga tresetišta i imaju bolju nosivost nego samo tresetište. Rezultati ovoga istraživanja upućuju na ALS kao obećavajuću tehnologiju koja će pripomoći u donošenju procje na nosivosti tresetišta prije početka sječe. Doduše, potrebno je provesti još dodatnih istraživanja kako bi se razvile što učinkovitije metode procjene. Uz to je ovo istraživanje, što je jedan od važnijih rezultata, pokazalo da se obujam stabla i relativna visina površine tla u tresetištu mogu okarakterizirati kao vrlo dobri čimbenici za procjenu nosivosti tla. Ključne riječi: pridobivanje drva, lasersko snimanje iz zraka, dubina kolotraga

Authors’ address – Adresa autorâ:

Received (Primljeno): June 1, 2012 Accepted (Prihvaćeno): August 13, 2012 Croat. j. for. eng. 33(2012)2

Prof. Jori Uusitalo, PhD. e-mail: jori.uusitalo@metla.fi Tomi Kaakkurivaara, MSc. e-mail: tomi.kaakkurivaara@metla.fi The Finnish Forest Research Insititute Western Finland Regional Unit Kaironiementie 15 39700 Parkano FINLAND Maarit Haavisto, MSc. e-mail: maarit.haavisto@stora-enso.com Stora-Enso Wood Supply FINLAND

337

Preliminary note – Prethodno priopćenje

Prospects of Research on Cable Logging in Forest Engineering Community Raffaele Cavalli Abstract – Nacrtak An analysis of researches on cable logging carried out in the past 12 years (2000–2011) as found in the scientific literature at international level is proposed in order to evaluate which have been the main topics of interest of the researchers and to evaluate the evolution of the research in the field of cable logging in the next future. International scientific literature on cable logging was extracted from the main databases, scientific journals and conference pro ceedings on forest engineering. A total of 244 references were retrieved and used to create a library implemented in Thompson Reuter EndNote® software. The analysis of the references through the use of some descriptors has allowed to define that in the period 2000–2011 most of the attention of forest engineers interested in cable logging was attracted by the study of the efficiency of the cable system, followed by the study and analysis of the impact produced by or derived from the environmental components by the use of cable logging. Cable system simula tion has played an important role, especially because of the variety of methodologies developed by forest engineers. Even if the number of references indicates some interest in cable system mechanics, most of the references deal with studies and tests about ropes, mainly synthetic ones. Carriage, yarders and supports seem to have been less attractive as objects of study. Ergonomics and safety in the field of cable logging have revealed a growing attention espe cially in the recent years, while an almost complete lack of interest in education and training was observed. Starting from such considerations, some prospects of the cable logging research for the next future were envisaged. Keywords: forest engineering, cable logging, scientific literature

1. Introduction – Uvod A way to assess the activity of a scientific community is to analyze its production in term of publications in scientific journals, communications at conferences, reports, etc. If such assessment is focused on a specific subject, the amount of the scientific production can be considered as a sort of indicator about the interest of scientists on such subject. Furthermore it is possible to focus the main issues that have directed the activity of the scientists and hence to evaluate which would be the prospects for the future. These considerations have driven the analysis of the researches on cable logging carried out in the past 12 years (2000–2011) as found in the scientific literature at international level. To retrieve the international scientific literature on cable logging the following databases were queried: Croat. j. for. eng. 33(2012)2

Þ Google Scholar, Þ Science Direct, Þ CAB Abstracts, Þ Current Contents, Þ Ingenta Connect, Þ Forest Science Database, Þ AGRIS International Information System for the Agricultural Sciences and Technology, Þ IUFRO On-line Library, Þ USDA National Forest Service Library, Þ USDA Treesearch Forest Service Research and Development. Each database was queried using the following keywords: »cable logging«; »cable yarding«; »cable crane«; »cable yarder«; »tower yarder« and the search for the records was only made considering those fully

339

R. Cavalli

Prospects of Research on Cable Logging in Forest Engineering Community (339–356)

written in English. The author was aware that this constraint excluded some important information sources of scientific literature on cable logging, e.g. the ones produced in Korea and in Japan, and so the survey was likely to be limited, but as explained below, the necessity to evaluate each record required that the linked manuscript was written in English. A further search was made on the main scientific journals related to forest engineering and available on the net and in the conference proceedings on forest engineering. For each reference the manuscript was examined and, if of interest to the research, it was imported into Thompson Reuter EndNote® software tool (2010) to create a library. Each reference contains the main bibliographic information together with the URL (Uniform Resource Locator), where a printed sample of the paper is available, and one or two keywords to allow the following elaboration on the information content of the library. For the purpose of the research, the keywords must be considered as a sort of descriptors useful to manage the references according to the analysis carried out; from this point of view two main subjects were considered: »Cable system« and »Cable logging«. The former includes all the descriptors dealing with cable extraction considered as a system; the latter considers the descriptors that refer to the relationship between the logging method and the environment, the operators and the management. Cable system efficiency: the descriptor or keyword refers to papers in which productivity and/or cost of cable system are considered; Cable system design: the descriptor or keyword refers to papers in which design, arrangement and patterns of cable system are considered; Cable system planning: the descriptor or keyword refers to papers in which not only the planning approach for cable system is considered, but also the relationship with forest road network and ancillary infrastructures; Cable system simulation: the descriptor or keyword refers to papers in which simulation techniques and modeling are considered to different extent; Cable system mechanics: the descriptor or keyword refers to papers in which mechanical features of cable system are concerned; in order to increase the discriminating capacity, second-level descriptors were introduced: »yarder«, »carriage«, »rope«, »support«, each one referring to mechanical features of a specific component of the cable system; Cable logging impact: the descriptor or keyword refers to papers in which the disturbance produced by

340

cable logging on environmental resources is considered; in order to increase the discriminating capacity, second-level descriptors were introduced: »vegetation«, »soil«, »water«, »air«, each one referring to the main environmental resource involved in the analysis of the cable logging impact; Cable logging ergonomics and safety: the descriptor or keyword refers to papers in which ergonomics and safety of cable logging, as a whole or as individual components, are considered; Cable logging education and training: the descriptor or keyword refers to papers in which educational and training approaches are applied to cable logging both as educational programs and teaching tools. Cable logging management: the descriptor or keyword refers to papers in which the relationship between cable logging and management practices is considered especially from the contractor’s perspective. The library was set up with a total of 243 recorded references; even though the number of records did not cover all the scientific literature about cable logging because of the English language constraint adopted in the literature review, the sample obtained could be considered valuable and fulfilling the aim of the research. The 243 recorded references refer to papers produced by a total of 354 authors.

2. Analysis of the research on cable logging – Analiza istraživanja iznošenja drva žičarama 2.1 Time distribution – Vremenska raspodjela The 243 recorded references are not homogenously distributed in the time period considered (2000–2011); as reported in Table 1, the average number of references per year was 20.3, but in one year the number of references per year was very much greater. This is the case of the year 2001 when 51 references were registered; the reason of such number is due to the fact that in 2001 a conference devoted to harvesting with cable systems was organized in Austria, accounting for 28 out of the total of 2001 references. Another year that deserves to be considered, not for the total number of references but for the source of such references, is the year 2011: even though the number of references was almost the same as the average of the period (20 vs. 20.3), 12 of them belong to the same conference. A similar situation was recorded in the year 2007, when half (11) of the annual references (22) came from the same conference. Croat. j. for. eng. 33(2012)2

Prospects of Research on Cable Logging in Forest Engineering Community (339–356)

Table 1 Recorded references in the period 2000–2011 Tablica 1. Zabilježene objave u razdoblju 2000–2011. Year – Godina

References – Objave

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

Total – Ukupno

243

Mean – Prosječno

20.3

The analysis of each reference related to the »Cable system« through the main descriptor showed that the »Cable system efficiency« is the most frequent descriptor (Table 3) with 78 references out of 172. The second largest descriptor is »Cable system simulation«. It can be concluded that the interest of the researchers was mainly devoted to the analysis of cable system from productivity and/or economic point of view and it should be noted that 15 references out of 80 dealing with the descriptor »Cable system efficiency« are related to experiments carried out in Turkey, highlighting the interest of the Turkish forest engineering researchers in such studies.

The main sources of scientific literature were the conference proceedings (Table 2), followed by the journal articles and reports. Under the form of book sections only two references were retrieved, while no book was available concerning cable logging. One gets the distinct impression that, compared to 10 or 20 years before the analyzed period, cable logging tended to be regarded less as the subject of a book, and more as the subject of journal articles or conference proceedings. From a certain point of view,

Table 2 Scientific literature sources Tablica 2. Izvori znanstvene literature References – n Objave – n

Journal article – Članak u časopisu

Conference proceedings – Zbornik radova

153

Book section – Poglavlje u knjizi

Report – Izvještaj

Total – Ukupno

243

Croat. j. for. eng. 33(2012)2

writing and publishing a book seems to be a time consuming activity, and however in the field of cable logging, fundamental books like the ones published in the ‘70s and ‘80s of the past century are highly needed.

2.3 »Cable system« – »Žični sustavi«

2.2 Scientific literature sources – Izvori znanstvene literature

Source – Izvori

R. Cavalli

»Cable system simulation«, accounting for around 21% of the references related to the »Cable system«, reveals its potential as a research issue. Cable system simulation is a broad term that considers Optimization Techniques, DDS (Decision Support System), Network Analysis, Dynamic Model, Finite Element Modeling; fields of application include cable logging layout, cable crane location, forest road spacing (using cable logging), cable tension analysis, fuel consumption, productivity, harvesting cost. »Cable system planning« accounts for 13% of the references related to the »Cable system« and generally refers to a harvest layout planning approach for cable-based system. Some references also refer to updated photogrammetric techniques through which the reliability of the data used in planning logging operation can be enhanced. Logging planning is considered essential to successful implementation of cable systems and also to effective implementation of BMPs (Best Management Practices). To analyze »Cable system mechanics«, which accounts for 16% of the references of the »Cable system«, it is necessary to consider four second-level descriptors, which represent the main components of a cable system from a mechanical point of view (Table 4). It is impressive to note that the »rope« descriptor includes 20 out of 28 references of the main descriptor (»Cable system mechanics«); it seems that most of the scientific interest on the mechanics of the cable system has been drawn by ropes if compared to other components. This situation can be understood if one considers that 16 references refer to research papers dealing with synthetic ropes; starting from the end of the last

341

R. Cavalli

Prospects of Research on Cable Logging in Forest Engineering Community (339–356)

Table 3 Recorded references in the period 2000–2011 broken down according to different main descriptors related to »Cable system« Tablica 3. Zabilježene objave teme »žični sustavi« u razdoblju 2000–2011. razdijeljene prema različitim deskriptorima Cable system – Žični sustavi Year

Efficiency

Design

Simulation

Planning

Mechanics

Godina

Djelotvornost

Projektiranje

Simulacija

Planiranje

Mehanika

n 2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

Total – Ukupno

Table 4 Recorded references in the period 2000–2011 related to »Cable system mechanics« descriptor and broken down according to second-level descriptors Tablica 4. Zabilježene objave vezane uz deskriptor »mehanika žičnih sustava« u razdoblju 2000–2011. te razdijeljene prema drugoj razini deskriptora Year

Rope – Uže

Carriage – Kolica

Godina

Yarder – Žičara

Support – Potporanj

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

Total – Ukupno

342

Croat. j. for. eng. 33(2012)2

Prospects of Research on Cable Logging in Forest Engineering Community (339–356)

R. Cavalli

table clearly reveal that the references mainly refer to the impact of cable logging on vegetation and on soil, while very few of them are linked to the impact on water and air. However it must be said that the boundary between soil impact and water impact is really vague because of the interaction that links soil disturbance to sediment production and hence stream water quality. Regarding the impact of the cable logging on the air, it is interesting to observe that the papers dealing with such topic are quite new, the oldest ones being published in 2006. »Cable logging ergonomics and safety« accounts for 17 references, many of which are related to the workload benefit using synthetic ropes and radio controlled chokers. It is interesting to note that such references refer to the last two years of the considered period and include 10 out of 17 papers. It is disheartening to notice that »Cable logging education and training« contains only three references, two of them referring to papers published at the beginning of the period! The descriptor »Cable logging management« has offered the opportunity to include references that deal with the connection between cable logging and man-

century synthetic ropes have been introduced into forest use and, due to their potential to replace steel wire rope for selected logging operations, they have required comprehensive analysis and evaluations. During the same period only four references consider steel wire rope and their behavior during operation.

2.4 »Cable logging« – »Iznošenje drva žičarama« »Cable logging« accounts for 71 references and the descriptor »Cable logging impacts« covers 44 of them (62%). The descriptor »Cable logging ergonomics and safety« includes 19 (27%) references, while the references covered by the descriptors »Cable logging education and training« and »Cable logging management« are minimal. »Cable logging impacts« refers to a wide range of papers, some of them strictly reporting the consequences of cable logging on soil or on residual stands, other analyzing the effects of forest practices, in which cable logging is considered, on the environmental components, including also visual disturbance. Four second-level descriptors were used to improve the accuracy of the classification: »vegetation«, »soil«, »water« and »air« (Table 6). The data from the

Table 5 Recorded references in the period 2000–2011 broken down according to different main descriptors related to »Cable logging« Tablica 5. Zabilježene objave teme »iznošenje drva žičarama« u razdoblju 2000–2011. razdijeljene prema različitim deskriptorima Cable logging – Iznošenje drva žičarama Year

Impact

Ergonomics and safety

Education and training

Management

Godina

Utjecaj

Ergonomija i sigurnost

Obrazovanje i obuka

Upravljanje

n 2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

Total – Ukupno

Croat. j. for. eng. 33(2012)2

343

R. Cavalli

Prospects of Research on Cable Logging in Forest Engineering Community (339–356)

Table 6 Recorded references in the period 2000–2011 related to »Cable logging impact« descriptor and broken down according to secondlevel descriptors Tablica 6 Zabilježene objave vezane uz deskriptor »utjecaj iznošenja drva žičarama« u razdoblju 2000– 2011. te razdijeljene prema drugoj razini deskriptora Year

Vegetation – Vegetacija

Soil – Tlo

Godina

Air – Zrak

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

Total – Ukupno

agement practices and to provide the contractors’ perspective about different problems that must be considered when cable logging is carried out. The references covered by the descriptor »Cable logging management« do not refer to scientific papers in the strict sense of the word, but they contribute to understanding how cable logging is managed and which the daily problems are that must be faced by contractors.

3. Future prospects of the research on cable logging – Budućnost istraživanja iznošenja drva žičarama The author is aware of the limits of the library he has set up due to the language constraint and to the number of information sources that were queried; he is also aware of the subjectivity that may have affected the evaluation of the main content of each paper trying to assign the proper descriptor. Anyway the methodology and the analysis meet the requirements of a scientific approach and hence it is possible to draw some conclusions and to point out some prospects to the forest engineering community. In the period 2000–2011 most of the attention of forest engineers interested in cable logging was at-

344

Water – Voda

tracted by the study of the efficiency of the cable system, followed by the study and analysis of the impact produced by or derived from the environmental components by the use of cable logging. Cable system simulation has played an important role, especially for the variety of methodological approaches developed by forest engineers; similar considerations can be made when cable system planning is concerned. Different conclusions can be drawn analyzing cable system mechanics; even if the number of references indicates some interest in such topic, it must be emphasized that most of the references deals with studies and tests about ropes, mainly synthetic ones. Carriage, yarders and supports seem to have been less attractive as objects of study. Ergonomics and safety in the field of cable logging have revealed a growing attention especially in the recent years, while education and training were affected by an almost total lack of interest. In order to suggest which vision of the cable logging research will characterize the years to come, it is important to recall a concept by Prof. Heinimann (2000), that still retains its full validity after a decade and can be repeated again as a basis for the development of cable logging: it must be considered that individual technologies will not be enough to face the challenges of the coming years but total systems, which Croat. j. for. eng. 33(2012)2

Prospects of Research on Cable Logging in Forest Engineering Community (339–356)

include expertise, procedures, goods and services, and equipment as well as organizational and managerial procedures, must be envisaged. Starting from such holistic considerations, the prospects of the cable logging research for the next future could include: Þ To develop new materials for the ropes (both synthetic and steel wire ropes) in order to increase the rope strength, allowing an increment of system transport capacity; Þ To develop mechatronics application on carriages and yarders, increasing work efficiency and ergonomics and empowering human-machine interface; Þ To improve the feasibility of computerized methods for cable logging planning; Þ To improve mathematical methods to optimize structural analysis of a cable structure; Þ To optimize the energy consumption, reducing as far as possible the energy derived from fossil fuels and increasing the utilization of electric energy and gravitational energy; Þ To investigate the cable assisted ground vehicle technology that represents a border area between terrestrial and air logging systems; Þ To improve the use of tools like Life Cycle Assessment (LCA) to evaluate the environmental issues of cable-based technologies; Þ To improve education and training in cable logging operations, developing educational methodologies and teaching tools that acknowledge the critical importance of the learner in all aspects of the learning process.

4. References – Literatura EndNote Rel. X4.0.2. Thompson Reuter Heinimann, H. R., 2000: Forest operations under mountainous conditions. In: Forests in Sustainable Mountain Development – a State of Knowledge Report for 2000, M.F. Price and N. Butt, Editors. CABI Publishing: Wallingford, UK. Vol. IUFRO Research Series No. 5: p. 224–230.

5. References library – Literatura zbirke Acar, H. H., 2006: Timber Extraction by Cable Cranes, Monorail and Chute Systems in Turkish Forestry. In: Proceedings of COFE Conference, July 30 – August 2. Coeur d’Alene Resort, Coeur d’Alene, Idaho. Croat. j. for. eng. 33(2012)2

R. Cavalli

Acar, H. H., Eroglu, H., Ozkaya, M. S., 2010: An investigation on roundwood extraction and determination of the physical damages on residual trees and seedlings due to logging operation using URUS MIII forest skyline on snow. In: Proceedings of FORMEC Conference, July 11–14, Padova, Italy. Acar, H. H., Topalak, O., Eroglu, H., 2006: Forest skylines in Turkish forestry. Meh. šumar. 2001–2004, Special Issue of the Journal Nova meh. šumar., Vol. 26(2005), Issue 2: 137–140. Acar, H. H., Unver, S., Ozkaya, M. S., Kilic, H., 2011: Determination of efficiency of the forest skylines in Artvin Forest Region of Turkey. In: Proceedings of FORMEC Conference, October 9–13, Graz, Austria. Ackermann, P., Talbot, B., 2004: Reverting urban exotic pine forests to Macchia and indigenous forest vegetation using cable-yarders on the slopes of Table Mountain, South Africa. In: Proceedings of Conference Forestry Serving Urbanised Societies, August 27–30, Copenhagen, Denmark. Adnan, A. M., 2002: Cable logging technique using a mobile tower yarder for low impact logging in Malaysian forest. In: Proceedings of International seminar on new roles of plantation forestry requiring appropriate tending and harvesting operations, September 29 – October 5, Tokio, Japan. Amishev, D., Evanson, T., 2010: Innovative methods for steep terrain harvesting. In: Proceedings of FORMEC Conference, July 11–14, Padova, Italy. Anderson, L., Temen, K. L., 2000: Cable Thinning as a Business Partnership between Landowner and Contractor. In: Proceedings of International Mountain Logging and 10th Pacific Northwest Skyline Symposium, March 28 – April 1, Corvallis, Oregon. Aricak, B., 2002: Transportation of timber by cable systems in Turkish forestry. In: Proceedings of International seminar on new roles of plantation forestry requiring appropriate tending and harvesting operations, September 29 – October 5, Tokio, Japan. Arriagada, R., Cubbage, F. W., Abt, K. L., Huggett, R. J. Jr., 2008: Estimating harvest costs for fuel treatments in the West. Forest Products Journal 58(7–8): 24–30. Aruga, K., Tasaka, T., Yoshioka, T., 2008: Long-term feasibility of timber and forest biomass resource extraction in a mountainous area – reducing harvesting costs with new harvesting systems. In: Proceedings of IUFRO All-D3 Conference, June 15–20, Sapporo, Japan. Asikainen, A., Stampfer, K.,Talbot, B., 2010: An evaluation of skyline systems in Norwegian conditions using discreteevent simulation. In: Proceedings of Precision Forestry Symposium, March 1–3, Stellenbosch, South Africa. Asikainen, A., Stampfer, K., Talbot, B., Belbo, H., 2010: Simulation of skyline systems in Norwegian conditions. In: Proceedings of 2010 Nordic Baltic Conference on Forest Operations, October 20–22, Honne, Norway. Aulerich, S., 2000: Commercial Thinning with Cable Yarding Systems. In: Proceedings of COFE Conference, September

345

R. Cavalli

Prospects of Research on Cable Logging in Forest Engineering Community (339–356)

11–14, Joint Meeting with Canadian Woodlands Forum, Kelowna, British Columbia. Aulerich, S., 2000: Raising a Spar Tree – Technique of the Past Solves Challenge of Today. In: Proceedings of COFE Conference, September 11–14. Joint Meeting with Canadian Woodlands Forum, Kelowna, British Columbia. Bailey, P., 2007: A contractor’s perspective on skyline thinning equipment and logging innovations. In: Proceedings of International Mountain Logging and 13th Pacific Northwest Skyline Symposium, April 1–6, Corvallis, Oregon. Baker, S., Sloan, H., Visser, R., 2001: Cable logging in Appalachia and opportunities for automated yarder equipment. In: Proceedings of COFE Conference. July 15–19, Snowshoe Resort, Snowshoe, West Virginia. Baldini, S., Calvani, P., Picchio, R., 2001: Winch use in work with extra light cable systems in South-Central Italy. In: Proceedings of Workshop New trends in wood harvesting with cable systems for sustainable forest management in the mountains, June 18–24, Ossiach, Austria. Beese, W. J., Dunsworth, B. G., Zielke, K., Bancroft, B., 2003: Maintaining attributes of old-growth forests in coastal BC through variable retention. The Forestry Chronicle 79(3): 570–578. Beschta, R. L., Jackson, W. L., 2008: Forest practices and sediment production in the Alsea watershed study. Hydrological and Biological Responses to Forest Practices: 55–66. Beschta, R. L., Rhodes, J. J., Kauffman, J. B., Gresswell, R. E., Minshall, G. W., Karr, J. R., Perry, D. A., Hauer, F. R., Frissell, C. A., 2004: Postfire management on forested public lands of the western United States. Conservation Biology 18(4): 957– 967. Bolding, M. C., Lanford, B. L., Kellogg, L. D., 2003: Forest fuel reduction: current methods and future possibilities. In: Proceedings of COFE Conference, September 7–10. Atlantic Oakes by the Sea Resort, Bar Harbor, Maine. Borz, S. A., Birda, M., Ignea, G., Oprea, I., 2011: Technological aspects regarding timber exploitation using Mounty 4100 cable yarder. In Serie II: Forestry-Wood Industry-Agricultural Food Engineering. Boswell, B., 2001: Partial cutting with a cable yarding system in coastal British Columbia. Advantage 42(2). Boyce, J. A., 2005: Blending stand-level treatments and landscape planning with opportunities for research in a working forest. Gen. Tech. Rep. PNW-GTR-635, U.S. Department of Agriculture, Forest Service, Pacific Northwest Research Station, Portland, Oregon, 389 p. Caglar, S., 2009. An investigation on productivity of Gantner yarder at windblown forest stand in Turkey. In: Proceedings of FORMEC Conference, June 2–5, Schmallenberg, Germany. Caglar, S., Acar, H. H., 2009: A review on the forest skylines and their productivities at mountainously eastern Black Sea region of Turkey. In: Proceedings of FORMEC Conference, June 21–24, Kostelec nad Cernymi lesy, Czech Republic.

346

Cavalli, R., Grigolato, S., 2010: Influence of characteristics and extension of a forest road network on the supply cost of forest woodchips. Journal of Forest Research 15(3):202–209. Cavalli, R., Grigolato, S., Lubello, D., 2006: Planning logging systems through site analysis. In: Proceedings of Precision Forestry Symposium, March 5–10, Stellenbosch, South Africa. Cavalli, R., Ientile, F., Menegus, G., 2004: Cable crane use under sustainable forest managment in North-eastern Italian Alps. In: Proceedings of Conference Cable yarding suitable for sustainable forest management, September 23, Idrija, Slovenia. Cavalli, R., Lubello, D., 2006: Firewood cable yarding on steep terrain in Alpine conditions. In: Proceedings of FORMEC Conference, September24–28, Sofia, Bulgaria. Cavalli, R., Menegus, G., De Benedet, B., 2001: Cable crane utilization in the close-to-nature silviculture in the upper Piave river valley (Belluno province, Northern Italy). In: Proceedings of Workshop New trends in wood harvesting with cable systems for sustainable forest management in the mountains, June 18–24, Ossiach, Austria. ChihLung, C., YuLun, S., BingShu, G., ShaoJyun, H., ShihYao, G., 2009: Operational efficiency of thinning for the mixed plantation stand located at Tai-Ping-Shan working circle. Forest Products Industries 28(3): 149–166. Chung, W., 2002: Optimization of cable logging layout using a heuristic algorithm for network programming. In: Proceedings of COFE Conference, June 16–20. Auburn University, Auburn, Alabama. Chung, W., Seielstad, C., Sessions, J., 2008: Sloperunner2.0: a computer program for cable logging unit planning. In: Proceedings of IUFRO All-D3 Conference, June 15–20, Sapporo, Japan. Chung, W., Sessions, J., 2003: A computerized method for determining cable logging feasibility using a DEM. In: Proceedings of COFE Conference, September 7–10. Atlantic Oakes by the Sea Resort, Bar Harbor, Maine. Chung, W., Sessions, J., Heinimann, H. R., 2001: Optimization of cable harvesting equipment placement and road locations using digital terrain models. In: Proceedings of Workshop New trends in wood harvesting with cable systems for sustainable forest management in the mountains, June 18– 24, Ossiach, Austria. Chung, W., Sessions, J., Heinimann, H. R., 2004: An application of a heuristic network algorithm to cable logging layout design. International Journal of Forest Engineering 15(1). Chung, W., Sessions, J., Holub, J., 2007: Sloperunner 1.0: a program to evaluate the effectiveness of landing and road locations for cable logging. In: Proceedings of International Mountain Logging and 13th Pacific Northwest Skyline Symposium, April 1–6, Corvallis, Oregon. Clinton, B. D., 2011: Stream water responses to timber harvest: Riparian buffer width effectiveness. Forest ecology and management 261: 979–988. Croat. j. for. eng. 33(2012)2

Prospects of Research on Cable Logging in Forest Engineering Community (339–356) Covert, S. A., Robichaud, P. R., Elliot, W. J., Link, T. E., 2005: Evaluation of runoff prediction from WEPP-based erosion models for harvested and burned forest watersheds. Transactions of the ASAE 48(3):1091–1100. Deal, R. L., Hennon, P. E., Orlikowska, E. H., D’Amore, D. V., 2004: Stand dynamics of mixed red alder-conifer forests of southeast Alaska. Canadian Journal of Forest Research 34(4):969–980. Dicus, C., 2003: Fueling loading and potential fire behavior after selective harvest in coast redwood stands. In: Proceedings of 2nd International Wildland Fire Ecology and Fire Management Congress, November 16–20, Orlando, Florida. Drews, E. S., Hartsough, B. R., Doyal, J. A., Kellogg, L. D., 2001: Harvester-forwarder and harvester-yarder systems for fuel reduction treatments. International Journal of Forest Engineering 12(1): 81–91. Dykstra, D. P., Monserud, R. A., 2007: Influence of Forest Operations on Timber Quality. In: Proceedings of International Conference Forest Growth and Timber Quality: Crown Models and Simulation Methods for Sustainable Forest Management, August 7–10, Portland, Oregon. Eker, M., Acar, H. H., 2001: Gantner skyline for timber extraction in Turkish forestry. In: Proceedings of Workshop New trends in wood harvesting with cable systems for sustainable forest management in the mountains, June 18–24, Ossiach, Austria. Eklund, A., Wing, M. G., Sessions, J., 2009: Evaluating economic and wildlife habitat considerations for snag retention policies in burned landscapes. Western Journal of Applied Forestry 24(2): 67–75. Epstein, R., Weintraub, A., Sapunar, P., Nieto, E., Sessions, J. B., Sessions, J., Bustamante, F., Musante, H., 2006: A combinatorial heuristic approach for solving real-size machinery location and road design problems in forestry planning. Operations Research54(6): 1017–1027. Eroglu, H., 2009: Timber harvesting by URUS M III forest skyline on snow in steep terrain; a case from Artvin, Turkey. In: Proceedings of FORMEC Conference, June 2–5, Schmallenberg, Germany. Eroglu, H., Ozkaya, M. S., Acar, H. H., Karaman, A., Yolasigmaz, H. A., 2010. An investigation on roundwood extraction of Fagus orientalis Lipsky, Abies nordmanniana (Stew.) Spach. and Picea orientalis (L.) Link. by Urus M III forest skyline on snow. African Journal of Biotechnology 8(6): 1082–1089. Evanson, T., Amishev, D., 2010. Productivity impacts of bunching for yarder extraction. In: Proceedings of FORMEC Conference, July 11–14, Padova, Italy. Fabiano, F., Magagnotti, N., Neri, F., Piegai, F., Spinelli, R., 2010: Safety in mechanised forest operations: a Tuscan project. In: Proceedings of FORMEC Conference, July 11–14, Padova, Italy. Fabiano, F., Marchi, E., Neri, F., Piegai, F., 2011: Skyline tension analysis in yarding operation: cases studies in Italy. In: Croat. j. for. eng. 33(2012)2

R. Cavalli

Proceedings of FORMEC Conference, October 9–13, Graz, Austria. Fight, R. D., 2003: Users guide for STHARVEST: software to estimate the cost of harvesting small timber. Gen. Tech. Rep. PNW-GTR-582, U.S. Department of Agriculture, Forest Service, Pacific Northwest Research Station, Portland, Oregon, 11 p. Fight, R. D., Hartsough, B. R., Noordijk, P., 2006: Users guide for FRCS: fuel reduction cost simulator software. Gen. Tech. Rep. PNW-GTR-668, U.S. Department of Agriculture, Forest Service, Pacific Northwest Research Station, Portland, Oregon, 23 p. Fischbacher, M., Mairhofer, M., 2007: New skyline logging technology for yarding and tree processing with a two person crew. In: Proceedings of International Mountain Logging and 13th Pacific Northwest Skyline Symposium, April 1–6, Corvallis, Oregon. Fladl, H., Loschek, J., 2001: Employment of contractors in cable crane operations in Austria. In: Proceedings of Workshop New trends in wood harvesting with cable systems for sustainable forest management in the mountains, June 18– 24, Ossiach, Austria. Gandaseca, S., Hulusi, A., Yoshimura, T., 2001: Occupational safety and health of forestry workers cable harvesting in Turkey. In: Proceedings of Workshop New trends in wood harvesting with cable systems for sustainable forest management in the mountains, June 18–24, Ossiach, Austria. Garland, J. J., 2001: The basis for productivity improvement in cable harvesting from human resources. In: Proceedings of Workshop New trends in wood harvesting with cable systems for sustainable forest management in the mountains, June 18–24, Ossiach, Austria. Garland, J. J., Pilkerton, S. J., 2007: Synthetic rope reduces workloads in logging. In: Proceedings of International Mountain Logging and 13th Pacific Northwest Skyline Symposium, April 1–6, Corvallis, Oregon. Garland, J. J., Sessions, J., Pilkerton, S. J., Hartter, J., Leonard, J. M., 2003: Synthetic rope to replace wire rope in mountain logging operations. In:Proceedingsof AUSTRO 2003 Conference, October, 5–9, Schlaegl, Austria. Garland, J. J., Sessions, J., Pilkerton, S. J., Leonard, J. M., Hartter, J., 2003: Synthetic rope to replace wire rope in logging. In: Proceedingsof 2nd Forest Engineering Conference, May 12–15, Växjö, Sweden. Garland, J. J., Sessions, J., Pilkerton, S. J., Stringham, B., 2001: Synthetic rope used in logging: Some Potentials. In: Proceedings of COFE Conference, July 15–19. Snowshoe Resort, Snowshoe, West Virginia. Ghaffariyan, M. R., Sessions, J., 2011. Error Impact of Regression Models on Forest Road Spacing. Silva Balcanica 1(12):97–111. Ghaffariyan, M. R., Stampfer, K., Sessions, J., 2010: Optimal road spacing of cable yarding using a tower yarder in South-

347

R. Cavalli

Prospects of Research on Cable Logging in Forest Engineering Community (339–356)

ern Austria. European Journal of Forest Research 129(3): 409–416.

Hamilton, P. S., 2008: Life expectancy of synthetic fibre mainlines in cable skidding applications. In: Advantage 10(10).

Ghaffariyan, M. R., Stampfer, K., Sessions, J., 2009: Using network analysis to optimize forest road network for cable logging. In: Proceedings of FORMEC Conference, June 2–5, Schmallenberg, Germany.

Han, H. S., 2007: Economics of Soil Disturbance. In: Proceedings of Conference on Volcanic-Ash-Derived Forest Soil in the Inland Northwest: Properties and Implications for Management and Restoration, November 9–10, 2005, Coeur d’Alene, ID

Ghaffariyan, M. R., Stampfer, K., Sessions, J., 2009: Production Equations for Tower Yarders in Austria. International Journal of Forest Engineering 20(1): 17–21. Ghaffariyan, M. R., Stampfer, K., Sessions, J., 2010. Optimal road spacing of cable yarding using tower yarder in Southern Austria. European Journal of Forest Research 129(3): 409–416. Ghaffariyan, M. R., Stampfer, K., Sessions, J., Durston, T., Kuehmaier, M., Kanzian, C. H., 2010: Road network optimization using heuristic and linear programming. Journal of Forestry Science 56(3): 137–145. Gingras, J. F., 2000: Update on the viability of cable yarding in eastern Canada. Advantage 1(9). Giovannini, G., Cavalli, R., Grigolato, S., 2009: Improving felling and thinning for coppice and young high forest stands in alpine condition – the case of Trento province. In: Proceedings of FORMEC Conference, June 2–5, Schmallenberg, Germany. Gomi, T., Sidle, R. C., Swanston, D. N., 2004: Hydrogeomorphic linkages of sediment transport in headwater streams, Maybeso Experimental Forest, southeast Alaska. Hydrological processes 18(4): 667–683. Greulich, F. E., 2000: Calculation and use of effective external boundary and related setting parameters in cable yarding production estimation. Gen. Tech. Rep. NC-205, U.S. Department of Agriculture, Forest Service, North Central Forest Experiment Station. pp 7. Greulich, F. E., 2003: Transportation networks in forest harvesting: early development of the theory. In: Proceedings of International seminar on new roles of plantation forestry requiring appropriate tending and harvesting operations, September 29 – October 5, Tokio, Japan. Grulois, S., 2007: Cable yarding in France: past, present and perspective. In: Proceedings of International Mountain Logging and 13th Pacific Northwest Skyline Symposium, April 1–6, Corvallis, Oregon. Gumus, S., Acar, H. H., Ozturk, M., 2010: Evaluation of consecutive skylines yarding and gravity skidding systems in primary forest transportation on steep terrain. Journal of Environmental Biology 31(1–2): 213–218. Hakso, A., 2009: Biofuels from precommercial thinning operations – A feasibility survey. In: Proceedings of 1st Annual Forest Engineering, Resources and Management Department Graduate Student Symposium, June 4, Corvallis, Oregon. Halpern, C. B., McKenzie, D., 2001: Disturbance and postharvest ground conditions in a structural retention experiment. Forest ecology and management 154(1–2): 215–225.

348

Han, H. S., Kellogg, L. D., 2000: Damage characteristics in young Douglas-fir stands from commercial thinning with four timber harvesting systems. Western Journal of Applied Forestry 15(1): 27–33. Hartley, D. S., Han, H. S., 2007: Effects of alternative silvicultural treatments on cable harvesting productivity and cost in western Washington. Western Journal of Applied Forestry 22(3): 204–212. Hartsough, B. R., Zhang, X., Fight, R. D., 2001: Harvesting cost model for small trees in natural stands in the interior northwest. Forest Products Journal 51(4): 54–61. Hartter, J., Garland, J. J., 2006: Synthetic rope end connections for use in timber harvesting. International Journal of Forest Engineering 17(1): 39–51. Hartter, J., Leonard, J. M., Garland, J. J., Pilkerton, S. J., 2006: Research on synthetic rope and its future in timber harvesting. New Zealand Journal of Forestry 51(2): 23–30. Haynes, H., Visser, R., 2001: Productivity improvements through professional training in Appalachian cable logging operations. In: Proceedings of The International Mountain Logging and 11th Northwest Pacific Skyline Symposium, December 10–12, Seattle, Washington. Haynes, H., Visser, R., Sloan, H., 2002: A Case Study on Cable Yarding for a Stream Habitat Treatment and the Utility of a »Roadside« Log Sale. In: Proceedings of COFE Conference, June 16–20, Auburn University, Auburn, Alabama. Heinimann, H. R., 2001: Approaches to the design of forest cable systems. In: Proceedings of Workshop New trends in wood harvesting with cable systems for sustainable forest management in the mountains, June 18–24, Ossiach, Austria. Heinimann, H. R., 2000: Forest Operations under Mountainous Conditions. In: Forests in Sustainable Mountain Development – a State of Knowledge Report for 2000, edited by M. F. Price and N. Butt. Wallingford: CABI Publishing. Heinimann, H. R., 2004: Forest Operations under Mountainous Conditions. In: Encyclopedia of Forest Sciences, edited by J. Burley, J. Evans and J. Youngquist. Amsterdam: Elsevier Academic Press. Heinimann, H. R., Stampfer, K., 2003: Harvest Layout Planning for High-Altitude Protection Forests. In: AUSTRO 2003 Conference Proceedings, October, 5–9, Schlaegl, Austria. Heinimann, H. R., Stampfer, K., Loschek, J., Caminada, L., 2001: Perspectives on Central European Cable Yarding Systems. In: Proceedings of The International Mountain Logging and 11th Northwest Pacific Skyline Symposium, December 10–12, Seattle, Washington. Croat. j. for. eng. 33(2012)2

Prospects of Research on Cable Logging in Forest Engineering Community (339–356) Hessburg, P., Goheen, D. J., Koester, H., 2001: Association of black stain root disease with roads, skid trails, and precommercial thinning in Southwest Oregon. Western Journal of Applied Forestry 16(3):127–135. Horek, P., Mauer, P., 2001: Forest cableways in shelterwood system. In: Proceedings of Workshop New trends in wood harvesting with cable systems for sustainable forest management in the mountains, June 18–24, Ossiach, Austria. Horek, P., Novak, J., Neruda, J., 2009: Forest cableways and their use in forest management. In: Proceedings of FORMEC Conference, June 2–5, Schmallenberg, Germany. Hosseini, S. M., Madjnonian, B., Nieuwenhuis, M., 2000: Damage to Natural Regeneration in the Hyrcanian Forests of Iran: A Comparison of Two Typical Timber Extraction Operations. International Journal of Forest Engineering 11(2): 69–73. Huettner, W., 2001: Aspects of mechanical engineering for cable systems. In: Proceedings of Workshop New trends in wood harvesting with cable systems for sustainable forest management in the mountains, June 18–24, Ossiach, Austria. Hulusi, A., Unver, S., 2004: The environmental impact assessment of timber extraction by cable cranes, monorail and chute systems in Turkish forestry. In: Science Bulletin, edited by Ukrainian National Forestry University. Ishikawa, T., Ishidoh, Y., Yamazaki, H., Shiba, M., Yarugi, Y., 2004: Development of simplified manual cable yarding methods for thinning operations by volunteer workers. Journal of the Japan Forest Engineering Society 18(4): 259–262. Ito, T., Uemura, T., 2011: Automatic control for a self-propelled carriage to enable one-man cable yarding. In:Proceedings of FORMEC Conference, October 9–13, Graz, Austria. Jaross, W. S., 2001: Techniques for Preparing Alternative Road Access Policies for Steep and Mountainous Terrain. In: Proceedings of The International Mountain Logging and 11th Northwest Pacific Skyline Symposium, December 10– 12, Seattle, Washington. Johnson, L. R., Lippke, B., Marshall, J. D., Comnick, J., 2005: Life-cycle impacts of forest resource activities in the Pacific Northwest and Southeast United States. Wood and fiber science 37: 30–46. Kato, A., Schiess, P., 2007: LIDAR-derived tree parameters for optimal cable logging system design. In: Proceedings of International Mountain Logging and 13th Pacific Northwest Skyline Symposium, April 1–6, Corvallis, Oregon. Kerstetter, J. D., Lyons, J. K., 2001: Logging and agricultural residue supply curves for the Pacific Northwest. Washington State University Energy Publication. King, R., 2007: Long span skyline logging: past application and current niche. In: Proceedings of International Mountain Logging and 13th Pacific Northwest Skyline Symposium, April 1–6, Corvallis, Oregon. Croat. j. for. eng. 33(2012)2

R. Cavalli

Kirk, P. M., Sullman, M. J. M., 2001: Heart rate strain in cable hauler choker setters in New Zealand logging operations. Applied Ergonomics 32(4): 389–398. Kirth, R., Schimier, S., Nemestóthy, N., Sperrer, S., 2007: Further developments of synthetic ropes for logging applications in forestry. In: Proceedings of FORMEC Conference, October 7–11, Vienna and Heiligenkreutz, Austria. Klepac, J., Reutebuch, S. E., 2003: Preliminary Soil Disturbance Assessment of a Skyline Logging Operation Performing Five Silvicultural Prescriptions. In: Proceedings of ASAE International Meeting, July 27–30, Las Vegas, Nevada. Klun, J., Robek, R., 2009. GPS-based design of skyline corridors and software solutions for analyzing of cableway system. In: Proceedings of FORMEC Conference, June 2–5, Schmallenberg, Germany. Klvac, R., Fischer, R., Skoupy, A., 2009: Emissions from the »Larix« cableway system operation phase. In: Proceedings of FORMEC Conference, June 2–5, Schmallenberg, Germany. Knobloc, C., 2011: Cost-effective short-wood logging cable way system for lowland stands. In: Proceedings of FORMEC Conference, October 9–13, Graz, Austria. Kolsek, M., 2004: Cable yarding in Slovenian State Forests. In: Proceedings of Conference Cable yarding suitable for sustainable forest management, September 23, Idrija, Slovenia. Košir, B., 2000: Cable crane skidding courses for professional workers and private forest owners in Slovenia. In: Proceedingsof Conference Working Under a Dynamic Framework – Forest Ownership Structures and Extension, October 4–8 1999, Bled, Slovenia. Košir, B., 2001: Optimal line lengths when skidding wood with the Syncrofalke cable crane in Slovenian conditions. In: Proceedings of Workshop New trends in wood harvesting with cable systems for sustainable forest management in the mountains, June 18–24, Ossiach, Austria. Krogstad, F., Schiess, P., 2004: The allure and pitfalls of using LiDAR topography in harvest and road design. In: Proceedings of Joint Conference of IUFRO 3.06 Forest Operations under Mountainous Conditions and 12nd International Mountain Logging Conference, June 13–16, Vancouver, Canada. Krpan, A., Poršinsky, T., Šusnjar, M., 2001: Timber extraction technologies in Croatian mountainous selection forests. In: Proceedings of Workshop New trends in wood harvesting with cable systems for sustainable forest management in the mountains, June 18–24, Ossiach, Austria. Kurosaki, H., Iwaoka, M., Minematsu, H., 2008: Intervals to determine extra yarding distance of small scale cable yarding systems. In: Proceedings of IUFRO All-D3 Conference, June 15–20, Sapporo, Japan. Laffan, M., Jordan, G., Duhig, N., 2001: Impacts on soils from cable-logging steep slopes in northeastern Tasmania, Australia. Forest ecology and management 144(1–3): 91–99.

349

R. Cavalli

Prospects of Research on Cable Logging in Forest Engineering Community (339–356)

Largo, S., Han, H. S., Johnson, L., 2004: Productivity and Cost Evaluation for Non-guyline Yarders in Northern Idaho. In: Proceedings of COFE Conference, April 27–30, Clarion Resort on the Lake, Hot Springs, Arkansas.

terrain. In: Proceedings of Decision support for multiple purpose forestry. A transdisciplinary conference on the development and application of decision support tools for forest management, April 23–25, Vienna, Austria.

LeDoux, C. B., Gopalakrishnan, B., Pabba, R. S., 2002: An expert system for estimating production rates and costs for hardwood group-selection harvests. Proceedings 13th Central Hardwood Forest Conference, April 1–3, Urbana-Champaign, Illinois.

Loschek, J., 2001: Development of mechanized logging. In: Proceedings of Workshop New trends in wood harvesting with cable systems for sustainable forest management in the mountains, June 18–24, Ossiach, Austria.

LeDoux, C. B., Huyler, N. K., 2000: Cost comparisons for three harvesting systems operating in northern hardwood stands. Res. Pap. NE-715, U.S. Department of Agriculture, Forest Service, Northeastern Research Station, pp 4. LeDoux, C. B., Wilkerson, E., 2008: Assessing the ecological benefits and opportunity costs of alternative stream management zone widths for eastern hardwoods. Proceedings 2007 National Silviculture Workshop, May 7–10, Ketchikan, Arkansas. Leitner, T., Visser, R., Stampfer, K., 2010: Efficiency and ergonomic benefits of using radio controlled chokers in cable yarding. Croatian Journal of Forest Engineering 31(1): 1–10. Leonard, J. M., 2000: Combining Cut-to-Length and Cable Yarding Operations. In: Proceedingsof International Mountain Logging and 10th Pacific Northwest Skyline Symposium, March 28 – April 1, 1999 Corvallis, Oregon. Leonard, J. M., Garland, J. J., Pilkerton, S. J., 2003: Evaluation of synthetic rope for static rigging application in cable logging. In: Proceedings of COFE Conference, September 7–10, Atlantic Oakes by the Sea Resort, Bar Harbor, Maine. Lewis, J., Keppeler, E. T., 2004: Trends in streamflow and suspended sediment after logging, North Fork Caspar Creek. In: Proceedings of The Redwood Region Forest Science Symposium: What does the future hold?, March 15–17, Albany, California. Lexter, M. J., Stampfer, K., Vacik, H., Hochbichler, E., Limbeck-Lilienau, B., Hagauer, D., Durrstein, H., Spork, J., 2002: A computer based decision support tool for timber harvest planning under multiple objectives in steep terrain. In: Proceedings of Conference Mountain Forests: Conservation and Management, July 28 – August 1, Vernon, British Columbia. Li, Y., Wang, J., Miller, G.,McNeel, J. F., 2006: Production economics of harvesting small-diameter hardwood stands in central Appalachia. Forest Products Journal 56(3): 81–86. Li-hai, W., 2000: Environmentally sound timber extracting techniques for small tree harvesting. Journal of Forestry Research 11(4): 269–272.

Loschek, J., 2004: Cable yarding management in mountainous areas – ecological and economical aspects on silviculture. In: Proceedings of Conference Cable yarding suitable for sustainable forest management, September 23, Idrija, Slovenia. Lyons, C. K., 2007: A design criterion for guyed backspars. In: Proceedings of International Mountain Logging and 13th Pacific Northwest Skyline Symposium, April 1–6, Corvallis, Oregon. Lyons, C. K., 2008: Considering Cable Stretch in Logging Applications. International Journal of Forest Engineering 19(1): 29–35. Magaud, P., 2011: SIMULCABLE, a new software to optimise the line implantation for cable yarding. In: Proceedings of FORMEC Conference, October 9–13, Graz, Austria. Mahon, R., 2000: Experience in Cable Thinning. In: Proceedingsof International Mountain Logging and 10th Pacific Northwest Skyline Symposium, March 28 – April 1, 1999 Corvallis, Oregon. Manzone, M., Balsari, P., 2010: Assessing the performance of a cable crane using single cable. In: Proceedings of FORMEC Conference, July 11–14, Padova, Italy. Marenče, J., Vadjal, J., Košir, B., 2009: Bucking of conifers using the woody H60 processor on the Syncrofalke 3 T cable yarder in Alpine foothills. Zbornik gozdarstva in lesarstva (88): 3–10. Markewitz, D., 2006: Fossil fuel carbon emissions from silviculture: impacts on net carbon sequestration in forests. Forest ecology and management 236(2): 153–161. Masi, M., Nannicini, C., Amati, R., Fabiano, F., Neri, F., Piegai, F., Magagnotti, N., Spinelli, R., Laurendi, V., Pirozzi, M., 2010: Safety and Prevention in Mechanised Forest Operations: a Tuscan Project. In: Proceedings of International Conference Ragusa SHWA 2010, September 15–18, Ragusa, Italy. McNeel, J. F., 2000: Modeling Production of Longline Yarding Operations in Coastal British Columbia. International Journal of Forest Engineering 11(1): 29–38.

Limbeck-Lilienau, B., 2003: Residual stand damage caused by mechanized harvesting systems. In: Proceedings of AUSTRO 2003 Conference, October, 5–9, Schlaegl, Austria.

Messingerova, V., Stanovsky, M., Ferencik, M., Kovacik, P., 2009: Technological planning in cableway terrains in Slovakia. In: Proceedings of FORMEC Conference, June 21–24, Kostelec nad Cernymi lesy, Czech Republic.

Limbeck-Lilienau, B., Lexter, M. J., Stampfer, K., Vacik, H., Palmetzhofer, D., Hagauer, D., Hochbichler, E., Durrstein, H., Spork, J., 2003: CONES – A computer based decision support tool to explore stand management options in steep

Murphy, G., Brownlie, R., Kimberley, M., Beets, P., 2009: Impacts of forest harvesting related soil disturbance on end-ofrotation wood quality and quantity in a New Zealand radiata pine forest. Silva Fennica 43(1): 147–160.

350

Croat. j. for. eng. 33(2012)2

Prospects of Research on Cable Logging in Forest Engineering Community (339–356)

R. Cavalli

Murphy, G., Firth, J. G., Skinner, M. F., 2004: Long-term impacts of forest harvesting related soil disturbance on log product yields and economic potential in a New Zealand forest. Silva Fennica 38(3): 279–289.

Olund, D., 2001: The Future of Cable Logging. In: Proceedings of The International Mountain Logging and 11th Northwest Pacific Skyline Symposium, December 10–12, Seattle, Washington.

Murphy, G., Vanderburg, M., 2007: Modelling the economics of extended shift and 24/7 forest harvesting. New Zealand Journal of Forestry 52(2): 15.

Ottaviani Aalmo, G., Kyllo, N. O., Talbot, B., 2011: The Owren mini 400: a unique 1: 3 scale electrically powered tower yarder for research, training and demonstration. In: Proceedings of FORMEC Conference, October 9–13, Graz, Austria.

Nemestóthy, N., 2010: Synthetic fibre ropes for forestry use – criteria for the replacement of fibre ropes. In: Proceedings of FORMEC Conference, July 11–14, Padova, Italy. Nemestóthy, N., 2011: Synthetic fibre ropes for forestry use – further developments in finding criteria for the replacement state of fibre ropes. In: Proceedings of FORMEC Conference, October 9–13, Graz, Austria. Neri, F., Piegai, F., Marchi, E., Nati, C., 2009: Cableway logging operations and residual harvesting: cases study in windthrow areas in the Eastern Alps – Italy. In: Proceedings of FORMEC Conference, June 2–5, Schmallenberg, Germany. Newton, M., Cole, E. C., 2006: Harvesting impacts on understory regeneration in two-storied Douglas-fir stands. Western Journal of Applied Forestry 21(1): 14–18. Nitami, T., Suk, S., Kataoka, A., Mitsuyama, T., 2011: Tower yarder operation in Japan and the performance analysis by GPS-based system. In: Proceedings of FORMEC Conference, October 9–13, Graz, Austria. Nitteberg, M. A., 2003: Mechanised harvesting in cable operation. In: Proceedingsof 2nd Forest Engineering Conference, May 12–15, Växjö, Sweden. Nitteberg, M. A., 2007: From Off-road to On-road harvesting in steep terrain in Norway. In: Proceedings of International Mountain Logging and 13th Pacific Northwest Skyline Symposium, April 1–6, Corvallis, Oregon. Nitteberg, M. A., Stampfer, K., Kindernay, D., 2010: Continuous work strain measurement in cable yarder rigging. In: Proceedings of 2010 Nordic Baltic Conference on Forest Operations, October 20–22, Honne, Norway. Nitteberg, M. A., Talbot, B., 2011: Developing a new yardercontrolled mechanical slack-pulling carriage for doubledrum winches. In: Proceedings of FORMEC Conference, October 9–13, Graz, Austria. Oberholzer, F., Howe, D., 2007: The use of monocables in the harvesting of small timber: a South African perspective. In: Proceedings of International Mountain Logging and 13th Pacific Northwest Skyline Symposium, April 1–6, Corvallis, Oregon. Oka, M., Inoue, M., Kobayashi, H., 2006: Studies of the pattern classification of the logging systems for thinning on the basis of the applicability of forestry machines and features of logging system. Journal of the Japan Forest Engineering Society 21(2): 115–124. Okonski, J., 2004: Skyline Systems in Appalachia – 101 Tribulations. In: Proceedings of COFE Conference, April 27–30, Clarion Resort on the Lake, Hot Springs, Arkansas. Croat. j. for. eng. 33(2012)2

Ottaviani Aalmo, G., Nitteberg, M. A., Stampfer, K., Talbot, B., 2011. Workload benefits of using synthetic rope strawline in cable yarder rigging in Norway. Croatian Journal of Forest Engineering 32(2): 561–569. Ottaviani Aalmo, G., Nitteberg, M. A., Stampfer, K., Talbot, B., 2011: Workload benefits of using synthetic ropes in cable yarder rigging in Norway. In: Proceedings of FORMEC Conference, October 9–13, Graz, Austria. Owende, P. M., Tiernan, D., Ward, S. M., Lyons, J. K., 2001: Is there a role for cable extraction on low gradient sensitive sites? In: Proceedings of Workshop New trends in wood harvesting with cable systems for sustainable forest management in the mountains, June 18–24, Ossiach, Austria. Ozturk, M., Aykut, T., Acar, H. H., 2001: Time analyses on Koller K300 cable system on difficult terrain in Turkey. In: Proceedings of Workshop New trends in wood harvesting with cable systems for sustainable forest management in the mountains, June 18–24, Ossiach, Austria. Ozturk, M., Senturk, N., Hulusi, H. H., 2007: Forest skylines in Turkey. In: Proceedings of International Mountain Logging and 13th Pacific Northwest Skyline Symposium, April 1–6, Corvallis, Oregon. Ozturk, T., 2009: Productivity of timber extraction by Urus MIII cable crane from selective spruce forests. African Journal of Agricultural Research 4(9): 852–858. Ozturk, T., Demir, M., 2007: Transporting of spruce timbers by URUS MIII cable system from selective forests of Artvin region. Building and environment 42(3): 1278–1282. Ozturk, T., Senturk, N., 2006. Extraction of spruce timber by Gantner cable crane from selective forests of Artvin region. Croatian Journal of Forest Engineering 27(1): 59–66. Park, S-J., Kim, J-W., Park, M-S., Song, T-Y., Cho, K-H., 2006: An analysis of the yarding operation system with a mobile tower-yarder in Korea. In: Proceedings of COFE Conference, July 30 – August 2, Coeur d’Alene Resort, Coeur d’Alene, Idaho. Park, S-J., Son, B-S., Lee, S-H., Han, S-Y., Kim, D-G., 2008: Analysis of a mobile tower-yarder operation in Korea. In: Proceedings of IUFRO All-D3 Conference, June 15–20, Sapporo, Japan. Pertlik, E.,2009: Do Synthetic Ropes change the design principles of standing skylines? In: Proceedings of COFE Conference,June 15–18, Kings Beach (Lake Tahoe), California. Phillips, E. J., 2001: Alternative harvesting for visually sensitive areas. Advantage 2(42).

351

R. Cavalli

Prospects of Research on Cable Logging in Forest Engineering Community (339–356)

Pičman, D., Pentek, T., Poršinsky, T., 2001: Relation between forest roads and extraction machines in sustainable forest management. In: Proceedings of Workshop New trends in wood harvesting with cable systems for sustainable forest management in the mountains, June 18–24, Ossiach, Austria. Pilkerton, S. J., Garland,J. J.,Hartter, J., 2004: Applications of synthetic rope for improved ergonomic, economic and environmental performance in mountainous logging. In: Proceedings of 2004 International Mountain Logging Conference, June 14, Vancouver, British Columbia. Pilkerton, S. J., Garland, J. J., Leonard, J. M., Sessions, J., 2004: Synthetic rope use in logging winching applications. In: Proceedingsof 2004 International Mountain Logging Conference, June 14, Vancouver, British Columbia. Pilkerton, S. J., Garland,J. J., Sessions,J., Stringham, B., 2001: Prospects for using synthetic rope in logging: First look and future research. In: Proceedings of The International Mountain Logging and 11th Northwest Pacific Skyline Symposium, December 10–12, Seattle, Washington. Pollini, C., Schmiedhofer, J., 2001: The use of cable cranes in South Tyrol in the light of recent legislation. In: Proceedings of Workshop New trends in wood harvesting with cable systems for sustainable forest management in the mountains, June 18–24, Ossiach, Austria. Pyles, M. R., Lyons, K. C., 2001 : Analysis of unguyed spartrees. International Journal of Forest Engineering 12(2): 11– 17. Rahim, A. R., Shahwahid, M. S., Zariyawati, M. A., 2009: A comparison analysis of logging cost between conventional and reduce impact logging practices. International Journal of Economics and Management 32(2): 354–366. Rashin, E. B., Clishe, C. J., Loch, A. T., Bell, J. M., 2006: Effectiveness of timber harvest practices for controlling sediment related water quality impacts. JAWRA Journal of the American Water Resources Association 42(5): 1307–1327. Rieger, G., 2001: Costs and performance of a Koller K300 yarder. In: Proceedings of Workshop New trends in wood harvesting with cable systems for sustainable forest management in the mountains, June 18–24, Ossiach, Austria. Rivera, S., Kershner, J. L., Dobrowolski, J. P., 2010: Evaluation of the surface erosion from different timber yarding methods in Honduras. Revista Arvore 34(4): 577–586. Robek, R., Medved, M., 2001: Implementation of cable logging requirements in environmentally sound road construction. In: Proceedings of Workshop New trends in wood harvesting with cable systems for sustainable forest management in the mountains, June 18–24, Ossiach, Austria.

In: Proceedings of Workshop New trends in wood harvesting with cable systems for sustainable forest management in the mountains, June 18–24, Ossiach, Austria. Saidi-Mehrabad, M., 2002. Application of Markov Processes to the machine delays analysis. International Journal of Engineering, Transactions B: Applications 15: 63–72. Saravi, A. A., Lyons, K. C., 2004: Finite element modeling of guyed backspars in cable logging. Canadian Journal of Forest Research 34 (4):817–828. Saunders, J. C., 2001: Current trends in cableway systems in the United Kingdom. In: Proceedings of Workshop New trends in wood harvesting with cable systems for sustainable forest management in the mountains, June 18–24, Ossiach, Austria. Schiess, P., 2001: Road management strategies to reduce habitat impacts – A case for engineered cable yarding operations and harvest schedules. In: Proceedings of Workshop New trends in wood harvesting with cable systems for sustainable forest management in the mountains, June 18–24, Ossiach, Austria. Schiess, P., 2004: Forest transportation issues in relation to stream habitat conservation: trade offs between environmental and economic concerns. In: Proceedings of Conference Cable yarding suitable for sustainable forest management, September 23, Idrija, Slovenia. Schiess, P., Krogstad, F., 2003: LiDAR-based topographic maps improve agreement between office-designed and fieldverified road locations. In: Proceedings of COFE Conference, September 7–10, Bar Harbor, Maine. Schmimer, K. R., Nemestóthy, N., Sperrer, S., 2009: Further developments of synthetic ropes for logging applications in forestry. In: Proceedings of FORMEC Conference, June 2–5, Schmallenberg, Germany. Senturk, N., Ozturk, T., Demir, M., 2007: Productivity and costs in the course of timber transportation with the Koller K300 cable system in Turkey. Building and environment 42(5): 2107–2113. Sessions, J., Chung, W., Heinimann, H. R., 2001: New algorithms for solving large-scale transportation planning problems. In: Proceedings of Workshop New trends in wood harvesting with cable systems for sustainable forest management in the mountains, June 18–24, Ossiach, Austria. Sloan, H., 2001: Appalachian hardwood logging systems: managing change for effective BMP implementation. In: Proceedings of COFE Conference, July 15–19, Snowshoe, West Virginia.

Roberts, B., 2004: A comparison of landslide rates following helicopter and conventional cable-based clear-cut logging operations in the Southwest Coast Mountains of British Columbia. Geomorphology 61(3–4): 337–347.

Smeets, P., 2007: The use of high performance synthetic fibers in synthetic ropes for logging applications in the forestry. In: Proceedings of FORMEC Conference, October 7–11, Vienna and Heiligenkreutz, Austria.

Sabadyr, A., Sibtsev, S., 2001: Characteristics of the cable yarding system application in the Ukrainian Carpathians.

Smidt, M., 2011: A description of forest industries and occupations with focus on forestry workers’ jobs and injury

352

Croat. j. for. eng. 33(2012)2

Prospects of Research on Cable Logging in Forest Engineering Community (339–356) and illness surveillance. In: Proceedings of COFE Conference, June 12–15, Quebec City, Canada. Solmie, D. K., Kellogg, L. D., Kiser, J. D., Wing, M. G., 2003: Comparing strategies for skyline corridor layout. In: Proceedings of COFE Conference, September 7–10, Bar Harbor, Maine. Sonne, E., 2006: Greenhouse gas emissions from forestry operations: a life cycle assessment. Journal of environmental quality 35(4): 1439–1450. Spinelli, R., Magagnotti, N., Lombardini, C., 2010: Performance, capability and costs of small-scale cable yarding technology. Small-Scale Forestry 9(1): 123–135. Spinelli, R., Magagnotti, N., Nati, C., 2010: Benchmarking the impact of traditional small-scale logging systems used in Mediterranean forestry. Forest ecology and management 260(11): 1997–2001. Spinelli, R., Visser, R., 2008: Analyzing and estimating delays in harvester operations. International Journal of Forest Engineering 19(1): 34–61. Stampfer, K., 2002: Combining harvester with cable yarder on steep terrain thinning operations. In: Proceedings of International seminar on new roles of plantation forestry requiring appropriate tending and harvesting operations, September 29 – October 5, Tokio, Japan. Stampfer, K., 2004: Perspectives on whole tree cable yarding systems for thinning operation in Austria. In: Proceedings of Conference Cable yarding suitable for sustainable forest management, September 23, Idrija, Slovenia. Stampfer, K., Karpf, F., Visser, R., 2010: Efficiency and ergonomic advantages of synthetic rope for guying cable yarders. In: Proceedings of 2010 Nordic Baltic Conference on Forest Operations, October 20–22, Honne, Norway. Stampfer, K., Leitner, T., Visser, R., 2009: Efficiency and ergonomic benefits of using radio controlled chokers in cable yarding. In: Proceedings of FORMEC Conference, June 21– 24, Kostelec nad Cernymi lesy, Czech Republic. Stampfer, K., Leitner, T., Visser, R., 2010: Efficiency and ergonomic benefits of using radio controlled chockers in cable yarding. Croatian Journal of Forest Engineering 31(1): 1–9. Stampfer, K., Lexter, M. J., 2001: Multicriteria evaluation of thinning operations on steep terrain. In: Proceedings of Workshop New trends in wood harvesting with cable systems for sustainable forest management in the mountains, June 18–24, Ossiach, Austria.

R. Cavalli

Streif, A., 2001: View of the contractor for forest companies. In: Proceedings of Workshop New trends in wood harvesting with cable systems for sustainable forest management in the mountains, June 18–24, Ossiach, Austria. Sullman, M. J. M., Kirk, P. M., 2001: Harvesting wind damaged trees: a study of the safety implications for fallers and choker setters. International Journal of Forest Engineering 12(2): 67–77. Suzuki, Y., 2000: Damage to residual stands from thinning with short-span tower yarders: re-examination of wounds after five years. Journal of Forest Research 5(3): 201–204. Suzuki, Y., Shiobara, M., Kondo, M., 2000: Plane-Type Logging Cable Systems in Japan – Past, Present and Future. In: Proceedings of International Mountain Logging and 10th Pacific Northwest Skyline Symposium, March 28 – April 1, Corvallis, Oregon. Thompson, M., Stander, H., John, S., 2008: GuylinePC: An Interactive Guyline Tension Analysis Program for Guyed Logging Towers. Western Journal of Applied Forestry 23(3): 133–141. Torgersen, H., 2000: Endurance of compacted steel wire ropes. International Journal of Forest Engineering 11(2): 43–49. Torgersen, H., 2002: Service life of compacted steel wire ropes on varying block sheave diameters. International Journal of Forest Engineering 13(1): 51–56. Torgersen, H., Lisland. T., 2002: Excavator-based cable logging and processing system: a norwegian case study. International Journal of Forest Engineering 13(1): 11–16. Tsioras, P. A., Rottensteiner, C., Stampfer, K., 2011: Analysis of accidents during cable yarding operations in Austria 1998–2008. Croatian Journal of Forest Engineering 32(2): 549–560. Tucek, J., Pacola, E., 2001: Skyline yarding distance modelling for logging in mountains. In: Proceedings of Workshop New trends in wood harvesting with cable systems for sustainable forest management in the mountains, June 18–24, Ossiach, Austria. Tunay, M., Melemez, K., 2001: Work performance of Koller K300 cable system on difficult terrain in Turkey. In: Proceedings of Workshop New trends in wood harvesting with cable systems for sustainable forest management in the mountains, June 18–24, Ossiach, Austria. Unver, S., Acar, H. H., 2009: Evaluation of residual tree damage in sloping areas due to harvesting operations by manually. Austrian Journal of Forest Science 126(3): 119–132.

Stampfer, K., Lexter, M. J., Vacik, H., Hochbichler, E., Durrstein, H., Spork, J. 2001: CONES – A computer based multiple criteria decision support tool for timber harvest planning in steep terrain. In: Proceedings of COFE Conference, July 15–19, Snowshoe, West Virginia.

Unver, S., Ozkaya, M. S., Kilic, H., Acar,H. H., 2011: Determination of efficiency of the forest skylines in Artvin Forest Region of Turkey. In: Proceedings of FORMEC Conference, October 9–13, Graz, Austria.

Stampfer, K., Visser, R., Kanzian, C. H., 2006: Cable corridor installation times for European yarders. International Journal of Forest Engineering 17(2): 71–77.

Vacik, H., Lexter, M. J., Palmetzhofer, D., Stampfer, K., Limbeck-Lilienau, B., 2004: Application of the spatial decision support system CONES for regeneration planning in moun-

Croat. j. for. eng. 33(2012)2

353

R. Cavalli

Prospects of Research on Cable Logging in Forest Engineering Community (339–356)

tain forests. In: Proceedings of Joint Conference of IUFRO 3.06 Forest Operations under Mountainous Conditions and 12nd International Mountain Logging Conference, June 13– 16, Vancouver, Canada. Vacik, H., Lexter, M. J., Palmetzhofer, D., Stampfer, K.,Limbeck-Lilienau, B., 2004: Application of the spatial decision support system CONES for regeneration planning in mountain forests. In: Proceedingsof Conference Cable yarding suitable for sustainable forest management, September 23, Idrija, Slovenia. Valente, C., Spinelli, R., Hillring, B. G., 2011: LCA of environmental and socio-economic impacts related to wood energy production in alpine conditions: Valle di Fiemme (Italy). Journal of Cleaner Production19(17–18): 1931–1938. Vana, T. T., Giambelluca, T. W., Sutherland, R. A., Senock, R. S., 2001: Impacts of Forest Harvesting on Soil Hydraulic Characteristics at Hakalau, Hawaii. In: Proceedings of American Geophysical Union Fall Meeting, December 10–14, San Francisco, California. Vasilev, V., Rossnev, B., Kitanova, S., Alexandrov, A., Raev, I., Tsakov, H., Dimitrov, V., Grozeva, M., Petrova, R., Popov, G., 2003: Possibilities for the of use the light cable crane Pioneer for removal of faggot-wood at gravitation regime of work. In: Proceedings of International scientific conference marking 75 years of the Forest Research Institute of the Bulgarian Academy of Sciences, October 110–14 5, Sofia, Bulgaria. Visser, R., 2010: Landings size and characteristics. Future Forests Research. University of Canterbury, School of Forestry, pp 8. Visser, R., Adams, J., 2002: Risk management of steep terrain harvesting. In: Proceedings of COFE Conference, June 16–20, Auburn, Alabama. Visser, R., Baker, S., Sloan, H., 2001: Cable logging opportunities in the Appalachian mountains. In: Proceedings of Workshop New trends in wood harvesting with cable systems for sustainable forest management in the mountains, June 18–24, Ossiach, Austria. Visser, R., Spinelli, R., Magagnotti, N., 2010: Landing size and landing layout in whole-tree harvesting operations in New Zealand. In: Proceedings of FORMEC Conference, July 11–14, Padova, Italy. Wing, M. G., Kellogg, L. D., 2001: Using a Laser Range Finder to assist harvest planning. In: Proceedings of First Inter-

354

national Precision Forestry Cooperative Symposium, June 17–20, Seattle, Washington. Worrell, W. C., Bolding, M. C., Aust, W. M., 2010: Comparison of potential soil erosion following skyline yarding versus tracked skidding on bladed skid trails in the Appalachian Region of Virginia. In:Proceedings of COFE Conference, June 6–9, Auburn, Alabama. Worrell, W. C., Bolding, M. C., Aust, W. M., 2011: Potential soil erosion following skyline yarding versus tracked skidding on bladed skid trails in the Appalachian Region of Virginia. Southern Journal of Applied Forestry 35(3): 131–135. Yoshimura, T., Hartsough, B. R., 2007: Conceptual evaluation of harvesting systems for fuel reduction and biomass collection on steep terrain using system dynamics. In: Proceedings of International Mountain Logging and 13th Pacific Northwest Skyline Symposium, April 1–6, Corvallis, Oregon. Yoshimura, T., Hartsough, B. R., 2010: Simulation-based validation of new conceptual cable harvesting systems. In: Proceedings of FORMEC Conference, July 11–14, Padova, Italy. Yoshimura, T., Hartsough, B. R., 2010: Validation of a new conceptual cable harvesting system using an independent device for lateral yarding. In: Proceedings of COFE Conference, June 6–9, Auburn, Alabama. Yoshimura, T., Otsu, H., Hartsough, B. R., 2011: Developing and validating the automatic log transfer mechanism between two carriage. In: Proceedings of FORMEC Conference, October 9–13, Graz, Austria. Zambelli, P., Lora, C., Ciolli, M., Spinelli, R., Tattoni, C., Vitti, A., Zatelli, P., 2010: A FOSS4G model to estimate forest extraction methods and biomass availability for renewable energy production. In: Proceedings of FOSS4G Conference, September 6–9, Barcelona, Spain. Zimbalatti, G., Proto, A. R. 2007: Cable Logging opportunities for firewood in Calabrian forestry. In: Proceedings of FORMEC Conference, October 7–11, Vienna and Heiligenkreutz, Austria. Zimbalatti, G., Proto, A. R., 2009: Cable logging opportunities for firewood in Calabrian forests. Biosystems Engineering 102(1): 63–68. Zimbalatti, G., Proto, A. R., 2010: Timber extraction with a cable crane in South Italy (Calabria). In: Proceedings of FORMEC Conference, July 11–14, Padova, Italy.

Croat. j. for. eng. 33(2012)2

Prospects of Research on Cable Logging in Forest Engineering Community (339–356)

R. Cavalli

Sažetak

Perspektiva istraživanja iznošenja drva žičarama u šumarskoj inženjerskoj zajednici Namjera je analize provedenih istraživanja iznošenja drva žičarama u posljednjih dvanaest godina (2000–2011), zasnovane na znanstvenim međunarodnim objavama, ocjena glavnih istraživanih tema te procjena razvoja istraživanja vezanih uz šumske žičare u budućnosti. Međunarodna znanstvena literatura prikupljena je iz baza podataka znanst venih časopisa, odnosno zbornika radova iz šumarskoga inženjerstva. Pretražene su ove baze podataka: 1) Google Scholar, 2) Science Direct, 3) CAB Abstracts, 5) Current Contents, 6) Ingenta Connect, 7) Forest Science Database, 8) AGRIS International Information System for the Agricultural Sciences and Technology, 9) IUFRO On-line Li brary, 10) USDA National Forest Service Library, 11) USDA Treesearch Forest Service Research and Development. U pretraživanju smo se služili ključnim riječima: iznošenje drva žičarom (»cable logging«, »cable yarding«), žičara (»cable crane«, »cable yarder«), stupna žičara (»tower yarder«). Pretraživanje se odnosilo isključivo na radove na pisane engleskim jezikom. Dodatno je pretraživanje obuhvatilo dostupne objave s mrežnih stranica znanstvenih časopisa i međunarodnih znanstvenih savjetovanja iz šumarskoga inženjerstva. Ukupno su 243 objave uvrštene te korištene za stvaranje zbirke u računalnom programu Thompson Reuter EndNote®. Svakoj su objavi pridružene jedna ili dvije ključne riječi radi daljnje razradbe sadržaja zbirke. U istraživanju su se ključne riječi smatrale deskriptorima pomoću kojih se objave mogu razvrstavati tijekom analiza prema potrebi. S toga su gledišta ključne riječi razvrstane u dvije glavne teme: »sustav žičara« i »iznošenje drva žičarama«. Prva tema uključuje sve deskriptore koji iznošenje drva žičarom opisuju kao sustav: »djelotvornost žičnih sustava«, »pro jektiranje žičnih sustava«, »planiranje žičnih sustava«, »simulacija žičnih sustava«, »mehanika žičnih sustava«. Za analizu »mehanike žičnih sustava« dodana su četiri deskriptora druge razine koji opisuju glavne sastavnice žičnoga sustava s mehaničke točke gledišta: »žičara«, »kolica«, »uže«, »potporanj«. Druga tema uključuje deskriptore koji se odnose na: 1) utjecaj načina (metode) iznošenja drva na okoliš (»utjecaj iznošenja drva žičarama«), 2) rukovatelje žičarama (»ergonomija i sigurnost u radu pri iznošenju drva žičarom« i »izobrazba i obuka za iznošenje drva žičarom«), 3) upravljanje (»upravljanje iznošenjem drva žičarom«). Radi lakšega razlučivanja deskriptoru »utjecaj iznošenja drva žičarama« dodana su četiri deskriptora druge razine, od kojih svaki opisuje jedan od glavnih okolišnih resursa uključenih u analizu utjecaja iznošenja drva žičarom: »vegetacija«, »tlo«, »voda« i »zrak«. Objave uvrštene u istraživanje nisu ravnomjerno raspoređene u promatranom razdoblju (2000–2011); prosječno su bile 20,3 objave godišnje, s izuzetkom 2001. godine kada ih je bilo mnogo više (50) zbog održane konferencije posvećene organizaciji pridobivanja drva žičarama. Potrebno je izdvojiti i 2007. i 2011. godinu ne zbog broja objava, već zbog činjenice da većina objava potječe s dviju konferencija održanih tih godina. Glavni su izvori znanstvenih objava bili zbornici radova, zatim članci u časopisu te izvještaji. Objave u obliku knjiga ili poglavlja iz knjiga gotovo su zanemarive. Primjenom deskriptora u analizi objava zaključeno je da su u razdoblju od 2000. do 2011. godine šumarski inženjeri zainteresirani za iznošenje drva žičarama najviše istraživali djelotvornost žičnih sustava te zatim utjecaj iznošenja drva žičarom na različite sastavnice okoliša. Simulaciji je žičnih sustava također pridana važnost, pretežno u njihovim različitim varijantama izvedbe koje su razvili šumarski inženjeri. Iako broj objava pokazuje da postoji zanimanje za mehaniku žičnih sustava, većina se znanstvenika bavi istraživanjem užadi, posebice sintetičke. To se može objasniti činjenicom da se, zahvaljujući mogućnosti da zamijeni čeličnu užad prilikom određenih radova pri dobivanja drva, sintetička užad u šumarstvu počela upotrebljavati krajem prošloga stoljeća te zbog toga zahtijeva opsežne analize i procjene. Kolica, žičare i potpornji rjeđe su bili predmetom istraživanja. Što se tiče utjecaja iznošenja drva žičarom na okoliš, objave su se uglavnom bavile utjecajem na vegetaciju i tlo, a samo nekoliko njih utjecajem na vodu i zrak. Treba ipak priznati da je teško razgraničiti utjecaj na tlo i na vodu od ostaloga štetnoga utjecaja jer oštećenje (gaženje) tla uzrokuje nastanak i taloženje sedimenta u vodotocima te posljedično narušava kakvoću vode. Ergonomiji i sigurnosti pri iznošenju drva žičarama posljednjih se godina pridaje sve veća pažnja, dok za izobrazbu i obuku ne postoji značajan interes. Na temelju tih razmatranja predviđa se sljedeća budućnost istraživanja iznošenja drva žičarama: Þ Razvoj novih materijala za užad (i sintetičku i čeličnu) radi povećanja čvrstoće i nosivosti užadi, Þ Razvoj upravljanja mehaničkim sklopovima na kolicima i žičarama, s povećanjem radne djelotvornosti i er gonomije te poboljšanjem upravljanja strojevima, Croat. j. for. eng. 33(2012)2

355

R. Cavalli

Prospects of Research on Cable Logging in Forest Engineering Community (339–356)

Þ Povećanje izvedivosti računalnih metoda planiranja iznošenja drva žičarom, Þ Poboljšanje matematičkih metoda za optimizaciju strukturnih analiza žičnih sustava, Þ Optimizacija potrošnje energije, smanjenje uporabe energije iz fosilnih goriva na najmanju moguću mjeru, povećanje iskoristivosti električne energije i energije gravitacije, Þ Istraživanje tehnologije privlačenja drva po tlu kretnim vozilima potpomognutim žičarom koja predstavljaju prijelaz između zračnih i po tlu kretnih sustava, Þ Povećanje uporabe alata poput Analize životnoga ciklusa (LCA) za procjenu štetnoga djelovanja žičnih tehnologija na okoliš, Þ Poboljšanje izobrazbe i obuke za radove na iznošenju drva žičarom razvojem edukacijskih metodologija i alata za učenje koji uzimaju u obzir važnost učenika u čitavom procesu učenja. Ključne riječi: šumarsko inženjerstvo, iznošenje drva žičarom, znanstvena literatura

Author’s address – Autorova adresa:

Received (Primljeno): August 2, 2012 Accepted (Prihvaćeno): September 14, 2012

356

Prof. Raffaele Cavalli, PhD. e-mail: raffaele.cavalli@unipd.it Forest Operation Management Unit Department Land, Environment, Agriculture and Forestry University of Padova Viale dell’Università 16 35020 Legnaro ITALY Croat. j. for. eng. 33(2012)2

Subject review – Pregledni rad

Life Cycle Assessment (LCA) in Forestry – State and Perspectives Hans Rudolf Heinimann Abstract – Nacrtak Environmentally sound technologies are a key to reduce resource use and environmental impact. The paper reviews the state of knowledge of an analysis tool, life cycle assessment (LCA), by addressing three issues: 1) methodological foundations of LCA, 2) lifecycle inven tory modeling, and 3) environmental performance indicators for wood supply systems. The study results in the following findings: 1) LCA is still not widely used and accepted in the forest operations engineering. 2) Only a few studies are based on state-of-the-art life cycle inventory analysis. 3) The boundaries of the studied systems are often too narrow, limiting comparability with standard LCA studies. 4) Most forest-related studies are based on direct process energy input only and are neglecting environmental burdens of upstream processes. 5) »Truncated LCAs«, neglecting embodied burdens of road infrastructure and forest machines always result in an underestimation of environmental impacts or an overestimation of envi ronmental performance, respectively. There is a need for LCA capacity building in the forest operations community, on the basis of which forest-related LCA studies should become more comprehensive and comparable with studies of the core LCA community. Keywords: LCA, environmental performance, wood supply, eco-efficiency, industrial ecology

1. Introduction – Uvod There is a broad consensus that mankind has to explore and implement pathways of development that minimize resource use while reducing emissions, waste and impacts to structures and functions of the environmental system to »near zero«. Environmentally sound technologies (ESTs) were identified as a key to achieve this broad, long-term goal. ESTs encompass technologies that have the potential to significantly improve environmental performance, compared with other technologies (UNEP-IETC, online). Following the quote »what gets measured gets done«, which is attributed to Peter Drucker, there is a need to apply and improve environmental analysis tools to produce reliable, comprehensible environmental performance indicators. The development since the early 1990s led to a whole set of tools, such as 1) lifecycle assessment (LCA) for product systems (ISO 2006b), 2) regional material flow analysis for geographical regions (Hendriks et al. 2000), or 3) carbon budget models for large-scale geographical areas (Kurz et al. 2009). Those Croat. j. for. eng. 33(2012)2

tools, designed for different contexts, are somewhat embedded in the umbrella concept of »industrial ecology«, which looks at industrial systems the same way as ecologists have been looking at ecosystems (Erkman 1997). The key issue is to understand, model and manage the »industrial metabolism«, particularly the flow of materials and energy, to continuously increase environmental performance (for the history of the industrial metabolism thinking, see (Fischer-Kowalski 1998 a, b)). The ideas go back to pioneers such as Robert Ayres (Ayres and Kneese 1969); Charles Hall (Hall et al. 1979); and Howard T. Odum (Odum et al. 1977), whose thoughts stimulated Ulf Sundberg, a forest operations engineering scholar, to perform some preliminary energy analysis studies of forest operations systems (Sundberg and Svanqvist 1987). Forestry has been a traditional supplier of renewable raw materials for industrial use (sawmilling, pulp and paper, particle boards, etc.), for household fuel wood, particularly in the Third World, and increasingly for biofuels. From a production context point of

357

H. R. Heinimann

Life Cycle Assessment (LCA) in Forestry – State and Perspectives (357–372)

view, LCA is a suitable tool to assess wood supply systems, because it was designed for product systems (ISO 2006b). Recent efforts fostering the development and deployment of energy systems that are based on the renewable resources should follow eco-efficient pathways, whose performance has to be based on comprehensive assessment methods. The present paper aims to critically review environmental performance assessment from a LCA-perspective. The paper looks at wood-based raw material pathways only, and also neglects production impacts on forest soils. Assuming that LCA has not been widely applied and accepted within the forest operations engineering, the paper first reviews the methodological foundation of LCA, then looks at the state of lifecycle inventory modeling, and finally synthesizes the state of knowledge on environmental performance indicators for wood product, forest road and bioenergy product systems.

2. LCA Methodology – Metode analize životnoga ciklusa Although LCA was standardized at the beginning of the 1990s, the underlying methodology is not widely known and accepted within the forest science community. Therefore, it is useful to position LCA within the whole set of environmental tools, to explain the LCA framework, and to present the generic approach to model life cycle inventories in the following paragraphs.

2.1 Positioning of LCA within the set of environmental tools – Položaj analize životnoga ciklusa unutar grupe alata za procjenu utjecaja na okoliš Life cycle assessment (LCA) is a method to quantify and improve our understanding of possible impacts associated with products aiming 1) to identify

Table 1 Positioning of LCA within set of environmental assessment and policy tools. LCA addresses the level of product systems by modeling the life cycle of a product Tablica 1. Položaj analize životnoga ciklusa unutar grupe alata okolišne politike i alata analize utjecaja na okoliš. Analiza životnoga ciklusa odnosi se na razinu sustava proizvoda, određujući životni ciklus proizvoda Level of Action Razina djelovanja Policy Politika

Program Program

Plan

Definition – Definicija

Definirane smjernice aktivnosti za donošenje sadašnjih i budućih odluka, a rezultat su političkoga interesa

Područje aktivnosti usmjereno na politiku sektora i dodjelu financijskih sredstava

Sustainability Impact Assessment SIA

Localization and temporal definition how and with what priority public actions should be implemented

Procjena učinka na potrajnost

A set of activities that is 1) limited in time, 2) directed to create a clearly defined output, 3) considering financial constraints, and 4) fulfilling quality requirements

Product System Sustav proizvoda

358

Strategic Environmental Assessment SEA Strateška procjena okoliša

Project (public)

Privatni projekt

Alati analize učinka na okoliš

A portfolio of actions, directed to a sectorial policy and usually allocating financial resources

Lokaliziranje i privremeno definiranje aktivnosti koje će se i s kojim prioritetom provesti

Project (private)

Environmental Analysis Tools

Alati okolišne politike

A defined course of action, for guiding present and future decisions, as a result of a political weighting of interest

Plan

Javni projekt

Environmental Policy Tools

Skup aktivnosti koji je 1) ograničen vremenom, 2) ima jasan zadatak, 3) ograničena financijska sredstva i 4) zadovoljava uvjete kakvoće Collection of materially and energetically connected unit processes which perform one or more defined functions (ISO 14050) Skup materijalnih i energetskih postupaka kojima se izvršava točno zadana zadaća (ISO 14050)

Environmental Impact Assessment EIA Procjena učinka na okoliš Sustainability Impact Assessment SIA Procjena učinka na potrajnost Eco-Labeling Ekoobilježavanje Eco-Auditing Ekoispitivanje

Life Cycle Assessment LCA (E-LCA, S-LCA) Procjena životnoga ciklusa

Croat. j. for. eng. 33(2012)2

Life Cycle Assessment (LCA) in Forestry – State and Perspectives (357–372)

opportunities for environmental performance improvement, 2) to inform industrial decision-makers on the development of products and on the design or redesign of manufacturing processes, 3) to select and quantify environmental performance indicators, and 4) to prove environmental soundness for eco-labels or environmental claims (ISO 2006a). However, LCA is embedded in a whole set of environmental tools (Table 1), and there is still some confusion about the purpose and scope of different tools and about which tool is most appropriate to tackle a specific problem. The success of environmental policies, aiming to mitigate and limit exhaustive resource use, pollution and waste disposal, is only visible after actions affecting the environment have been implemented on the ground. A whole set of environmental tools was developed to foster a process of change, leading to environmentally more sound behavior. Table 1 presents an overview, organized along two dimensions. The first dimension represents the level of action, from policies to programs, projects and product systems, whereas the second dimension characterizes the type of tools: 1) analysis tools and 2) policy tools (Udo de Haes 1996a). Analysis tools address the quantification of eco-efficiency metrics, aiming at answering questions like »how does a package system perform compared to a new, alternative system«. Policy tools, on the other hand, are »instruments through which governments seek to influence citizen behavior and achieve policy purposes« (Schneider and Ingram 1990). Strategic environmental assessment is a policy tool to influence environmental soundness of policies, programs and plans, whereas environmental impact assessment aims at improving the environmental soundness of projects (Anonymous 2002, 2009). The purpose of impact assessment is to ensure that environmental considerations are explicitly incorporated into the development decision-making process (Anonymous 1999, 2002, 2009); it therefore aims to influence the behavior of mainly public decision-makers, e.g. the authorities responsible to approve project proposals. Sustainability impact assessment has its origins in the family of environmental assessment processes (SEA and EIA) and reflects the »triple bottom line« approach to sustainability by concurrently assessing environmental, economic and social impacts of proposed interventions (Pope et al. 2004). Eco-labels and eco-auditing are two policy tools addressing to change the values and perceptions of consumers and producers, respectively. Eco-labeling is a statement, symbol or graphic that indicates an environmental aspect of a product, a component or packaging, whereas ecoauditing is a systematic verification process to deterCroat. j. for. eng. 33(2012)2

H. R. Heinimann

mine where activities or management systems comply with normative standards (ISO 2002). Lifecycle assessment began to emarge at the end of the 1960s when the Midwest Research Institute conducted a study on resource requirements, emissions and waste of different beverage containers for the Coca-Cola Company (Guinée et al. 2010). Similar studies followed in the US and in Switzerland, investigating environmental burdens of containers made of PVC, glass, sheet metal and cardboard. However, there was a lack of a common theoretical framework and of methodological consistency (different names were in use, such as »eco-balance« or »resource and energy production analysis«), affecting the comparability of results and preventing LCA to become an accepted analytical tool (Guinée et al. 2010). By the end of the 1980s, SETAC, the society of environmental toxicology and chemistry, took responsibility for the theoretical foundation and the standardization of LCA, which resulted in a code of practice (Consoli et al. 1993). In Europe, the two guidelines developed by the center of environmental science in the Netherlands (Heijungs 1992a, b) had a considerable impact on the LCA practice because they presented a mathematical framework to handle even huge sets of interrelated processes. The mathematical formalism is based on former work of (Ayres and Noble 1978; Koopmans 1951a, b). In 1994, ISO – international standards organization – started to develop the ISO 14000 standards series (Marsmann 1997), addressing all aspects of lifecycle management (ISO 2006a), particularly LCA principles and framework (ISO 2006b), LCA requirements and guidelines (ISO 2006c), LCA vocabulary (ISO 2002), and environmental performance evaluation (ISO 1999). In 2002, SETAC together with UNEP – United Nations environment program – started the so-called lifecycle initiative (Guinée et al. 2010), aiming to foster lifecycle thinking and to integrate the »triple bottom line« (economic, social, and environmental) philosophy for goods and services. Lifecycle assessment, now called environmental lifecycle assessment – E-LCA has mainly been focusing on impacts on the natural environment, whereas lifecycle costing – LCC addresses the direct costs and benefits for people, planet and prosperity (UNEP and SETAC 2009). A tool to assess the impact of a product system on human well-being and corporate social responsibility was missing, and the lifecycle initiative resulted in framing and conceptualizing a new tool, social lifecycle assessment – SLCA to fill this gap (UNEP and SETAC 2009). This broader, more holistic approach covers all aspects of the triple bottom line – people, planet, and prosperity

359

H. R. Heinimann

Life Cycle Assessment (LCA) in Forestry – State and Perspectives (357–372)

Table 2 Definition and positioning of LCA phases as defined by the ISO standard Tablica 2. Određivanje i pozicioniranje faza analize životnoga ciklusa, opisanih u normama ISO Phase – Etapa

Required knowledge

Definition – Opis

Potrebno znanje

Goal definition, e.g. environmental performance comparison of alternative systems Goal and scope definition Cilj i značaj definicije

Definicija cilja, npr. usporedba okolišne pogodnosti alternativnih sustava Scope definition, particularly determination of 1) system boundaries, 2) functional unit.

Systems theory Teorija sustava

Značaj definicije, poglavito određivanje 1) granica sustava, 2) funkcionalnih jedinica Inventory analysis – LCI Analiza zaliha – LCI Impact assessment – LCIA Procjena učinaka – LCIA Interpretation Tumačenje

Inventory modeling of input/output flows for specified product system(s)

Systems theory, process engineering

Modeliranje ulaznih i izlaznih tokova za određeni proizvodni sustav

Teorija sustava, tehnike postupaka

Impact assessment: assignment of LCI results to specific impact categories

Environmental science, Eco toxicology

Procjena utjecaja: zadatak utjecaja životnoga ciklusa na određene kategorije

Znanost o okolišu, Ekotoksikologija

Conclusions and recommendations for process improvement.

Critical thinking – Kritičko promišljanje

Zaključci i prijedlozi za poboljšanje postupka

Decision making – Odlučivanje

– and is also called lifecycle sustainability analysis – LCSA (Guinée et al. 2010). The standardization efforts triggered the introduction of LCA into forestry and forest industries. One of the fathers of LCA (Udo de Haes 1996b) presented the »bottlenecks« that primary production was facing, particularly the definition of the upstream system boundary and co-production, which require special allocation rules. At the same time a first conference was organized in Germany (Frühwald 1995), and the main LCA activities related to forest sector emerged in LCA pioneering countries (Guinée et al. 2010), such as the US, Sweden, Switzerland, Germany and Finland. The Nordic pulp and paper industry started an initiative to develop a joint methodology for lifecycle inventories for forest industry in 1993 (Kärnä and Ekvall 1997), resulting in the definition of parameters and units of measure and in a proposition of allocation rules. At the same time, the first LCA studies on forest operations and long-distance transportation of timber (Karjalainen and Asikainen 1996) and on eco-inventories of forest machines and processes (Berg 1995; Knechtle 1997; Zimmer and Wegener 1996) appeared, yielding operational performance indicators, particularly related to energy consumption and CO2 emission. Harmonization efforts continued with the European COST-action »lifecycle assessment of forest and forest products«, which published the findings in 2001 (Karjalainen et al. 2001). Lifecycle inventories of forestry and forest industry processes started to be investigated systematically and entered into lifecycle-inven-

360

tory databases, such as ecoinvent (ECOINVENT, online), or ProBas (PROBAS, online). A new wave of LCA-type studies for forest operations appeared after 2005, e.g. (Valente et al. 2011a; Valente et al. 2011b), probably triggered by the increasing interest in renewable energy. However, the stream of research seems somewhat to be delinked from the LCA community.

2.2 LCA framework – Okosnica analize životnoga ciklusa The procedural LCA framework, consisting of 1) goal and scope definition, 2) inventory analysis, 3) impact assessment and 4) interpretation has been the foundation of LCA. Table 2 presents the four phases of LCA and the definition following the ISO 14,000 standards. There are two critical issues in scope definition, 1) the determination of system boundaries and 2) the definition of the functional unit. As we will show later, system boundaries are often not clear, particularly in the »upstream« direction. The environmental system has to be part of the analysis, characterized by input flows such as CO2, solar energy, mineral resources and land (occupied and transformed). Inventory analyses consists of mapping the structure and functions of the product system, usually in the form of a process flow diagram that is the basis for the following modeling of materials, energy, emission and waste flows. Inventory analysis is the heart of LCA, taking a considerable amount of time and being extremely data intensive. Lifecycle impact assessment assigns the result of inventory analysis to specific impact categories, such Croat. j. for. eng. 33(2012)2

Life Cycle Assessment (LCA) in Forestry – State and Perspectives (357–372)

as depletion of nonrenewable and renewable resources, greenhouse effect, ozone depletion, human toxicity, acidification, etc. (Heijungs 1992b). Finally, lifecycle interpretation formulates conclusions and recommendations as to how the environmental performance of product systems may be improved, or what alternative of different product systems performs best.

2.3 Life Cycle inventory modeling of a product system – Životni ciklus zaliha proizvodnoga sustava Inventory analysis consists of three major steps: first the identification of functions and flows of the product system; second, the quantification of flows; and third, the quantitative modeling of all flows converting into a specific set of output flows. The simplest approach, tabular balancing, is very limited to handle complex networks of flows, resulting in so-called

H. R. Heinimann

»truncated« system inventories (Joshi 1999) that do not fulfill the cradle to grave requirement. Therefore, we will present the formal mathematical approach below, which to our knowledge is not well known in the forest science community, and illustrate it with a practical example. Production economics provides a formal approach to investigate process networks or even economic sectors, which relies on two fundamental concepts: 1) commodities, and 2) activities (Koopmans 1951b, 1951c). An activity, also called a process, consists of a specific technology, which transforms specific input-flows into output-flows according to well-defined procedures. The mapping of process networks as flows on a graph has become an important approach to analyze environmental impacts (Koopmans 1951b, 1951c). Activities are represented as nodes, while arcs represent flows of goods, resources, emissions, and wastes. The resulting

Table 3 Flow of materials, equipment components and services into the functional unit – one productive machine hour PMH of a Stihl 026C chainsaw (Knechtle 1997). Source flows are positive, whereas sink flows are negative Tablica 3. Tok materijala, sastavni dijelovi i servis radne jedinice na temelju jednoga proizvodnoga sata Stihl 026C (Knechtle 1997). Ulazni su tokovi pozitivni, dok su izlazni negativni

Steel, high alloyed, kg Visokolegirani čelik, kg Steel, non-alloyed, kg Nelegirani čelik, kg Aluminum, kg Aluminij, kg Plastic, kg Plastika, kg Gasoline, kg Gorivo, kg Chainsaw oil, kg Ulje za motornu pilu, kg Manufacturing, unit Proizvodnja, komada Guide bar, unit Vodilica, komada Chain, unit Lanac, komada Maintenance, unit Održavanje, komada Chainsaw deployment, PMH Rad motorne pile, radni sat Croat. j. for. eng. 33(2012)2

X10

X11

–1.166

–1.06

–0.24

–0.35

–0.088

–0.026

–2.268

–0.681

–1.127

–0.338

–1.245

–0.296

–8E-04

–0.005

–0.017

–8E-04

361

H. R. Heinimann

Life Cycle Assessment (LCA) in Forestry – State and Perspectives (357–372)

graph is non-cyclic, directed, and finite. Additionally, several source nodes and sink nodes may exist, being located outside of the system’s boundaries. This type of graph has also become known as GOZINTO-graph, following »the part that goes into« (Vazsonyi 1954). Life-cycle assessment began to emerge in the late 1960’s, when Ayres and Kneese published a paper on »production, consumption, and externalities« (Ayres and Kneese 1969). They started from the premise that the capacity of the environment to provide resources and assimilate emissions and waste has become scarce, and that there is a strong need to relief the concept of »free economic goods«, such as water, air, etc. Their guiding idea was that resource extraction and environmental pollution and its control is a »materials balance problem«. They formulated a mathematical approach to model the materials balance problem with an inputoutput approach, considering the products of photosynthesis and mineral resources as inputs, and CO2 and waste as final outputs. Flows on a graph may be represented by a system of linear equations, an example of which is presented in Table 3 for a Stihl 026C chainsaw from (Knechtle 1997). Each row represents the flow of a commodity from the source process (positive unit values) to sink processes (negative unit values).The first row represents the flow of high alloyed steel, of which –1.16 kg are flowing into the manufacturing process, –1.06 into the guide bar, –0.24 into the chain and –0.35 into maintenance (spare parts). Rows 2 to 11 follow the same representational logic, and the whole system is represented by 11 commodities (rows) flowing between 11 processes (columns). Assuming that each process can be scaled by a variable xi, whereas i = 1...11 the system of 11 equations can be solved for X, whereas X is the vector (x1,..., x11), if the total system output Y is known. For Table 3, the elements of the output vector Y are 0, except for x11, which is equal to one unity of the functional unit. The 11×11 matrix of Table 3 defines the chainsaw technology and is called »technology matrix« A (Koopmans 1951b). A unique solution requires 1) a quadratic technology matrix A, and 2) a known balance of inflows and outflows for all commodities. In matrix notation, the equation system (1) of Table 3 is written as A·X=Y

output vector Y as below, equation (2) results in the following solution for X.

The scaling vector X has the following meaning: x5, gasoline consumption, means that 2.96 kg of gasoline are used for a productive machine hour PMH; x1, consumption of high alloy steel, means that about 10 g of steel are consumed per PMH. However, the model just describes materials and energy flows, and it has to be enhanced to analyze the flows of environmental burdens. There is a well-documented approach (Koopmans 1951b; 1951c) that assumes the flow of commodities to be proportional to the flow of environmental burdens (linearity assumptions). A single type of environmental burden may be represented by a vector B, which has the same length as the scaling vector X. Table 4 presents eight environmental burden vectors, written as rows. The burdens of materials and energy carriers were taken from ecoinvent (ECOINVENT, online), whereas the burden associated with chainsaw deployment (x11) characterizes the engine-specific emission profile of a two-stroke engine. The x7 to x10 processes do not have direct burdens, but are used to logically link the flows. The figures in Table 4 define the so-called burden matrix B, which by multiplication with the scaling vector X results in the burden vector b of the total system (3), resulting in the following values for the Stihl 026C example. b = B· X

(3)

(1)

Model analysis is complete, if we know the vector X. It can be found by solving matrix equation (1) for X (2). X = A–1 · Y

(2)

Equation (2) completely describes the flow of commodities for a specific process network. Assuming an

362

Croat. j. for. eng. 33(2012)2

Life Cycle Assessment (LCA) in Forestry – State and Perspectives (357–372)

H. R. Heinimann

Table 4 Environmental burden matrix B for the Stihl 026C chainsaw. The figures for x1 to x6 were taken from the ecoinvent database (ECOINVENT, online) (grey shaded), whereas the values for x11 characterize the combustion process in the 2-stroke engine (dark shaded) Tablica 4. B matrica okolišnoga opterećenja motorne pile Stihl 026C. Izvor je podataka x1 do x6 baza Ecoinvent, dok je vrijednost x11 rad dvotaktnoga motora X1

X10

X11

109.7

33.3

231.4

93.32

9.216

9.21

42.7

CO2, kg

5.282

1.614

9.964

3.229

0.505

0.503

3.926

CO, kg

0.029

0.03

0.004

0.001

9E-04

0.285

CH4, kg

0.016

0.009

0.022

0.008

0.004

2E-04

HC, kg

0.005

0.001

0.011

0.009

0.16

NOx, kg

0.01

0.003

0.02

0.007

0.003

0.004

SOx, kg

0.343

0.005

0.058

0.019

0.003

0.001

Embodied energy, MJ Ugrađena energija, MJ Process energy, MJ Utrošak energ. pri radu

Imported from Ecoinvent (2012) – Izvor Ecoinvent (2012)

The use of a productive machine hour PMH of a chainsaw consumes 65.8 MJ of process energy, 16.1 MJ of embodied energy, and emits 4.79 kg of CO2, 0.29 kg of CO and 0.17 kg of HC. This chainsaw specific burden vector B can be reused for the analysis of production systems, which contribute to comparability of results.

3. State of Modeling Approaches – Pristupi modeliranja 3.1 Conceptual Models – Konceptualni modeli Life cycle inventory analysis has to be based on a conceptual model that defines the building blocks of analysis from »cradle to grave«. The author proposed a conceptual model that is presented in Fig. 1 (Heinimann et al. 2006). Product systems are hierarchically organized. The highest level of organization is the product system level (Fig. 1, level 2, right), consisting of a network of humans, machinery and facilities. The underlying level consists of machines, made of different kinds of materials during the manufacturing process. Combustion engines have been the backbone of forest machinery and the quality of the combustion process is crucial for all subsequent results. Machines consume resources through maintenance, which should also be considered in the analysis process. The materials of which a machine is manufactured emCroat. j. for. eng. 33(2012)2

No direct burden Bez izravnoga opterećenja

Engine combustion Sagorijevanje

body environmental burdens that have to be considered to fulfill their »cradle to grave« requirement. Environmental burdens of materials are documented in databases, such as ecoinvent (ECOINVENT, online). The ecoinvent database contains international industrial life cycle inventory data on energy supply, resource extraction, material supply, chemicals, metals, agriculture, waste management services, and transport services. It is used by about 4’500 users in more than 40 countries worldwide and is included in the leading LCA software tools as well as in various eco-design tools for building and construction, waste management or product design (ECOINVENT, online). Similar databases are spine@cpm (SPINE@CPM, online), ELCD (ELCD, online), or ProBas (PROBAS, online). Fig. 2 presents the product system for solid wood production, as represented in the ecoinvent database (Werner et al. 2007). The first function, biomass growth, consists of three input flows from the environment, the capture of solar energy, the sequestration of CO2, and land occupation. The second function, forest management, has three input flows, too, energy input (diesel combustion), the use of gravel for road maintenance, and the conversion of forest land into industrial land, which is caused by the area occupied by forest roads. The third function, harvesting, has two input flows, energy input (diesel combustion) and chainsaw use (in PMH).

363

H. R. Heinimann

Life Cycle Assessment (LCA) in Forestry – State and Perspectives (357–372)

Fig. 1 Components of the framework of life cycle inventory analysis (Heinimann et al. 2006). Recycling of materials, such as steel is assumed to be included in raw material burdens Slika 1. Dijelovi okvira analize zaliha životnoga ciklusa. Recikliranje materijala, poput čelika, pretpostavlja se da ulazi u granice sirovih materijala

Fig. 2 Conceptual mapping of energy, material and area flow for wood product systems (Werner et al. 2007) Slika 2. Konceptualno kartiranje tokova energije, materijala i područja u proizvodnji drva (Werner i dr. 2007) The harvesting function is a typical case of co-production, which is a term for multiple products coming out of one function (ISO 2002). Co-production requires rules as to how to allocate upstream environmental burdens to a set of products, which is often done proportionally to the economic value of the output products. In the ecoinvent database, the allocation factors are 0.86 for roundwood, 0.09 for industrial wood, and 0.05 for residual wood. The nature of allocation factors is normative, and there is no right or wrong solution to this problem. The re-presentation in ecoinvent does not fully comply with the »cradle to grave« requirement that would request to include environmental burdens that are embodied in forest machines, tools,

364

and in forest roads. Taking into account these shortcomings, (Heinimann et al. 2006; Knechtle 1997) ecoinventories for forest machines and for the construction and maintenance of forest roads were modeled.

4. State of environmental performance knowledge – Stanje znanja o okolišnoj učinkovitosti The selection and quantification of operational performance indicators (OPIs) is a main purpose of LCA (ISO 2006a). OPIs are indicators to measure and compare eco-efficiency (Huppes and Ishikawa 2007), Croat. j. for. eng. 33(2012)2

Life Cycle Assessment (LCA) in Forestry – State and Perspectives (357–372)

which is a relationship between resource consumption, emitted pollutants or deposed waste per a unity of the functional unit. Materials-related indicators are the mass of materials, or the quantity of water used per product unit, which is e.g. a performance indicator for pulp product systems. Energy-related OPIs specify the quantity of energy used per product unit, and emission-related OPIs the mass of specific emissions per product unit (ISO 1999). Below, we present two typical OPIs, energy consumption and CO2 emission per functional unit, as reported in the scientific literature.

4.1 OPIs for solid wood product systems – Pokazatelji ekološke učinkovitosti za drvne proizvode Table 5 presents OPIs for the »solid wood product system«. A first problem that we encountered during the screening of the literature was that the system boundaries were either not clearly specified or differing across studies. The upstream boundary should include the environmental system (Udo de Haes

H. R. Heinimann

1996b) as represented in the ecoinvent model (Fig. 2). However, some of the studies start with the forest management function (see Fig. 2), whereas others seem to consider the harvesting function only. In the Nordic countries, the downstream boundary is usually at the mill gate, whereas in Central European studies, the boundary is at the forest road. Considering the early findings of (Karjalainen and Asikainen 1996) that about two third of the environmental burden of the forest to mill product system are caused by road construction and long distance transport, it would make sense to report two separate OPIs, one for forest management and harvesting, and the other one for long distance transportation, including road infrastructure. A second problem that we encountered was the fact that most of the forestry studies were of level 2 only (see Fig. 1), considering only direct materials and energy consumptions during system deployment and neglecting embodied environmental burdens of upstream functions. According to (Knechtle 1997), the embodied energy in machines like harvesters and forwarders equals to about 40 to 50% of the direct process energy. The higher emission values from the eco-in-

Table 5 Energy input and CO2 output for harvesting operations. The underlying harvesting systems are of harvester-forwarder type, with a motor-manual system for ecoinvent Tablica 5. Utrošak energije i emisija CO2 pri pridobivanju drva sustavom harvester – forvarder, osim u slučaju »Ecoivent« gdje je u pitanju ručno-strojni rad CO2

Source – Izvor

–3

(kg m

Ecoinvent 2012

12.41

ProBas 2012

17.92

Level – Razina

Energy – Energija )

SPINE@COM 2012

–3

(MJ m

)

4.07 kg crude oil

4,07 kg sirove nafte 501

2.4

–

~22

~90

~14

~40

Karjalainen and Asikainen 1996

6.4

–

Knechtle 19993

7.4

110

5.9

63–66

2.4–4.3

4.4

Berg 1997 3,4

González-García et al. 2009, Spain

3,4

González-García et al. 2009, Sweden 3

Lindholm 2006

Schwaiger and Zimmer 2001 3

Valente et al. 2011

1+2

road maintenance included – uključeno održavanje cesta functional unit: 1 kg, conversion with 0.007 m3 per kg – funkcionalna jedinica 1 kg, pretvorba s 0,007 m3 po kg 3 without silvicultural operations – bez uzgojnih radova 4 including transport to mill – uključujući prijevoz do pilane 2

Croat. j. for. eng. 33(2012)2

365

H. R. Heinimann

Life Cycle Assessment (LCA) in Forestry – State and Perspectives (357–372)

ventory databases (Table 5) (ECOINVENT, online, PROBAS, online) may also be explained by the embodied burden. Median values of the studies reported in Table 2 are about 7.5 kg CO2.m-3ub and about 60 MJ.m-3ub, both with considerable variability and uncertainty. There is a need to improve the quality and comparability of future studies by standardizing the definition of system boundaries for the solid wood production system and by defining and investigating identical flows. This has to be achieved on the conceptual level, as illustrated in Fig. 2.

4.2 OPIs for the construction and maintenance of forest roads – Pokazatelji ekološke učinkovitosti pri gradnji i održavanju šumskih cesta Taking the evidence that about 60% of the overall environmental burdens of forest production are caused by road network infrastructure and long-distance transport (Karjalainen and Asikainen 1996; Winkler 1997; Heinimann 1999) as a starting point, Heinimann and Maeda-Inaba (2004) presented an LCA study that followed the conceptual approach presented in Fig. 1, performing both level 1 and level 2 analysis and using eco-inventories for materials and energy carriers from ecoinvent (ECOINVENT, online). Table 6 presents a summary of the results. On moderate slopes up to 40 percent, construction of one unit length (m) of forest road consumes about 350 MJ of energy while emitting about 20 kg of greenhouse gases. This amount of energy consumption is equivalent to the heating value of about 10 l of diesel fuel, and about 10 kg of wood mass that has to be grown to sequestrate the amount emitted greenhouse gas. Transport distance of base course materials is the most sensitive factor of influence. Compared to on-site

preparation of aggregates, a 50-kilometer transport increases energy consumption by a factor of about five. Slope demonstrated to be the second important factor that shows a nonlinear influence on energy consumption and greenhouse gas emissions. Increasing slope to about 50 percent, doubles energy consumption and greenhouse gas emissions, while a slope of 70 percent almost triples them. Roadbed width is the third factor of influence. Energy consumption doubles by increasing it from 4.2 m to 6.2 m. Assuming a life cycle of a forest road of 40 years, a road density of 25 m ha–1, an average yield of 10 m3 ha–1 a—1, and allocating the OPIs of Table 6 to the functional unit (m3ub) of the wood product system results in an amount of embodied energy of 20–40 MJ m–3, which is considerable, compared to the energy input of harvesting operations (Table 5). If the annual yield is lowered to about 5 m3 ha–1 a–1, the roadinduced embodied energy reaches the same order of magnitude as the impact of the product system itself. These findings illustrate that neglecting the forest road infrastructure results in a considerable overestimation of environmenttal performance, particularly for lowyield forests and for difficult terrain.

4.3 OPIs for truck transport systems – Pokazatelji ekološke učinkovitosti pri prijevozu drva kamionima Road transportation with trucks has been the main mode for long-distance hauling. Allowable gross vehicle mass (GVM) was standardized within the European Union, defining the following limits for trucks: 2-axle 18 tons, 3-axle 26 tons, 4-axle 32 tons, and 5-axle 38 tons. Typical configurations for timber transport are a 3-axle motor vehicle plus 2- or 3-axle trailer with a GVM of 40 tons, and a 3-axle motor vehicle plus 4-axle trailer with a GVM of 60 tons in the Nordic countries.

Table 6 Energy input and CO2 output for the construction and maintenance of forest roads. Functional unit: 1 m of road length. Assumptions: (1) roadbed width of 4.2 m, (2) cut slope angle of 1:1, (3) fill slope angle of 4:5, (4) base course thickness of 0.3 m, (5) surface course thickness of 0.08 m, and (6) base course materials transport distance of 10 kilometers, (6) increasing rock excavation on slopes steeper than 50% Tablica 6. Utrošak energije i emisija CO2 prilikom gradnje 1 m šumske ceste širine planuma od 4,2 m, nagiba usjeka 1 : 1, nagiba nasipa 4 : 5, debljine donjega stroja 0,3 m, debljine gornjega stroja 0,08 m, s prijevozom građevnoga materijala na udaljenost od 10 km i povećan iskop materijala na nagibima > 50 % CO2

Energy – Energija

Level 2

Level 1+2

(kg m–1)

(MJ m–1)

Razina 2

Razina 1+2

Slope >10% – Nagib >10 %

315

–

Slope ~ 40% – Nagib ~ 40 %

405

–

Slope ~60% – Nagib ~ 60 %

735

–

Terrain conditions – Terenski uvjeti

366

Croat. j. for. eng. 33(2012)2

Life Cycle Assessment (LCA) in Forestry – State and Perspectives (357–372)

Table 7 gives an overview of energy input and CO2 emissions for these two configurations. The German Pro Bas database (PROBAS, online) provides the most up-to-date eco-inventories for truck transportation, considering euro-5 emission standards, however, without figures for 60 ton configurations. The Swedish Spine database (SPINE@CPM) provides figures for 60 ton configurations, which are based on analysis work done in the late 1990s. Forest-specific figures are only available for Sweden (Lindholm and Berg 2005) and for Finland (Karjalainen and Asikainen 1996), whereas some results, e.g. (Valente et al. 2011b), are not comparable. The ProBas data illustrate that environmental performance depends on the traffic mode (highway, over land, in the city), yielding an increasing gradient from highway to in the city transport. The Swedish data (Lindholm and Berg 2005) give a preliminary hint that timber haulage has its own, forestspecific traffic mode (in the forest, over land, highway) that is not well understood, but probably results in burdens that are close to the »in the city« mode. Assuming a 40 ton configuration, over land mode, and a timber load of 28 m3 yields an energy consumption of about 0.55 MJ m–3ub. Compared to an average energy input of 60 MJ m–3ub (Table 5), a road transport distance of about 100 km results in an environmental burden of the same order of magnitude as from the harvesting process, however, neglecting the embodied energy of the forest road infrastructure (see Table 6), which adds between 20 and 40 MJ m–3ub (see 4.2).

4.4 Environmental performance of bioenergy product systems – Okolišna učinkovitost sustava za proizvodnju bioenergije

H. R. Heinimann

IEA (2011, 2012), assumed that low-carbon technologies would contribute to the reduction of greenhouse gas emission. There are several bioenergy pathways (Fig. 3), 1) biomass in unprocessed form such as firewood, forest residues; 2) biomass intermediates, such as pellets or biomethane from manure or landfill; 3) first-generation biofuels, made of seed, grain, or sugar; 4) second-generation biofuels, manufactured from lignocellulosic biomass, and 5) third-generation bio fuels, manufactured from algae or seaweeds (IEA 2011, IEA 2012, Nigam and Singh 2011). Environmental performance, particularly CO2- and energy-efficiency, is a decisive criterion to choose the best course of action for future biomass-based energy supply. Comparability of results of environmental performance assessment requires first a clear definition of system boundaries, and second, an agreement on the functional unit, which is used to normalize the results. System boundaries should be of a »well to plant« for wood chips, »well to stove« for pellets, and »well to wheel« type for biofuels. As a consequence, functional units should be defined as a unit of energy produced in a plant or stove (MJ), or a unit of transportation service produced (t.km, person.km). The screening of the LCA literature for forest fuel supply indicates that system boundaries are far too narrow, particularly in the downstream direction, and the functional unit is, in most cases, defined in traditional forestry units, such as bulk volume, bulk mass, solid timber volume, etc. The key question still is what technology route out of the possibilities outlined in Fig. 3 is most cost-effective and most environmentally performing. The EIA technology roadmap hypothesizes that the following

Table 7 Energy input and CO2 output for long-distance truck transportation Tablica 7. Utrošak energije i emisija CO2 pri daljinskom transport drva kamionima Source – Izvor

Gross vehicle mass, t Masa vozila, t

Load capacity, t

Emission standard

CO2

Energy

Masa tovara, t

Standardi emisije

(kg· t · km )

(MJ· t–3· km–1)

–3

–1

ProBas 2012 (over land, no highway – izvan autoceste)

Euro 5

0.057

0.67

ProBas 2012 (highway – autocesta)

Euro 5

0.050

0.59

ProBas 2012 (in the city – gradska vožnja)

Euro 5

0.084

0.99

SPINE@COM 2012

Euro 2

0.050

0.68

SPINE@COM 2012

Euro 2

0.041

0.57

Karjalainen and Asikainen 1996

0.038

–

Lindholm and Berg 2005

–

0.99

Croat. j. for. eng. 33(2012)2

367

H. R. Heinimann

Life Cycle Assessment (LCA) in Forestry – State and Perspectives (357–372)

Fig. 3 Classification of Biofuels (Nigam and Singh 2011), modified Slika 3. Podjela biogoriva (modificirano prema Nigam i Singh 2011) bioenergy systems are the most promising: 1) the replacement of traditional biomass by advanced biomass cook stoves and household biogas systems, 2) cogenerating heat-power plants CHP (IEA 2012). EIA hypothesizes that advanced biofuels, such as cellulosic ethanol, advanced biodiesel, bio-syntactic gas BSG from lignocellulosic biomass or algae are more cost-and eco-efficient than first generation biofuels made of sugar and starch (Cherubini et al. 2009). A review paper (Cherubini et al. 2009) assesses the state of knowledge of different technology pathways. The authors conclude that the available studies indicate that electricity or heat generation of biomass have a better environmental performance than biofuels, and that bioenergy chains based on the waste and residue raw materials outperform chains based on dedicated crops. They also mentioned that the »cascading« use of biomass (e.g. first use as building material, followed by use for fiber, followed by energetic use) have the potential to further enhance greenhouse gas savings. A recent LCA-study (Stucki and Jungbluth 2012) on second-generation biofuels presents evidence that systems based on molasse, waste oil glycerine and purified biogas performed best with an amount of CO2emissions of about 120 g pkm–1, whereas oil-based diesel and gasoline emit about 180 g pkm–1 and 200 g pkm–1, respectively. However, there are still considerable challenges (Cherubini et al. 2011), particularly methodological inconsistencies due to the selection of different system boundaries, alternative approaches to estimate greenhouse gas emissions, or

368

alternative allocation rules. The authors also stress that an increasing number of LCA-studies on lignocellulosic biomass, sugarcane, or palm oil is available, whereas contributions on promising feedstocks, such as algae, or advanced biomass processing are still scarce. There are still interesting forestry short rotation crops, usually based on willow or poplar (Björeson 2006; Göranson 2009; Gyuricza et al. 2011). A comparison of alternative bioenergy cropping systems (Adler et al. 2007) showed that systems based on corn, soybean, and alfalfa outperformed a poplar-based system in terms of CO2 emissions by a factor of about 1.5. A study comparing two pellet production systems (sawdust, chips) (Heinimann et al. 2007) resulted in the finding that chip-based supply systems outperform sawdust-based systems in terms of energy efficiency, carbon dioxide emission, eco-toxicity, and particle emissions (PM10) caused by higher moisture content of sawdust. The pellet manufacturing process, consisting of raw materials supply, pellet manufacturing, and pellet distribution, caused about 60 to 80% of the total energy input. The drying process is the most important step of pellet manufacturing with a share of 60 to 80%, and therefore offering potential for efficiency improvement by using e.g. superheated steam dryer technology. IEA technology roadmaps (IEA 2011, IEA 2012) stress that alternative bioenergy pathways require raw material supply chains that are tailored to the end use and optimized for both economic and eco-efficiency. Croat. j. for. eng. 33(2012)2

Life Cycle Assessment (LCA) in Forestry – State and Perspectives (357–372)

However, to our knowledge, there are no comprehensive LCA-studies available that investigate supply chains tailored to specific bioenergy pathways as illustrated in Fig. 3. Available studies on supply chains are limited in scope, as they investigate the process energy use only, and calculate the emissions due to engine combustion with rules of thumb. Therefore, there is a strong need to do further research on biomass supply chains for different bioenergy pathways (Valente et al. 2011a; Valente et al. 2011b).

5. Conclusions – Zaključci The guiding idea of LCA is to improve our understanding of the impacts of alternative product systems on the environment and to quantify environmental performance indicators, characterizing the contribution of products to the main environmental risks. The present contribution reviewed the state of LCA-related research for product systems with forest biomass as raw material. The study resulted in the following findings. 1) Whereas LCA-methodology is looking back on about thirty years of experience, it is still not widely used and accepted within the forest operations engineering research community. 2) Only a few forest-related LCAstudies are based on a quantitative, mathematical stringent methodology that uses systems of linear equations to characterize and model commodity, energy, and substance flows from »cradle to grave«. 3) Although LCA is following a »well to use« philosophy, many forest related LCA-studies are based on quite narrow system boundaries, and on forest-specific functional units, thus limiting the comparability with state-of-the-art LCAstudies. 4) Most of the forest-related LCA-studies are based on direct process energy consumption, measured as fuel consumption, and emission figures that are calculated from fuel consumption by general assumptions. 5) »Truncated LCAs«, neglecting embedded environmental burdens of machines and forest road infrastructure, results in an underestimation of environmental impacts of forest product systems. Whereas environmental analysis tools seem to converge, for example by bringing exergy analysis, which is a traditional field in process engineering, together with LCA-methodology (Hau 2002; Hau and Bakshi 2004), a new, forest-specific research stream has been emerging, sustainability impact assessment of wood supply chains, for which a specific, made-to-purpose software tool was developed, ToSIA (Lindner et al. 2012). This new initiative seems not to be well linked to sustainability impact assessment (SIA) that emerged out of the strategic impact assessment (SIA) and enviCroat. j. for. eng. 33(2012)2

H. R. Heinimann

ronmental impact assessment (EIA) traditions, being designed as policy, and not as analysis tools. Additionally, it is only weakly linked to the ongoing lifecycle management initiative, which addresses the »triple bottom line« with three tools: environmental lifecycle assessment (E-LCA), social lifecycle assessment (S-LCA), and lifecycle costing (LCC) (UNEP and SETAC 2009). The study has several implications for the forest operations research community. First, it has to invest in capacity building to better understand mainstream lifecycle assessment methodology. Second, there is a strong need to develop standards for system boundaries and functional units of typical bioenergy pathways (see for example Fig. 3), which is the basis to improve the comparability of future studies. Third, lifecycle inventories for road construction, road maintenance and road transportation need to be updated because previous studies demonstrated that long-distance transportation and forest road infrastructure account for about two third of the total impact for typical forest productivity systems (Heinimann and Maeda-Inaba 2004; Karjalainen and Asikainen 1996). And forth, future lifecycle inventories have to be linked to LCI databases, such as Ecoinvent (ECOINVENT, online), ProBas (PROBAS, online), etc., to account for materials and energy systems of very upstream processes like the manufacturing of machines. And fifth, future studies should provide information on the standards for the assessment of compliance of machine engines, e.g. Tier 1 to Tier 4 standards in the US (DIESELNET, online-b), Stage I to IV for the European Union (DIESELNET, online-a), because there is a considerable decrease of allowable emission with increasing Tier and Stage.

6. References – Literatura Adler, P. R., Grosso, S. J. D., Parton, W. J., 2007: Life-cycle assessment of net greenhouse-gas flux for bioenergy cropping systems. Ecological Applications 17(3): 675–691. Anon., 1999: Principles of Environmental Impact Assessment Best Practice. IAIA special publications. Accessed [Feb04-2012]. [http://www.iaia.org/publicdocuments/specialublications/Principles%20of%20IA_web.pdf] Anon., 2002: Strategic Environmental Assessment. Performance Criteria. IAIA Special Publication Series, (1). Accessed [Feb-04-2012]. [http://www.iaia.org/publicdocuments/special-publications/sp1.pdf] Anon., 2009: What is Impact Assessment? IAIA special publications. Accessed [Feb-04-2012]. [http://www.iaia.org/publicdocuments/special-publications/What%20is%20IA_web. pdf] Ayres, R. U., Kneese, A. V., 1969: Production, consumption, and externalities. The American Economic Review 59(3): 282–297.

369

H. R. Heinimann

Life Cycle Assessment (LCA) in Forestry – State and Perspectives (357–372)

Ayres, R. U., Noble, S. B., 1978: Materials/energy accounting and forecasting models. In Resources, environment, and economics, R. U. Ayres, Editor. John Wiley & Sons: New York, etc.: p. 95–125. Berg, S., 1995: The environmental load of fossil fuels in Swedish forestry-an inventory for a LCA. In LCA–a challenge for forestry and forest products industry, eds. Frühwald, A. and Solberg, B., p. 57–65, Hamburg. European Forest Institute. Berg, S., 1997: Some aspects of LCA in the analysis of forestry operations. Journal of Cleaner Production 5(3): 211–217. Björeson, P., 2006: Livscykelanalys av Salixproduktion, Lunds University, Luns Tekniska Hôgskola. Lund. Master Thesis, 21 p. Cherubini, F., Bird, N. D., Cowie, A., Jungmeier, G., Schlamadinger, B., Woess-Gallasch, S., 2009: Energy-and greenhouse gas-based LCA of biofuel and bioenergy systems: Key issues, ranges and recommendations. Resources, Conservation and Recycling 53(8): 434–447. Cherubini, F., Strømman, A. H., Hertwich, E., 2011: Effects of boreal forest management practices on the climate impact of CO2 emissions from bioenergy. Ecological Modelling 223: 59–66. Consoli, F., Allen, D., Boustead I., Fava, J., Franklin, W., Jensen, A., De Oude, N., Parrish, R., Perriman, R., Postelthwaite, D., 1993: Guidelines for life-cycle assessment: a »Code of Practice«. Society of Environmental Toxicology and Chemistry (SETAC). Dieselnet, 2012a: Emission Standards – European Union – Nonroad Diesel Engines. [Accessed Aug-30-2012. Available from www: [http://www.dieselnet.com/standards/eu/nonroad.php] Dieselnet, 2012b: Emission Standards – United States – Nonroad Diesel Engines. [Accessed Aug-30-2012. Available from www: [http://www.dieselnet.com/standards/us/nonroad. php] Ecoinvent, 2012: Ecoinvent – Swiss Centre for Life Cycle Inventories. [‘Accessed’ Aug-30-2012]. [www.ecoinvent.org] ELCD, 2012: European Reference Life Cycle Database. [‘Accessed’ Aug-30-2012]. [www.ecoinvent.org] Erkman, S., 1997: Industrial ecology: an historical view. Journal of Cleaner Production 5(1): 1–10. Fischer-Kowalski, M., 1998a: Society’s Metabolism. The Intellectual History of Materials Flow Analysis, Part I, 1860–1970. Journal of Industrial Ecology 2(1): 61–78. Fischer-Kowalski, M., 1998b: Society’s Metabolism. The Intellectual History of Materials Flow Analysis, Part II, 1970–1998. Journal of Industrial Ecology 2(4): 107–136. Frühwald, A., 1995: LCA–a challenge for forestry and forest products industry. In LCA–a challenge for forestry and forest products industry, eds. Frühwald, A. and Solberg, B., p. 9–14. Hamburg. European Forest Institute. González-García, S., Berg, S., Feijoo, G., Moreira, M. T., 2009: Environmental impacts of forest production and supply of pulpwood: Spanish and Swedish case studies. The International Journal of Life Cycle Assessment 14(4): 340–353.

370

Göranson, M., 2009: LCA of transport fuels from short rotation forestry in a long term perspective. Department of Energy and Environment, Division of Environmental Systems Analysis, Chalmers University of Technology, Göteborg, Master Thesis, 56 p. Guinée, J. B., Heijungs, R., Huppes, G., Zamagni, A., Masoni, P., Buonamici, R., Ekvall, T., Rydberg, T., 2010: Life Cycle Assessment: Past, Present, and Future. Environmental science & technology 45(1): 90–96. Gyuricza, C., Hegyesi, J., Kohlheb, N., 2011: Experience of energy willow (Salix sp.) production in short rotation coppice (SRC) experiment and results of life cycle analysis (LCA). Növénytermelés 60(2): 45–65. Hall, C., Lavine, M., Sloane, J., 1979: Efficiency of energy delivery systems: I. An economic and energy analysis. Environmental Management 3(6): 493–504. Hau, J. L., 2002: Integrating life cycle assessment, exergy and emergy analyses. Department of Chemical Engineering, Ohio State University, Master Thesis, 142 p. Hau, J. L., Bakshi, B. R., 2004: Promise and problems of emergy analysis. Ecological Modelling 178(1): 215–225. Heijungs, R., 1992a: Environmental life cycle assessment of products: backgrounds [transl. from Dutch]. Leiden NL: Centrum voor Milieukunde, 130 p. Heijungs, R., 1992b: Environmental life cycle assessment of products: guide [transl. from Dutch]. Leiden NL: Centrum voor Milieukunde, 96 p. Heinimann, H. R., Haessig, J., Wollenmann, R. A., 2007: EcoEfficiency of Pellet Production – Comparison of Log Wood and Sawmill Byproduct Systems. In Austro2007/FORMEC. Meeting the Needs of Tomorrows Forests – New Developments in Forest Engineering, eds. Steinmuller, T. and Stampfer, K., Vienna. Forest Engineering Institute, University of Applied Life Sciences BOKU, Vienna, Austria. Heinimann, H. R., Maeda-Inaba, S., 2004: Environmental Performance Indicators EPIs for Forest Roads Network Systems. In 2004 International Mountain Logging Conference. A joint FERIC, UBC, and IUFRO D3 conference, eds. Hedin, I. B. and Krag, R., published on CD. Vancouver, BC, Canada. Forest Engineering Research Institute of Canada, FERIC. Heinimann, H. R., Wollenmann, R. A., Knechtle, N., 2006: Environmental Performance Analysis of Forest and Agricultural Production Systems. In Life cycle approaches for Conservation Agriculture. SETAC-Europe 16th Annual Meeting, eds. Guinée, J., et al., 156. The Hague, Netherlands, May 08, 2006. Department of Industrial Ecology & Department of Environmental Biology. Hendriks, C., Obernosterer, R., Müller, D., Kytzia, S., Baccini, P., Brunner, P. H., 2000: Material flow analysis: A tool to support environmental policy decision making. Case-studies on the city of Vienna and the Swiss lowlands. Local Environment 5(3): 311–328. Huppes, G., Ishikawa, M., 2007: Quantified eco-efficiency. An introduction with applications. Eco-efficiency in industry and science. Dordrecht: Springer, 330 p. Croat. j. for. eng. 33(2012)2

Life Cycle Assessment (LCA) in Forestry – State and Perspectives (357–372)

H. R. Heinimann

IEA, 2011: Technology Roadmap. Biofuels for Transport. [‘Accessed’ Aug-30-2012]. [http://www.iea.org/papers/2011/biofuels_roadmap.pdf]

Koopmans, T. C., 1951c: Introduction. In Activity Analysis of Production and Allocation, Koopmans, T.C., Editor. Yale University Press: New Haven, London: p. 1–12.

IEA, 2012: Technology Roadmap. Bioenergy for Heat and Power. [‘Accessed’ Aug-30-2012]. [http://www.iea.org/publications/freepublications/publication/2012_Bioenergy_ Roadmap_2nd_Edition_WEB.pdf]

Kurz, W., Dymond, C., White, T., Stinson, G., Shaw, C., Rampley, G., Smyth, C., Simpson, B., Neilson, E., Trofymow, J., 2009: CBM-CFS3: a model of carbon-dynamics in forestry and land-use change implementing IPCC standards. Ecological Modelling 220(4): 480–504.

ISO 1999: ISO 14031 Environmental management – Environmental performance evaluation – Guidelines. Geneva: ISO, 32 p. ISO 2002: ISO 14050 Environmental management – Vocabulary. Geneva: ISO, 29 p. ISO 2006a: ISO 14001 Environmental management – Life cycle assessment – Principles and framework (2nd ed.). Geneva: ISO, 29 p. ISO 2006b: ISO 14040 Environmental management – Life cycle assessment – Principles and framework (2nd ed.). Geneva: ISO, 20 p. ISO 2006c: ISO 14044 Environmental management – Life cycle assessment – Requirements and guidelines. Geneva: ISO, 46 p.

Lindholm, E. L., 2006: Energy Use in Swedish Forestry and its Environmental Impact. Department of Biometry and Engineering, Swedish University of Agricultural Sciences Uppsala, SE, Licenciate Thesis, 28+10+11+8 p. Lindholm, E. L., Berg, S., 2005: Energy requirement and environmental impact in timber transport. Scandinavian Journal of Forest Research 20(2): 184–191. Linder, M., Werhahn-Mees, W., Suominen, T., Vötter, D., Zudin, S., Pekkanen, M., Päivinen, R., Roubalova, M., Kneblik, P., Brüchert, F., 2012: Conducting sustainability impact assessments of forestry-wood chains: examples of ToSIA applications. European Journal of Forest Research 131(1): 21–34. Marsmann, M., 1997: Current LCA-ISO Activities. International Journal of Life-Cycle Assessment 2(1).

Joshi, S., 1999: Product Environmental Life – Cycle Assessment Using Input – Output Techniques. Journal of Industrial Ecology 3(2–3): 95–120.

Nigam, P. S., Singh, A., 2011: Production of liquid biofuels from renewable resources. Progress in energy and combustion science 37(1): 52–68.

Karjalainen, T., Apneseth, T., Esser, P., Finér, L., Jungmeier, G., Košir, B., Kvist, K. E., Roda, J. M., Schaiger, H., Berg, S., Zimmer, B., Welling, J., 2001: Identification of problems using case studies. In Energy, Carbon and Other Material Flows in the Life Cycle Assessment of Forestry and Forest Products. Achievements of the Working Group 1 of the COST Action E9Discussion Paper 10]. EFI, European Forest Institute.

Odum, H. T., Kemp, W., Sell, M., Boynton, W., Lehman, M., 1977: Energy analysis and the coupling of man and estuaries. Environmental Management 1(4): 297–315.

Karjalainen, T., Asikainen, A., 1996: Greenhouse gas emissions from the use of primary energy in forest operations and long-distance transportation of timber in Finland. Forestry 69(3): 215–228. Kärnä, A., Ekvall, T., 1997: Guidelines on life cycle inventory analysis in the Nordic pulp and paper industry. Paperi Ja Puu – Paper and Timber 79(2): 83–85. Knechtle, N., 1997: Materialprofile von Holzerntesystemen – Analyse ausgewählter Beispiele als Grundlage für ein forstechnisches Ökoinventar. [Material profiles of harvesting systems – analysis of selected cases as a foundation of an eco-inventory for forest operations]. Department of Forest Sciences, ETH Zurich, Diploma Thesis, 66 p.

Pope, J., Annandale, D., Morrison-Saunders, A., 2004: Conceptualising sustainability assessment. Environmental Impact Assessment Review 24(6): 595–616. ProBas, 2012. Prozessorientierte Datenbasis für Umweltmanagement-Instrumente. [‚Accessed‘ Aug-30-2012]. [http:// www.probas.umweltbundesamt.de/] Schneider, A., Ingram, H., 1990: Behavioral assumptions of policy tools. Journal of Politics 52(2): 510–529. Schwaiger, H., Zimmer, B., 2001: A comparison of fuel consumption and greenhouse gas emissions from forest operations in Europe. In Energy, Carbon and Other Material Flows in the Life Cycle Assessment of Forestry and Forest Products. Achievements of the Working Group 1 of the COST Action E9Discussion Paper 10]. EFI, European Forest Institute. SPINE@CPM, 2012: CPM Life Cycle Inventory Data. [‚Accessed‘ Aug-30-2012]. [http://cpmdatabase.cpm.chalmers.se/]

Knechtle, N., 1999: Materialprofile von Holzerntesystemen als Ausgangspunkt für Ökoinventare. Schweizerische Zeitschrift für Forstwesen 150(3): 81–87.

Stucki, M., Jungbluth, N., 2012: LCA of biogas from different purchased substrates and energy crops. Accessed [Aug-302012]. [http://www.lcaforum.ch/portals/0/df47/DF47-01%20 Stucki.pdf]

Koopmans, T. C., 1951a: Activity analysis of production and allocation, Cowles Foundation Monograph. Vol. 13. Yale University Press: New Haven, London, 404 p.

Sundberg, U., Svanvist, N., 1987: Fuel consumption as indicator of the economics in mechanization. Scandinavian Journal of Forest Research 2(1-4): 389–398.

Koopmans, T. C., 1951b: Analysis of production as an efficient combination of activities. In Activity Analysis of Production and Allocation, Koopmans, T.C., Editor. Yale University Press: New Haven, London: p. 33–97.

Udo de Haes, H., 1996a: Discussion of general principles and guidelines for practical use. In Towards a methodology for life cycle impact assessment, Udo de Haes, H., Editor. Society of Environmental Toxicology and Chemistry: Brussels: p. 7–30.

Croat. j. for. eng. 33(2012)2

371

H. R. Heinimann

Life Cycle Assessment (LCA) in Forestry – State and Perspectives (357–372)

Ude de Haes, H., 1996b: LCA: Approaches and bottlenecks when applied to agriculture, food and forestry. CML, NL, Leiden. UNEP-IETC, 2012: What are environmentally sound technologies (ESTs)? [Accessed Aug-30-2012. United Nations Environment Program. International Environmental Technology Centre. Available from www: [http://www.unep.org/ ietc/WhatareESTs/tabid/56309/Default.aspx] UNEP and SETAC, 2009: Guidelines for Social Life Cycle Assessment of Products. [‚Accessed‘ Aug-30-2012]. [http:// www.unep.fr/shared/publications/pdf/DTIx1164xPA-guidelines_sLCA.pdf] Valente, C., Hillring, B. G., Solberg, B., 2011a: Bioenergy from mountain forest: a life cycle assessment of the Norwegian woody biomass supply chain. Scandinavian Journal of Forest Research 26(5): 429–436.

Valente, C., Spinelli R., Hillring, B. G., 2011b: LCA of environmental and socio-economic impacts related to wood energy production in alpine conditions: Valle di Fiemme (Italy). Journal of Cleaner Production 19: 1931–1938. Vazsonyi, A., 1954: The Use of Mathematics in Production and Inventory Control. Management Science 1(1): 70–85. Werner, F., Althaus, H. J., Künniger, T., Richter, K., 2007: Life Cycle Inventories of Wood as Fuel and Construction Materials. ecoinvent report. 9. Accessed [Aug-30-2012]. [http:// db.ecoinvent.org/ecoquery/files/09_WoodFuelConstruction. pdf?area=463ee7e58cbf8] Zimmer, B., Wegener, G., 1996: Stoff-und Energieflüsse vom Forst zum Sägewerk. European Journal of Wood and Wood Products 54(4): 217–223.

Sažetak

Analiza životnoga ciklusa u šumarstvu – stanje i perspektiva Okolišno prihvatljive tehnologije ključne su za smanjenje potrošnje ograničenih resursa i smanjenje utjecaja na okoliš. U radu je opisano stanje poznavanja analitičkoga alata i analize životnoga ciklusa uz razradu triju problema: 1) metodološke postavke analize životnoga ciklusa, 2) modeliranja zaliha životnoga ciklusa i 3) pokazatelja okolišne učinkovitosti pri proizvodnji drva. Rezultati istraživanja ogledaju se u sljedećim nalazima: 1) Analiza životnoga ciklusa nema široku primjenu u šumarskoj zajednici. 2) Samo je nekoliko istraživanja napravljeno temeljem najsu vremenijih analiza zaliha životnoga ciklusa. 3) Postavljene granice istraživanja često su preuske, što smanjuje mogućnost usporedbe s uobičajenim istraživanjima životnoga ciklusa. 4) Većina se istraživanja analize životnoga ciklusa u šumarstvu zasniva samo na izravnom utrošku energije te time zanemaruje opterećenje okoliša daljinjim postupcima. 5) Takva »skraćena« analiza životnoga ciklusa zanemarivanjem opterećenja koja nastaju gradnjom šumskih cesta i uporabom šumskih vozila podcjenjuje učinak na okoliš ili precjenjuje okolišnu učinkovitost. U šumarstvu je potrebno dodatno razviti analizu životnoga ciklusa kako bi buduća istraživanja bila što obuhvatnija i kako bi se što lakše mogla usporediti s osnovnim istraživanjima analize životnoga ciklusa. Ključne riječi: analiza životnoga ciklusa, okolišna učinkovitost, proizvodnja drva, ekološka učinkovitost, indus trijska ekologija

Author’s address – Autorova adresa:

Received (Primljeno): July 14, 2012 Accepted (Prihvaćeno): September 14, 2012

372

Prof. Hans Rudolf Heinimann, PhD. e-mail: hans.heinimann@env.ethz.ch Institute of Terrestrial Ecosystems Department of Environmental Systems ETH Zürich Universitätsstrasse 22 8092 Zurich SWITZERLAND Croat. j. for. eng. 33(2012)2

ISSN 1845-5719

9 771845 571000