Environmental & Engineering Geoscience by Association of Environmental & Engineering Geologists (AEG)

Environmental & Engineering Geoscience NOVEMBER 2015

VOLUME XXI, NUMBER 4

THE JOINT PUBLICATION OF THE ASSOCIATION OF ENVIRONMENTAL AND ENGINEERING GEOLOGISTS AND THE GEOLOGICAL SOCIETY OF AMERICA SERVING PROFESSIONALS IN ENGINEERING GEOLOGY, ENVIRONMENTAL GEOLOGY, AND HYDROGEOLOGY

Environmental & Engineering Geoscience (ISSN 1078-7275) is published quarterly by the Association of Environmental and Engineering Geologists (AEG) and the Geological Society of America (GSA). Periodicals postage paid at AEG, 1100 Brandywine Blvd, Suite H, Zanesville, OH 43701-7301. Phone: 844-331-7867 and additional mailing offices. EDITORIAL OFFICE: Environmental & Engineering Geoscience journal, Department of Geology, Kent State University, Kent, OH 44242, U.S.A. phone: 330-672-2968, fax: 330-672-7949, ashakoor@kent.edu. CLAIMS: Claims for damaged or not received issues will be honored for 6 months from date of publication. AEG members should contact AEG, 1100 Brandywine Blvd, Suite H, Zanesville, OH 43701-7301. Phone: 844-331-7867. GSA members who are not members of AEG should contact the GSA Member Service center. All claims must be submitted in writing. POSTMASTER: Send address changes to AEG, 1100 Brandywine Blvd, Suite H, Zanesville, OH 43701-7301. Phone: 844331-7867. Include both old and new addresses, with ZIP code. Canada agreement number PM40063731. Return undeliverable Canadian addresses to Station A P.O. Box 54, Windsor, ON N9A 6J5 Email: returnsil@imexpb.com. DISCLAIMER NOTICE: Authors alone are responsible for views expressed in articles. Advertisers and their agencies are solely responsible for the content of all advertisements printed and also assume responsibility for any claims arising therefrom against the publisher. AEG and Environmental & Engineering Geoscience reserve the right to reject any advertising copy. SUBSCRIPTIONS: Member subscriptions: AEG members automatically receive the journal as part of their AEG membership dues. Additional subscriptions may be ordered at $40 per year. GSA members who are not members of AEG may order for $40 per year on their annual GSA dues statement or by contacting GSA. Nonmember subscriptions are $175 and may be ordered from the subscription department of either organization. A postage differential of $10 may apply to nonmember subscribers outside the United States, Canada, and Pan America. Contact AEG at 303-757-2926; contact GSA Subscription Services, c/o American Institute of Physics, 2 Huntington Quadrangle, Suite 1N01, Melville, NY 11747-4502. Single copies are $50.00 each. Requests for single copies should be sent to AEG, 1100 Brandywine Blvd, Suite H, Zanesville, OH 43701-7301. Phone: 844-331-7867. © 2015 by the Association of Environmental and Engineering Geologists All rights reserved. No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or by any information storage and retrieval system, without permission in writing from AEG. THIS PUBLICATION IS PRINTED ON ACID-FREE PAPER EDITORS ABDUL SHAKOOR Department of Geology Kent State University Kent, OH 44242 330-672-2968 ashakoor@kent.edu

BRIAN G. KATZ Florida Department of Environmental Protection 2600 Blair Stone Rd. Tallahassee, FL 32399 850-245-8233 eegeditorbkatz@gmail.com

EDITORIAL BOARD ROBERT H. SYDNOR JEROME V. DEGRAFF USDA Forest Service Consulant THOMAS J. BURBEY CHESTER F. WATTS (SKIP) Virginia Polytechnic Institute Radford University SYED E. HASAN University of Missouri, Kansas City ASSOCIATE EDITORS JOHN W. BELL PAUL M. SANTI Nevada Bureau of Mines and Colorado School of Mines Geology ROBERT L. SCHUSTER U.S. Geological Survey RICHARD E. JACKSON (Book Reviews Editor) ROY J. SHLEMON R. J. Shlemon Geofirma Engineering, Ltd. & Associates, Inc. JEFFREY R. KEATON AMEC Americas GREG M. STOCK National Park Service PAUL G. MARINOS National Technical University RESAT ULUSAY Hacettepe University, Turkey of Athens, Greece CHESTER F. “SKIP” WATTS JUNE E. MIRECKI U.S. Army Corps of Radford University Engineers TERRY R. WEST Purdue University PETER PEHME Waterloo Geophysics, Inc NICHOLAS PINTER Southern Illinois University SUBMISSION OF MANUSCRIPTS Environmental & Engineering Geoscience (E&EG), is a quarterly journal devoted to the publication of original papers that are of potential interest to hydrogeologists, environmental and engineering geologists, and geological engineers working in site selection, feasibility studies, investigations, design or construction of civil engineering projects or in waste management, groundwater, and related environmental fields. All papers are peer reviewed. The editors invite contributions concerning all aspects of environmental and engineering geology and related disciplines. Recent abstracts can be viewed under “Archive” at the web site, “http://eeg.geoscienceworld.org”. Articles that report on research, case histories and new methods, and book reviews are welcome. Discussion papers, which are critiques of printed articles and are technical in nature, may be published with replies from the original author(s). Discussion papers and replies should be concise. To submit a manuscript go to http://eeg.allentrack.net. If you have not used the system before, follow the link at the bottom of the page that says New users should register for an account. Choose your own login and password. Further instructions will be available upon logging into the system. Please carefully read the “Instructions for Authors”. Authors do not pay any charge for color figures that are essential to the manuscript. Manuscripts of fewer than 10 pages may be published as Technical Notes. For further information, you may contact Dr. Abdul Shakoor at the editorial office. Cover photo Air photo of the Lower San Diego River Valley showing the Qualcomm Stadium area north of downtown San Diego, California. The yellow outline is that of a plume of tertiary butyl alcohol (TBA), a biodegradation product of the gasoline additive MTBE, which was unintentionally released in the late 1980s from the fuel terminal shown at the upper right corner of the photo (see article on pages 000–000).

Environmental & Engineering Geoscience Volume 21, Number 4, November 2015 Table of Contents 249

The Late Quaternary History and Groundwater Quality of a Coastal Aquifer, San Diego, California Robert M. Sengebush, Dru J. Heagle, and Richard E. Jackson

277

The Timing of Susceptibility to Post-Fire Debris Flows in the Western United States Jerome V. DeGraff, Susan H. Cannon, and Joseph E. Gartner

293

Measuring Orientations of Individual Concealed Sub-Vertical Discontinuities in Sandstone Rock Cuts Integrating Ground Penetrating Radar and Terrestrial LIDAR Norbert H. Maerz, Adnan M. Aqeel, and Neil Anderson

311

Cut Slope Design for Stratigraphic Sequences Subject to Differential Weathering: A Case Study from Ohio Yonathan Admassu and Abdul Shakoor

325

Understanding Karst Leakage at the Kowsar Dam, Iran, by Hydrogeological Analysis Morteza Mozafari and Ezzatollah Raeisi

The Late Quaternary History and Groundwater Quality of a Coastal Aquifer, San Diego, California ROBERT M. SENGEBUSH INTERA Inc., 6000 Uptown Boulevard NE, Suite 220, Albuquerque, NM 87110, USA

DRU J. HEAGLE Geofirma Engineering, 1 Raymond Street, Suite 200, Ottawa, Ontario K1R 1A2, Canada

RICHARD E. JACKSON Geofirma Engineering, 11 Venus Crescent, Heidelberg, Ontario N0B 2M1, Canada

Key Terms: Hydrogeology, Quaternary Geology, Groundwater Quality, Coastal Aquifers

ABSTRACT Prior to World War II, the City of San Diego, California, extracted millions of gallons of high-quality groundwater daily from alluvial gravels in the lower San Diego River Valley that have since become contaminated with brackish water and hydrocarbons. The origin of this brackish groundwater and of the Quaternary sedimentary geology of the valley is interpreted through archived reports, journal articles, U.S. Geological Survey data, and samples from new city wells in the alluvial gravels. Eocene sediments were inundated by seawater during the last interglacial period (ca. 120 ka), when sea levels were ,19 ft (6 m) higher than present levels. The brackish groundwater present in these Eocene sediments appears to be relict seawater from this inundation. We hypothesize that the city’s pre–World War II well field—referred to herein as the Mission Valley Aquifer—was a buried channel gravel created following the Last Glacial Maximum of the Pleistocene Epoch (,20 ka). As such, it would have been similar to other long (,11 km, 7 mi) buried channel gravels along the southern Californian coast described in previous reports. We present evidence of groundwater freshening of the Eocene sedimentary rock that has led to increasing total dissolved solids in the Mission Valley Aquifer, which acts as a highpermeability drain for the valley. Freshening occurs as a Ca-HCO3 groundwater replaces a Na-Cl water, which we propose was derived from the marine inundation of 120 ka.

INTRODUCTION The City of San Diego is dependent for ,80 percent of its water supply from distant sources that could be interrupted by seismic events, severing the aqueducts from the Colorado River (San Diego Project) and northern California (State Water Project). In addition, surface-water sources such as the California State Water Project and the Colorado River are threatened by drought and thus are no longer as reliable as in the past. Prior to World War II (WWII), the City of San Diego utilized groundwater from a high-permeability alluvial aquifer in the lower San Diego River Valley (“the valley”). This aquifer, which is the subject of this article, was developed in 1914 but abandoned before WWII. It is the city’s goal to re-develop this groundwater resource once remediation of the Mission Valley Terminal (MVT) fuel release has been completed and thus provide further diversification of the city’s water supply. It is the intent of this remediation to restore background groundwater quality conditions (San Diego RWQCB, 2005). Our motivation in preparing this article is to identify the background groundwater quality (GWQ) in the valley to bring closure to the MVT remediation that began in 1992; a discussion of background GWQ concludes this article. The study area (Figure 1) consists of the lower San Diego River Valley, from its outlet near Mission Bay eastward and upstream along the San Diego River to the vicinity of Qualcomm Stadium, a distance of approximately 8 km (5 mi). The main axis of the valley contains the San Diego River and its related floodplain. Murphy Canyon Creek is now confined to a concrete channel that directs flow into the river and thence to the Pacific Ocean. The lower San Diego

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 249–275

249

Sengebush, Heagle, and Jackson

Figure 1. The lower San Diego River Valley showing the locations of the monitoring wells, stream sampling locations, the Mission Valley Terminal, and the Qualcomm Stadium, all of which are mentioned in the text.

River Valley is frequently identified as “Mission Valley”; however, we use this term only to refer to the high-permeability alluvial aquifer, i.e., the Mission Valley Aquifer (MVA), which was pumped as the pre-WWII groundwater supply. Our primary objective is to present a hypothesis describing the Quaternary history of the valley based upon similar histories elsewhere along the southern Californian coastline (e.g., Edwards et al., 2009) and evidence from borehole logs and cores collected in the valley. We then use that hypothesized history to explain the occurrence of brackish GWQ in the Quaternary-age alluvial deposits and the deeper, Eocene-age sedimentary rocks beneath the Valley. GEOLOGIC HISTORY OF THE LOWER SAN DIEGO RIVER VALLEY Sources of Information The valley contains Cretaceous through Holocene rocks and sediments deposited over the past 145 m.y. under a multitude of paleoenvironmental conditions. Understanding the geologic history and stratigraphic relationships of the rock layers is important to characterize the aquifer now intended for sustainable development. Accordingly, we summarize the stratigraphy and geologic history of the valley based mainly

250

on the investigations of Geofirma and INTERA from 2010 through 2014, on reconstructions of the broad geologic history of the San Diego area (Abbott, 1999; Kennedy and Peterson, 2001), and on interpretation of new, site-specific subsurface lithologic and GWQ data. Most of our detailed stratigraphic interpretations of the Quaternary alluvial deposits are based on geologic logs of groundwater monitoring wells prepared for Kinder Morgan Energy Partners (KMEP). KMEP is the owner and operator of the Mission Valley Terminal (MVT) bulk fuel storage facility located at the mouth of Murphy Canyon. In addition, the City of San Diego has installed since 2011 several monitoring wells in the buried channel aquifer at the intersection of Interstates 8 and 805, as well as one north of the MVT within Murphy Canyon and two more on the northern edge of the Qualcomm Stadium parking lot (see Figure 1). These new wells now provide site-specific information about the alluvium and Friars Formation bedrock through lithologic and geophysical logs, petrographic analysis, and laboratory grain-size analysis. In 2004, the U.S. Geological Survey (USGS, 2014) installed a monitoring well cluster known as the Aquaculture Well (SDAQ, San Diego), a multi-depth monitoring well consisting of five nested piezometers, which also provides geologic context. This well is located on the south side of the river across from Qualcomm Stadium (see

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 249–275

History and Groundwater Quality of a Coastal Aquifer, San Diego, California

Figure 2. The stratigraphic column for the San Diego River Valley (Kennedy and Peterson, 2001) showing units present within the study area.

Figure 1); it provides a vertical stratigraphic column to a depth of approximately 940 ft (286 m) below ground surface (bgs). The lithologic and geophysical logs of this well and periodic water sampling and analysis provide a deep vertical profile of Quaternary sediments, Eocene bedrock, and GWQ within the valley.

Stratigraphic History Figure 2 is a stratigraphic column illustrating the geologic units in the study area. The oldest deposits exposed in the area are of Late Jurassic age and of volcanic and marine origin, regionally known as Santiago Peak Volcanics (Jsp) (Kennedy and

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 249â&#x20AC;&#x201C;275

251

Sengebush, Heagle, and Jackson

Peterson, 2001), and outcrops of these rocks occur in the San Diego River Valley about 3 miles (4.8 km) east of Qualcomm Stadium. Kennedy and Peterson (2001, cross section B-B9, La Mesa Quadrangle Geologic Map) show these rocks underlying the entire cross section. The Santiago Peak formation consists of volcanic, volcaniclastic, and sedimentary rocks, but it also includes small plutons. These rocks are in nonconformable contact (sedimentary/volcanic rocks in contact with igneous rocks) with the Cretaceous-age southern California batholith (Tanaka et al., 1984). The depositional environment during the Eocene was that of an advancing and retreating shallow sea, which resulted in transgressive-regressive sedimentary sequences. Alluvial fans were built seaward, pushing the shoreline to the west (Abbott, 1999). Sediments deposited during this time consisted of both the La Jolla Group, west of the study area, and the Poway Group, which constitutes the rocks in the San Diego River Valley in the vicinity of Qualcomm Stadium. The La Jolla Group is only represented by the Friars Formation in the study area. The Poway Group includes the Stadium, Mission, and Pomerado Formations and consists of sediments laid down by the Eocene Ballena River, an ancestral west-flowing channel originating east of San Diego but now displaced by lateral and vertical movements of the San Andreas and related fault systems (Abbott and Smith, 1989). The late middle Eocene–age Friars Formation (Tf) of the La Jolla Group overlies the Santiago Peak Volcanics throughout the San Diego River Valley area. The Friars Formation crops out extensively on the walls of Murphy Canyon and San Diego River canyon and is named for exposures along the north side of the valley near Friars Road on the La Jolla and La Mesa geologic quadrangles (Kennedy and Peterson, 2001). The Friars Formation consists of sand- and clay-stone and contains both non-marine and lagoonal facies, up to 492 ft (150 m) thick in the study area. Table 1 presents the mineralogical composition of the Friars Formation. The Stadium Formation (Stadium conglomerate) lies directly over the Friars Formation in the study area and contains clasts rounded from fluvial transport and composed predominantly of rhyolite, dacite, and quartzite, according to Abbott and Smith (1989). The Stadium conglomerate is about 160 ft (50 m) thick and crops out on the sides of Murphy Canyon and the sides of Serra Mesa, bordering Qualcomm Stadium. Abbott and Smith (1989) postulated that the Stadium formation originated from volcanic sources near Sonora, Mexico, an area that has since been separated from San Diego by lateral slip along

252

the intervening San Andreas and related plateboundary faults. The Stadium conglomerate is overlain by the Mission Valley Formation (Tmv) and the Pomerado Conglomerate (Tp), completing the Poway Group of Eocene sand and gravel deposits. The Mission Valley Formation is about 200 ft (60 m) thick, while the Pomerado is about 180 ft (55 m) thick. These units inter-tongue from east to west. The Pliocene units directly overlie the Eocene strata. The Pliocene rocks consist of the San Diego Formation (Tsd) marine sandstone. During Pliocene time (5.3–1.8 m.y. ago), continental glaciers grew in the Northern Hemisphere, resulting in a sea-level fall of about 625 ft (190 m), according to Abbott (1999). Invertebrate, marine mammal, fish, and bird fossils are abundant in the 100-ft-thick (30-m-thick) San Diego Formation, readily seen in road cuts on the south side of the valley near the mesa top. Quaternary Geology and Hydrogeology The Pleistocene rocks and sediments in the study area consist of the Linda Vista Formation (Ql) and terrace alluvium (Qt). The Linda Vista Formation, which consists of redeposited sand and conglomerate derived from nearby older sediments, is mapped on the top of the mesa bordering the valley. The Pleistocene stream-terrace deposits within the valley, as mapped by Kennedy and Peterson (2001), are located at the foot of the mesas on the north side of the valley and extend from the mesa slopes to the banks of the river, a distance of up to ,3000 ft (1 km). Holocene deposits are San Diego River alluvium (Qal) and slopewash (Qsw) (Kennedy and Peterson, 2001). These sediments have been largely covered by urban development, including the paved parking lot of Qualcomm Stadium, and are described as “poorly consolidated, conglomeratic sand deposits” (Kennedy and Peterson, 2001, p. 50). Petrographic analysis of the terrace alluvium in MW-03 at 59 ft (18 m) bgs reveals these sands are composed of almost entirely polycrystalline porphyritic dacite volcanic and pyroclastic lithic grains, equigranular granite to granodiorite intrusives, and subordinate metamorphic lithic grains and monocrystalline quartz. Table 1 presents the mineralogical composition by DePangher (2014); locations of wells are shown in Figure 1. The geomorphic feature extending from the mouth of Murphy Canyon to the modern San Diego River has been interpreted as the Murphy Canyon alluvial fan (Geofirma Engineering Ltd. and INTERA, 2011), which is fed by a drainage basin of 13 mi2 (33.6 km2). The fan dimensions and morphology are presented in

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 249–275

History and Groundwater Quality of a Coastal Aquifer, San Diego, California Table 1. Mineralogical composition of the basal gravel unit of the MVA and of the Friars Formation at MW-3. Mineral

Percent

Photomicrograph

Lithologic Description

Quartz Plagioclase K-feldspar Clay Organic matter Carbonate Biotite Illite Chlorite

26 20 20 20 5 3 2 2 1

Friars sandstone Scale: ,27 mm across The sample is an unconsolidated sand derived almost entirely from a carbonaceous clayey arkose sandstone protolith From borehole at MW-3, sampled at 67 ft (20 m) bgs (DePangher, 2014).

Plagioclase Quartz K-feldspar Hornblende Actinolite Ferric oxide Sericite Biotite and chlorite

74 10 8 2 2 2 1 ,1

Basal gravel unit of the MVA Scale: ,27 mm across The sample is an unconsolidated lithic sand composed almost of entirely polycrystalline lithic grains: polycrystalline porphyritic dacite volcanic and pyroclastic lithic grains, equigranular granite to granodiorite intrusives, and subordinate metamorphic lithic grains and monocrystalline quartz From borehole at MW-3, sampled at 59 ft (18 m) bgs (DePangher, 2014).

Table 2. These dimensions suggest that the Murphy Canyon alluvial fan fits the type II fans identified by Blair and McPherson (1994) with sediment transport dominated by sheetflow. We believe the Murphy Canyon fan was derived primarily from the Poway Group and the Friars sandstone. The evidence for this interpretation stems mainly from lithofacies in the boring logs of monitoring wells throughout the area. In particular, the rounded volcanic cobbles described in the KMEP boring logs and observed in core from the city’s recent (INTERA, 2014) monitoring well borings—often described as up to 4 in. (100 mm) in diameter—are evidence of the correlation between the cobbles in the Table 2. Dimensions of the Murphy Canyon alluvial fan. Parameter

Value

Width of toe of fan, m (mi) Length from mouth to toe, m (mi) Slope from mouth to toe, m (ft) Angle of slope, deg

1,584 (1) 800 (0.5) 6 (20) 0.43

Stadium and Pomerado conglomerates and the cobbles in the basal and surficial gravel deposits in the Murphy Canyon alluvial fan. Textural analysis of the basal gravel within the alluvium (MW-3, 59 ft or 18 m bgs) describes the material as poorly graded gravel with sand: median grain size d50 5 12 mm and uniformity coefficient Cu 5 d60/d10 5 69. By comparison, the Friars Formation (MW-3, 67 ft or 20 m bgs) is finer grained: d50 5 0.17 mm and Cu 5 171 (Daniel B. Stephens and Associates, Inc., 2014). Figure 3 presents grain-size distribution curves for these samples. Lithologic logs from monitoring wells installed in these sediments for the MVT remediation project have been used to map the subsurface lithofacies of the Qt stream-terrace deposits (Geofirma Engineering Ltd. and INTERA, 2011). Schematic geologic cross section A-A9 (Figure 4) illustrates the vertical and lateral distribution of various sediment types within the Qt deposits. The cross section identifies three principal layers within the Qt beneath the MVT and off-terminal remediation area: the basal gravel

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 249–275

253

Sengebush, Heagle, and Jackson

Figure 3. Grain-size distribution in samples of the basal gravel (59 ft bgs, 18 m) and the Friars Formation (67 ft bgs, 20 m), MW-3 borehole, front entrance to the Qualcomm Stadium, Friars Road, San Diego.

(eroded into the Friars Formation); the middle sand layer (including zones of silt and clay within the sand); and the upper gravel layer. These three units are generally identifiable in the wells across the MVT site (although the upper gravel is absent in some of the lithologic logs). The basal gravel contains a channel structure (“paleochannel”) that trends northeast to southwest beneath the Qualcomm Stadium parking lot. This basal gravel deposit, including the paleochannel, together with the overlying middle sand unit comprise the MVA in the context of both the MVT remediation program and its future use as a city water-supply aquifer (Geofirma Engineering Ltd. and INTERA, 2011, 2013). This paleochannel, now obscured beneath the stadium parking lot, was the location of several of the City of San Diego watersupply wells completed and operated prior to WWII (Fay, 1914; Ellis and Lee, 1919). Additional city wells were located along the axis of the river downstream from the alluvial fan. This combination of unconsolidated gravel, sand, silt, and clay beneath Qualcomm Stadium represents a Pleistocene fluvial depositional environment with its major source in Murphy Canyon. A schematic block diagram (Figure 5) illustrates the various layers and their spatial relationships. We describe three Qt lithofacies—the basal gravel, middle sand, and upper gravel—using Miall’s (1985)

254

classification. We interpret the basal gravel deposit, with its principal lithofacies of massive, matrixsupported gravel (Gms) with massive or crudely bedded gravel (Gm) and coarse to very coarse sand (St), to be a channel (CH) element. The coarse sand matrix would result in high hydraulic conductivity for this basal gravel layer, which is consistent with the grain-size analysis and hydraulic conductivity estimates from sediment samples of the basal gravel in the Murphy Canyon and stadium parking lot wells installed by the City of San Diego (INTERA, 2014). In addition, the concave-up erosional base (as defined by the elongate and trough-like contact between the basal gravel and the Friars Formation) fits with the third-order contact within the hierarchy of bedding contacts as described by Allen (1983). Our classification is consistent with the CH architectural element interpretation of Miall (1985, 1992). Because Quaternary terrace deposits do not crop out in the immediate study area, we illustrate the depositional model from well-known examples near Socorro, New Mexico (Figure 6). The photograph shows a typical matrix-supported gravel overlying a sand unit similar to the Qt channel contact with the underlying Friars sandstone. Our lithofacies classification is provided in Table 3, and the stratigraphic architectural elements are given in Table 4 (Miall, 1985).

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 249–275

History and Groundwater Quality of a Coastal Aquifer, San Diego, California

Figure 4. Cross section A-A9 through the Qt stream-terrace and alluvial-fan deposits indicating the basal gravel paleochannel aquifer (MVA), Qualcomm Stadium area, San Diego.

Pleistocene Paleochannels The Murphy Canyon alluvial fan contains a Pleistocene gravel channel beneath the Qualcomm Stadium parking lot (the “paleochannel”), which extends from the mouth of Murphy Canyon southwest beneath the Qualcomm Stadium parking lot to the river. This paleochannel (within the area mapped as Quaternary terrace deposits, Qt; Kennedy and Peterson, 2001) incises the Friars Formation and is mapped by lines of equal gravel thickness (Figure 7) based on the gravel lithofacies thickness in monitoring well logs (Geofirma Engineering Ltd. and INTERA, 2011). Additionally, the elevation of the contact surface at the base of the gravel and the top of the Friars Formation was mapped (Figure 8). This shows a thalweg that spatially coincides with the axis of the gravel thickness isopach map. This subsurface structure ranges from about 300 ft (91 m) to 600 ft (183 m) wide and is 3,600 ft (1,097 m) long from

the mouth of Murphy Canyon to the toe of the alluvial fan. The slope of the channel parallel to the channel axis is as much as 29 ft (8.75 m) over the length of 3,637 ft (1109 m), or 0.46 degrees. The channel bank slope (perpendicular to the channel axis) near the southwest corner of the stadium parking lot is 3.8 degrees. Bull (1991, p. 172) observed that “unusually warm sea-surface temperatures at about 125 ka should have favored stronger and more frequent tropical storms in the San Gabriel Mountains” in Los Angeles County; this is no doubt true for San Diego County at the same time. Such storms would have accelerated the erosion of Cenozoic sediments in the coastal areas of southern California, potentially leading to the erosion of the San Diego River Valley and its tributaries. The paleochannel down cut through existing Eocene sediments in the valley, in particular, the Friars Formation, and was filled with an upward-fining sequence of gravels, sands,

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 249–275

255

Sengebush, Heagle, and Jackson

Figure 5. Conceptual block diagram of sedimentary lithofacies, Qualcomm Stadium area, San Diego.

and silts similar to modern gravel channels found throughout the Basin and Range Province. This buried-channel aquifer beneath the Qualcomm Stadium parking lot (within the Murphy Canyon alluvial

fan) and the buried channel beneath the San Diego River itself—collectively defined as the MVA—was used by the City of San Diego prior to World War II as its primary water supply, yielding from 2–5 million

Figure 6. Alluvial deposit outcrop showing channel gravel unit in erosional contact with sand unit, Socorro, NM, analogous to basal gravel contact with the Friars Formation. Vertical scale approximately 10 ft (3m).

256

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 249–275

History and Groundwater Quality of a Coastal Aquifer, San Diego, California Table 3. Lithofacies classification for Qt deposits (from Miall, 1985). Facies Code

Lithofacies

Sedimentary Structures

Gms Gm

Massive, matrix-supported gravel Massive or crudely bedded gravel

Gt Gp

Gravel, stratified Gravel, stratified

Grading Horizontal bedding, imbrication Trough cross-beds Planar cross-beds

Sand, medium to v. coarse, may be pebbly Sand, medium to v. coarse, may be pebbly

Solitary (theta) or grouped (pi) trough cross-beds Solitary (alpha) or grouped (omikron) planar cross-beds

gallons per day of high-quality groundwater (,400 mg total dissolved solids [TDS]/L; Ellis and Lee, 1919). The buried paleochannel within the Murphy Canyon alluvial fan is interpreted as analogous to the late Pleistocene buried channel of the lower Santa Margarita River, as described by Shlemon (1979), and to other buried Pleistocene channels along the southern California coast that grade to marine isotope stage 2, indicating the eustatically lowered sea levels of the Last Glacial Maximum (LGM) (Shlemon, 1979; Edwards et al., 2009; and Lee and Normark, 2009). The Santa Margarita River and its estuary are located in northern San Diego County, 40 mi (64 km) north of San Diego River Valley. The buried channel lies beneath the modern Santa Margarita River and is identified by buried gravels 75 ft (23 m) thick, extending to a depth of 150 ft (45 m) below sea level and 7 mi (11 km) long. The late Pleistocene shoreline was over 1.8 mi (3 km) west of the present coast, and the gradient in the channel was steep compared to that of the modern Santa Margarita River. Subsequent sea-level rise covered the channel with finer-grained sediments, resulting in a fining-upward sedimentary sequence. We believe that the paleochannel mapped beneath the Qualcomm Stadium parking lot intersects and is tributary to a larger and deeper paleochannel beneath the San Diego River (Figure 9), similar to the Santa Margarita River buried paleochannel. This paleochannel beneath the river was the location of several of the city’s pre-WWII MVA water-supply wells (Figure 10).

Interpretation Debris-flow deposits Longitudinal bars, lag deposits, sieve deposits Minor channel fills Linguoid bars or deltaic growths from older bar remnants Dunes (lower flow regime) Linguoid, transverse bars, sand waves (lower flow regime)

Last Interglacial Sea-Level Highstand In the San Diego area, and in the lower San Diego River Valley in particular, the Pleistocene was a time of repeated sea-level rise and fall as worldwide glaciers advanced and retreated (Abbott, 1999). Erosional features related to these sea-level changes may be seen today as terraces in and around La Jolla, San Diego, and on the mesas above San Diego River Valley. The number and spacing of terraces were determined by the rate of tectonic uplift and nature of coastal processes. The oldest terrace is generally correlated with the early Pleistocene (1.18 Ma to 120 ka) according to Muhs et al. (2002). Kern and Rockwell (1992) documented 16 separate marine terraces, ranging in age from 1.29 Ma to 80 ka. These erosional wave-cut platforms mark the highest sea-level elevations maintained during glacial/interglacial time. Muhs et al. (2002) also described marine terraces with a focus on those near Point Loma, the lower Bird Rock terrace at about 26 ft (8–9 m) above present sea level, and the higher, Nestor terrace, about 75 ft (23–24 m) above present sea level. The Nestor terrace dates to 120 ka, while the Bird Rock terrace is more recent, dated at 80 ka. The Bird Rock terrace formed at about 26 ft (22 m) relative to present sea level, while the higher Nestor terrace formed about 20 ft (6 m) above present sea level (Table 3; Kern and Rockwell, 1992). Based on the estimated tectonic land-surface uplift that has taken place over the past

Table 4. Architectural elements for the observed lithofacies in the Qt fluvial deposits (from Miall, 1985). Element

Symbol

Principal Lithofacies Assemblage

Channels

Any combination

Gravel bars and bed forms Sandy bed forms

Gm, Gp, Gt

St, Sp, Sh, Sl, Sr, Se, Ss

Geometry and Relationships Finger, lens or sheet; concave-up erosional base; scale and shape highly variable; internal concave-up secondary erosion surfaces common Lens, blanket; usually tabular bodies; commonly interbedded with SB Lens, sheet, blanket, wedge; occurs as channel fills, crevasse splays, minor bars

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 249–275

257

Sengebush, Heagle, and Jackson

Figure 7. Paleochannel gravel isopach map showing channel gravel thickness.

120,000 years, the inland extent of the 120 ka (Nestor) sea-level invasion may be estimated. Uplift resulting from offset on the Rose Canyon fault dominates the tectonic history of the San Diego coast. Uplift has occurred at a rate of 0.13–0.14 m/ k.y., with both higher and lower rates near the Rose Canyon fault zone (Kern and Rockwell, 1992). Consequently, the uplift is interpreted to be 55 ft (17 m) since the last interglacial ca. 120 ka. Additionally, the sea level represented by the Nestor terrace was 19 ft (6 m) above the present sea level prior to uplift. Therefore, the elevation of the marine inundation of the valley would have been 74 ft (23 m) above the present sea level, which is shown in Figure 11. This illustration shows the marine inundation of the valley to a maximum position in the vicinity of the San Diego Mission, with some inundation of Murphy Canyon. Abbott (1999, pages 202–203) has calculated and shown a similar marine highstand during this interglacial.

258

GROUNDWATER QUALITY IN THE LOWER SAN DIEGO RIVER VALLEY The late-middle Eocene–age sediments, such as the Friars Formation, were deposited ca. ,50 to ca. 34 Ma and were inundated repeatedly by rising sea levels during the Pleistocene. These Eocene sedimentary rocks contain brackish groundwater with TDS ,2,000 mg/L, which is either (1) connate water trapped during sedimentary deposition or (2) seawater that inundated the valley during Pleistocene time. In both cases, the residual salinity would have become diluted by freshwater recharge flowing through the valley flow system. In this section, we evaluate the role of two Pleistocene events: (1) the marine inundations during the last interglacial (marine oxygen isotope substage 5e; Shackleton, 1969) and earlier Pleistocene interglacials and (2) the subsequent MVA deposition during the LGM. Then, we attempt to determine how they might have influenced the GWQ in the valley since 1915, when measurements began.

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 249–275

History and Groundwater Quality of a Coastal Aquifer, San Diego, California

Figure 8. Paleochannel morphology map showing erosional surface at contact between Qt basal gravel and Tf sandstone.

We initially present USGS data for the San Diego Aquaculture monitoring-well cluster that provide a reference set of GWQ data for the sedimentary sequence beneath the valley. Other USGS, DWR, and city GWQ data are then presented, followed by stable isotope values of oxygen and deuterium to show the differences evident in groundwaters from monitoring wells throughout the valley. The GWQ data are subsequently interpreted to demonstrate “freshening” of the Eocene groundwaters reported in the 1965 California Department of Water Resources (DWR) GWQ data; this freshening is accompanied by increased salinity in the MVA itself. We then identify background GWQ conditions prior to the urban development of the valley during the 1960s based on the evidence of geological history and hydrogeochemical analysis. Sources of Information The GWQ in the valley prior to the development of both the MVT and the Qualcomm Stadium in the

1960s can be defined by reference to studies by Ellis and Lee (1919) and DWR. The DWR conducted a series of studies of the groundwater hydrology of the San Diego region during the 1950s and 1960s (DWR, 1959, 1965, 1967), prior to the urbanization of the valley. This information is supplemented with more recent data collected by the City of San Diego and by the USGS San Diego Hydrogeology Project. Ellis and Lee (1919) conducted an early survey of the GWQ in the city’s new well field. The sample collected was a composite sample from the 13 drilled wells of the city’s Mission Valley well field, which had depths ranging from 15 to 30 m (45 to 90 ft) bgs. Figure 10 shows the approximate position of the first 12 of the 13 wells based on records of the San Diego Water Department. Presumably, because just one sample from June 1915 was analyzed (see Table 5), the sample was collected from a manifold at the pumping station that mixed the groundwater from the 13 wells drilled the year before by the city (Fay, 1914).

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 249–275

259

Sengebush, Heagle, and Jackson

Figure 9. Block diagram of Pleistocene river channel with city wells and MVA.

The seven wells for which the data are presented in Table 6 (DWR, 1965) are those along or adjacent to the axis of the Murphy Canyon paleochannel (i.e., its thalweg) and thus constitute a set of data that is useful in reconstructing the ambient or baseline GWQ of the MVA prior to the establishment of the MVT in 1963. Three of the wells, 5D1, 5M1, and 5N1, are in Murphy Canyon itself (see Figure 1), while two others are at the mouth of Murphy Canyon (17D1 and 17D2; see Figure 10). The sixth well, 18Q3, was likely the city’s former well no. 6 in the pre-WWII well field (see Figure 10). A seventh well, 18N1, was outside the MVA and located in an area now built over near Friars Road and is shown in Figure 10. The USGS San Diego Hydrogeology Project began an extensive study of the San Diego region in 2001 (http://ca.water.usgs.gov/sandiego/) that has yielded much useful information on background GWQ in the region (Wright et al., 2005; Wright and Belitz, 2011; and Anders et al., in review). In particular, Anders et al. (2014) has studied GWQ in the Pliocene-age San Diego Formation. The data presented in Table 7 are

Ellis and Lee’s table 46 referred to this sample as K 47 and identified its origin as 13 drilled wells belonging to the City of San Diego located in the “Pueblo lands and Ex Mission of San Diego.” These data are reproduced as Table 5. DWR evaluated the GWQ in the valley prior to the construction of the MVT tank farm and the Qualcomm Stadium in the 1960s but following the abandonment of the city’s well field in the late 1930s. An initial report (DWR, 1959) presented the hydrogeology of the region, including the valley. A subsequent report (DWR, 1965) contained an account of the GWQ in that part of the valley shown in Figure 1, including data (see Table 6) from a number of residential and farm wells that are located on Figure 10. Well construction and screen depths for these wells were not reported. A final report (DWR, 1967) identified the variability in GWQ in both inland and coastal regions. These DWR studies were conducted over the same area “in support of the activities of the San Diego Regional Water Quality Control Board” (DWR, 1967, p. xiii), which exists to this day as the regional regulator.

Table 5. Groundwater quality analyses by the U.S. Geological Survey for a sample from the City of San Diego Mission Valley Aquifer well field (from Ellis and Lee, 1919). (All measurements are given in mg/L.) SiO2 24

260

Na + K

CO3

HCO3

SO4

NO3

TDS

Trace

0.0

151

1.0

394

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 249–275

History and Groundwater Quality of a Coastal Aquifer, San Diego, California Table 6. Groundwater quality data for the era prior to urban development of the Lower San Diego River Valley (DWR, 1965). Wells 5D1, 5M1, and 5N1 are located on Figure 1, while the others are shown on Figure 10. Well Parameter (Unit) Sample date Temperature (uC) pH (pH units) SEC (mS/cm) TDS{ (mg/L) Sodium (mg/L) Potassium (mg/L) Calcium (mg/L) Magnesium (mg/L) Chloride (mg/L) Sulfate (mg/L) Bicarbonate (mg/L) Nitrate-NO3 (mg/L) Boron (mg/L)

5D1

5M1

5N1

17D1

17D2

18N1

18Q3

Feb 1959 n.m. 7.3 1,400 1,039 132 2 149 35 216 288 250 7.6 0.13

May 1960 28 7.5 1,406 975 214 8 54 28 206 111 346 0 0.4

April 1959 n.m. 7.6 1,432 994 230 4 51 20 241 118 315 0 0.08

May 1960 26 7.2 3,160 2,155 363 5 226 68 615 383 432 9 0.6

Feb 1963 n.m. 7.1 n.m. 1,776 n.m. n.m. 189 93 593 360 n.m. 2.0 n.m.

April 1959 n.m. 7.8 2,931 1,944 340 5 122 115 703 154 416 4 0.44

April 1955 22 7.2 1,786 1,105 215 4 103 46 368 146 303 2.5 0.14

n.m. 5 not measured. SEC 5 specific electrical conductance. Total dissolved solids (TDS) by evaporation to 180uC.

{

from the data compilation of Anders et al. (in review). These analyses are considered “complete” in the hydrogeochemical sense in that a full suite of major inorganic ions and some stable and radiogenic isotopes were analyzed. Table 7 presents the results of three samples from the valley. The USGS Aquaculture (SDAQ) well cluster was installed in 2004 on the south side of the river opposite the Qualcomm Stadium (see Figure 1). This monitoring well cluster contains five 2 in. (5 cm) nested piezometers installed within a 17.5 in. (44 cm) borehole. These wells are referred to herein by their state well numbers 16S/ 2W-18J3 through J7 with the shallowest well being J7, which has its 20-ft-long (6 m) well screen set in the base of the MVA at 20 ft (6 m) above mean sea level (amsl) and penetrating the Friars Formation. Thus, SDAQ-J7 provides a mixed sample of MVA and Friars Formation Groundwater, including tritium from the Quaternary sediments and high TDS (1,900 mg/L) from the Friars Formation, as shown in Figure 12. In addition to sampling the SDAQ multi-level well, the USGS analyzed samples from several other wells in Mission Valley, including that of the River Walk Golf Course number 2 well (“RWGC2”), which is further down Mission Valley (see Figure 1). Both RWGC2 and SDAQ J7 have relatively high salinity, with RWGC2 (TDS ,3,579 mg/L) being higher than J7 (TDS ,1,840 mg/L), compared with those shown in Table 8. The proximity of the RWGC2 well to the San Diego River Floodway (0.5 mile, 800 m) and the typical extraction rates of irrigation wells suggest that the high salinity in this well is due to modern seawater intrusion. The Groundwater Quality Monitoring Act (California, 2001) initiated the Groundwater Ambient

Monitoring and Assessment (GAMA) Priority Basin Project “to assess and monitor the quality of groundwater in California. (Wright and Belitz, 2011, p. 2)” Wright and Belitz (2011, p. 1) indicated that the “GAMA San Diego study was designed to provide a statistically robust assessment of untreated-groundwater quality within the primary aquifer systems.” One of the four primary aquifer systems tested is identified as “Alluvial Basins,” which would include aquifer systems such as the San Diego River alluvial basin that contains the MVA. The USGS sampled a total of 17 alluvial basin wells in 2004, including two public watersupply wells in the San Diego River Valley farther upstream from Qualcomm Stadium. The range of measured GWQ parameters in these 17 wells is presented in Table 8. The rationale for including in this article groundwater samples from alluvial wells collected by the USGS outside the valley, i.e., Table 8, but within the San Diego Drainages Hydrogeological Province (Wright and Belitz, 2011), is that they are derived from sediments of similar geochemical nature to the alluvial sediments within the valley and thus are representative of GWQ within the valley. Table 9 presents data from two City of San Diego monitoring wells that were installed and sampled in 2011. These monitoring wells are situated down gradient of the SDAQ multi-level well but in the MVA and were, at the time of sampling, at the leading edge of a plume of contaminated groundwater containing the gasoline additive methyl tertiary-butyl ether (MTBE) and its biodegradation product tertiarybutyl alcohol (TBA). Since these samples were collected, the TBA and TDS concentrations have

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 249–275

261

Sengebush, Heagle, and Jackson

Figure 10. Historic pre-WWII MVA water-supply well field, DWR wells mentioned in text, and recent DB monitoring wells. The TBA plume precisely traces the paleochannel because of its high permeability contrast with the surrounding Friars Formation and the release of gasoline directly into the paleochannel at the MVT (NE corner of figure).

risen in this part of the MVA; thus, no more recent data from these wells are included in this assessment. These five data sets (Tables 5 to 9) contain increasingly large analyte lists from a limited number of water-supply and monitoring wells in the Lower San Diego River Valley. We now apply several methods of hydrogeochemical analysis in order to identify the origin and evolution of the groundwater in the valley. However, no single well has been continually sampled in the 100 year period since the 1915 USGS analysis shown in Table 5. Samples from the MVA prior to its contamination by the MVT gasoline releases of 1987â&#x20AC;&#x201C;1991 are limited to just the 1919 USGS (Table 5) and 1965 DWR (Table 6) data sets. Thus, we will compare data from this valley with that from elsewhere in the San Diego hydrogeologic region as collected, analyzed, and compiled by the USGS (Wright et al., 2005; Wright and

262

Belitz, 2011; Anders et al., 2014; and Anders et al., in review). Stable Water Isotopes Measurements of the stable water isotopes 18O and deuterium (2H) provide information on the origin the groundwaters in the valley. Stable water isotope data in this paper are reported in per mil (%) compared to Vienna Standard Mean Ocean Water (VSMOW), i.e., d18O and d2H, and shown in Figure 13 for July and October 2014. USGS San Diego River sampling locations in the lower valley are shown on Figure 1, while groundwater sampling was conducted with the completion of the cityâ&#x20AC;&#x2122;s monitoring well network shown in Figure 14. The San Diego River water samples (open triangles) represent storm runoff during five events in

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 249â&#x20AC;&#x201C;275

History and Groundwater Quality of a Coastal Aquifer, San Diego, California Table 7. Groundwater quality data of wells in the Lower San Diego River Valley wells sampled by the USGS (from Anders et al., in review). Well Parameter (Unit) Screen elevation (ft a.m.s.l.) Sample date Temperature (uC) pH (pH units) Dissolved oxygen (mg/L) Specific electrical conductance (mS/cm) Total dissolved solids, residue (mg/L) Alkalinity (mg/L CaCO3) Sodium (mg/L) Potassium (mg/L) Calcium (mg/L) Magnesium (mg/L) Iron (mg/L) Manganese (mg/L) Chloride (mg/L) Sulfate (mg/L) Bicarbonate (mg/L) Nitrate-N (mg/L) Boron (mg/L) Arsenic (mg/L) Dissolved organic carbon (mg/L) Oxygen-18, d18O (%) Deuterium, d2H (%) Tritium, 3H (tritium units) SI CaCO3, calcite SI Fe(OH)3, iron hydroxide SI MnO2, pyrolusite Na/Cl ratio (mmol/L) Cl/Br ratio (mmol/L)

SDAQ-J7

RWGC-2

20 Aug 2010 24.5 7.0 (field) 2.1 2,950 (field) 1,930 294 300 2.6 221 71.1 0.524 2.79 742 224 338 ,0.04 0.22 0.0026 4.0 25.2 240 5.8 0.2 ,0.0 ,0.0 0.62 814

20 May 2005 22 7.1 0.5 3,000 1,840 340 301 3.9 219 75 0.92 3.05 631 237 414.4 0.022 0.23 0.00713 No sample 25.58 241.1 5.6 0.33 2.0 210.19 0.74 661

250 to 280 Jan 2004 21 7.1 (field) 0.7 4,600 (field) 2,813 755 644 11.2 303 137 2.43 3.25 1061 477 852 ,0.06 0.359 0.0118 No sample 25.5 237 2.9 0.7 ,0.0 210.47 0.94 644

Note: SI indicates the saturation index of the sample, where 0.0 indicates equilibrium with the mineral, negative values indicate mineral dissolution, and positive values indicate precipitation. SI and bicarbonate values were calculated by PHREEQC (Parkhurst and Appelo, 1999). Estimated Eh 5 +100 mV for all samples. a.m.s.l., above mean sea level.

2004–2010, which was obtained from the USGS National Water Information Database for the three locations shown in Figure 1. These samples fall on or about the global meteoric water line (GMWL), although two plot to the right of the GMWL. Such a displacement is often considered evidence of evaporation prior to sampling (Clark and Fritz, 1997); however, it appears that here it reflects the annual or seasonal variability in the isotopic character of the winter rains (Williams and Rodoni, 1997). The groundwater samples are mainly from city monitoring wells (DB series) and multi-level wells (Einarson and Cherry, 2002) identified as MW-1, MW-2, and MW-3 and shown in Figure 14 as “MVA MW-2 MW-3” or “MVA MW-1” for that well in Murphy Canyon. These are within the MVA paleochannel, except for the deepest sampling port in each case, which was installed in the Friars Formation just beneath the paleochannel gravels; these are identified as “Friars MW-1-2-3.” A few samples are included from the USGS database (Anders et al., in review),

e.g., those from SDAQ, RWGC 2, and the Mission (see Figure 1 for locations). Groundwater stable isotopes from the MVA (diamond-shaped data) fall on or around a “GW correlation” line, in which the most depleted samples (i.e., most negative) are from furthest up Murphy Canyon at multi-level well MW-1 (see Figure 14). The shallow SDAQ J7 sample also plots on this GW correlation line, as do MW-2 and MW-3 samples from the MVA. These samples appear to indicate that the MVA, i.e., both the paleochannel beneath Qualcomm Stadium and that beneath the main channel of the river (DB data), is recharged by discrete storm runoff events producing the unique spatial pattern in the MVA shown in Figure 13. It is also possible that this pattern in some way reflects the infiltration of irrigation at various golf courses above the stadium and the DB site and perhaps infrastructure leaks. More data are required to resolve this uncertainty, because the runoff data and the groundwater data are from different times; consequently, it is difficult to

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 249–275

263

Sengebush, Heagle, and Jackson

Figure 11. The 74 ft (23 m) a.m.s.l. topographic contour showing the marine inundation of the San Diego River Valley, approximately 120,000 years before present. The inset shows the marine isotope stages (d18O) for benthic foraminifers in Hole 893A, Santa Barbara Basin, against age (ka) from Kennett (1995). Substage 5e represents the last interglacial at 120 ka for which the marine invasion is shown in this figure.

define reliable end points for mixing calculations. Nevertheless, a comparison of hydraulic heads in SDAQ J7 with the stage height of the river shows evidence of recharge of the shallow alluvium by winter

storms, which is not the same as recharge of the MVA. The recharge of the MVA appears to occur at discrete times and localities that cannot be identified with our present data and monitoring well locations.

Figure 12. (Left) Variation of TDS (mg/L), specific electrical conductance (mS/cm), and chloride/boron mass ratio for the depth profile of the USGS SDAQ multi-depth monitoring well, August 2010. (Right) Isotopic data for 14C in percent modern carbon (PMC), for tritium in tritium units (TU), and sulfate-sulfur isotope ratio in per mil (%). Both the Cl/B ratio and d34SO4 show trends towards the seawater values of 4,300 and 21%, respectively, with elevation. The uppermost data are for sample SDAQ J7.

264

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 249â&#x20AC;&#x201C;275

History and Groundwater Quality of a Coastal Aquifer, San Diego, California

Those groundwater samples from sampling ports located in the Friars Formation (yellow circles) appear to be a mixture of MVA and deep bedrock groundwaters but have low TDS values (,1,200 mg/L), rather than Friars Formation groundwaters, which exhibited higher salinity from the inundation of the valley. It appears that low TDS groundwater present beneath the Quaternary/Friars contact may discharge upwards into the MVA from the deeper Eocene sedimentary bedrock (see TDS data in Figure 12). Thus, the paleochannels appear to be acting as focused linear discharge areas through which the regional groundwater flow system discharges into the MVA under the artesian conditions noted above. This has resulted in the deepest sampling ports from MW-1, MW-2, and MW-3 plotting midway between the deep SDAQ samples and the DB samples on Figure 13. Impacts of the MVT Gasoline Release on Groundwater Quality An additional complication has been posed by the presence of high TDS concentrations induced by the biodegradation of the very large gasoline leak (,200,000 gallons or ,800 m3) that occurred in 1987–1991 at the MVT. The MVT is situated (see Figure 1) at the neck of Murphy Canyon, which allowed the gasoline to directly penetrate the MVA gravels and for the dissolved phase contamination— principally MTBE, which biodegraded to TBA—to be transported throughout the MVA to the DB monitoring wells. The TBA plume shown in Figure 10 exactly traces the MVA paleochannel due to its very high permeability relative to the Friars Formation. Biodegradation of the gasoline within this plume resulted in an increase in TDS due to the production of protons caused by the hydrolysis of the dissolved carbon dioxide, i.e., CO2 + H2O 5 H2CO3 5 H+ + HCO32. The acid produced attacks mineral surfaces (see Bennett et al., 1993; Borden et al., 1995; and McMahon et al., 1995), thus causing their dissolution and an increase in TDS. Such biodegradationinduced TDS may be identified by the simultaneous occurrence of fuel hydrocarbons in the groundwater sample, e.g., MTBE and TBA. For this reason, we have restricted our analysis of groundwater freshening to samples collected by DWR prior to the construction of the MVT in 1963 and gasoline releases. Freshening Process and Its Effect on the Mission Valley Aquifer We have defined the MVA as the Pleistocene (LGM) paleochannel deposit underlying both the Qualcomm Stadium and the Lower San Diego River

Valley (see section on “Pleistocene Paleochannels”). Because of the cutting and filling of the paleochannel during the Pleistocene lowstand of the sea level, the Eocene sediments, which were inundated by seawater during the last interglacial, now surround the MVA paleochannel throughout its length as shown in Figure 9. The hydrogeochemical consequence of this Pleistocene aquifer (i.e., MVA) being embedded in an Eocene aquitard is that the aquifer acts as a natural hydraulic drain throughout the valley. When the city began to extract groundwater from the MVA in 1914 (Fay, 1914), the natural process of freshening the Eocene sedimentary rock was enhanced through induced seepage to the MVA, causing the MVA to become somewhat brackish while the Eocene sediments underwent freshening. This process of drainage via the MVA and freshening of the Eocene sediments was accelerated by the heavy pumping that occurred during the remediation of the MTBE/TBA plume that migrated from the MVT to the DB monitoring wells shown in Figure 1. Not only was brackish water induced to flow into the MVA by this pumping, but also the groundwater became more brackish (i.e., higher TDS) due to the effects of the bioremediation of the gasoline released from the MVT (see above). Head measurements in SDAQ indicate a strong upward hydraulic gradient across the Friars Formation into the Quaternary sediments. This gradient produces an artesian head, exhibited by the piezometers beneath the Quaternary/Friars contact being ,10–15 ft (3–5 m) above ground surface. Thus, SDAQ monitors hydraulic head and GWQ in a regional groundwater discharge area being recharged in the Peninsular Range to the east. The Friars Formation, a poorly indurated sandstone with 20 percent clay-sized particles, is regarded as an aquitard (K , 1E205 m/s) in the valley. By contrast, the Pleistocene sands and gravels of the MVA have much higher hydraulic conductivities (K . 1E204 m/s). The MVA acts hydraulically as a line sink through the center of the Friars Formation such that Friars’ groundwater has slowly drained into the MVA naturally or has been induced to seep more rapidly by groundwater extraction in the MVA. Depth profiles of TDS, specific electrical conductance, chloride/boron ratios, 14C, tritium and sulfur isotopic values (d34SO4), and specific electrical conductance (SEC) for SDAQ are presented in Figure 12. The non-detect tritium results in the lower four ports (J3 through J6) of the SDAQ piezometers indicate the groundwater was recharged before 1953. The uppermost port, J7, had small amounts of tritium (near 6 TUs), which suggests some groundwater recharge occurred after 1953. The 14C results in the

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 249–275

265

Sengebush, Heagle, and Jackson Table 8. Ranges of results for the parameters determined for the alluvial wells in the San Diego Hydrogeologic Province as part of the GAMA Project (Wright et al., 2005).

Table 9. Groundwater quality data for the city’s DB monitoring wells near the intersection of Interstate routes I-8 and I-805. Well

Groundwater Quality Parameter (Unit) Dissolved oxygen (mg/L) pH (standard units, field measured) Specific conductance (mS/cm @ 25uC, field) Total hardness (mg/L as CaCO3) Alkalinity (mg/L as CaCO3) Nitrate + nitrite (mg/L as N) Dissolved organic carbon (mg/L) Major ions in ppm TDS (residue on evaporation, mg/L) Calcium (mg/L) Magnesium (mg/L) Potassium (mg/L) Sodium (mg/L) Bromide (mg/L) Chloride (mg/L) Sulfate (mg/L) Trace elements in ppb Arsenic (mg/L) Barium (mg/L) Boron (mg/L) Iron (mg/L) Manganese (mg/L) Strontium (mg/L) Uranium (mg/L)

Range in Alluvial Basin Wells 0.1–5.5 6.8–7.5 805–2,787 201–922 133–300 0.04–9.14 0.4–2.1 685–1,800 43–234 22.5–81.6 2.51–9.1 68–295 0.17–1.74 113–540 61.7–421 0.5–2.0 21–144 51–228 4–2,120 0.1–492 409–1,130 0.46–7.91

lower four ports (, 10 pmc) indicates the groundwater is relatively old and has not been recently recharged. We note that SEC, TDS, and the chloride/boron ratio indicate a clear trend in the upper 300 ft (100 m) towards a more saline shallow groundwater at the Friars/Quaternary contact (i.e., J7, 12 m or 40 ft bgs), which is confirmed by a similar trend in the sulfur isotope data towards the seawater d34SO4 value of 21% (Clark and Fritz, 1997, p. 140). Sulfate-sulfur isotope results are reported relative to the Vienna Canyon Diablo Troilite. We associate these features with the marine inundation of 120 ka, which, we propose, produced the brackish groundwaters of the Friars Formation. In this hypothesis, the high TDS at the Friars/Quaternary contact reflects inundation of the Friars Formation by seawater during the last interglacial (see Figure 11) of 120 ka, when there was a sufficiently high head of seawater (specific gravity 5 1.02) to sink through the Quaternary sediments and be trapped at the Friars/Quaternary contact. This trapping is clearly evident in the resistivity and gamma logs from the SDAQ well construction diagram (Figure 9H; Aqua Culture Monitoring Well, http://ca.water.usgs. gov/projects/sandiego/wells/summary.html).

266

Parameter (Unit) Screen elevation (ft amsl) Sample date Temperature (uC) pH (pH units) Dissolved oxygen (mg/L) Specific electrical conductance (mS/cm) Total dissolved solids (mg/L) Alkalinity (mg/L CaCO3) Sodium (mg/L) Potassium (mg/L) Calcium (mg/L) Magnesium (mg/L) Iron (mg/L) Manganese (mg/L) Chloride (mg/L) Sulfate (mg/L) Bicarbonate (mg/L) Nitrate-N (mg/L) Arsenic (mg/L) SI CaCO3, calcite SI FeCO3, siderite SI MnCO3, rhodochrosite SI Fe(OH)3, iron hydroxide SI MnO2, pyrolusite Na/Cl

DB-1

DB-2

+6 to 214 Apr-11 — 7.7 (lab) —

4.5 to 220.5 Jun-11 — 7.1 (lab) —

2,710 (lab) 1,540 416 233 — 160 67.6 4.11 1.66 545 203 472 ,0.05 ,0.002 0.9 1.4 1.2 ,0.0 ,0.0 0.66

2,650 (lab) 1,640 325 263 — 172 62.1 7.83 2.61 555 212 374 0.14 ,0.002 0.3 1.1 0.8 ,0.0 ,0.0 0.73

Note: See Figure 1 for locations. SI indicates the saturation index of the sample, where 0.0 indicates equilibrium with the mineral, negative values indicate mineral dissolution, and positive values indicate precipitation.

In this assessment, we use ionic ratios and concentrations, as well as the stable water isotopes discussed above (Figure 13), to elucidate processes affecting the observed patterns of GWQ. The relationships between bromide and chloride (Davis et al., 1998), as well as sodium and chloride (see Appelo, 1994; Ravenscroft and McArthur, 2004; Andersen et al., 2005; and chapter 6 in Appelo and Postma, 2005), are common tools used to identify freshening of brackish aquifers. Bromide and chloride both have high aqueous solubilities, and their movement in brackish or fresh groundwater is considered to be conservative (Davis et al., 1998). Their relationship (in mg/L) is shown in Figure 15, although the DWR (1965) data cannot be shown because bromide was not analyzed by DWR. The seawater ratio of Cl:Br is 284 (by mass) based on the seawater concentrations shown in Hem (1985). The ratio was developed for the range of Cl shown in the figure. The data from the MVA and the USGS GAMA wells plot on or slightly below the seawater ratio line and suggest there is a seawater Cl:Br component in the groundwater.

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 249–275

History and Groundwater Quality of a Coastal Aquifer, San Diego, California

Figure 13. Stable water isotopes in the Lower San Diego River Valley. The groundwater data from city monitoring wells (DB-2A, MVA at DB, Friars GW, MVA Qualcomm, and MW-1) are from July and October 2014. The USGS SDAQ data are from 2010, while the USGS data from the River Walk Golf Course (RWGC2) and that from the Mission wells are from 2005. River water samples are USGS data from 2004–2010, with one collected in March 2015 at the DB site (identified as DB Mar-15); sampling locations are shown in Figure 1.

Figures 16 through 18 present evidence of groundwater within the Lower San Diego River Valley being freshened over the past 20,000 years. The exception is well 18Q3, which was installed in the MVA (see Figure 10); the paleochannel aquifer acts as the drain for the Eocene sediments and thus becomes increasingly saline over time. Ellis and Lee’s (1919) composite sample is used to represent the freshwater end member, i.e., background conditions. Figure 16 shows the sodium and chloride concentrations for the DWR, SDAQ, RWGC2, and DB samples. A two-component mixing line is shown between seawater from Hem (1985; Na 5 10,500 mg/L, Cl 5 19,000 mg/L) and the Ellis and Lee (1919) sample (Na 5 54 mg/L and Cl 5 85 mg/L). Ravenscroft and McArthur (2004) and Anders et al. (2014) have used a similar graphical technique to identify aquifers that are being either freshened or salinized. Figure 16 shows that 5D1, 18Q3, and 17D1 plot on the mixing line and have a seawater signature. The data points above this line include SDAQ J3, J4, J5, J6, and the Murphy Canyon 5M1 and 5N1 wells, representing brackish water that is being freshened, which has resulted in an increase in the Na concentration relative to Cl. The data points below the line represent groundwater that is being influenced by the

addition of more saline water. RWGC2 shows relatively less evidence of freshening, which is consistent with the likelihood that it is undergoing modern seawater intrusion that comprises about 5 percent of the sample. Those samples close to the freshwater end member, the Ellis and Lee (1919) sample, indicate that freshening is well advanced. 18Q3 reflects saline seepage into the MVA and is thus more saline than would otherwise be expected. Freshening of seawater-inundated sediments is also apparent in Figure 17, which presents evidence of the cation exchange processes that are to be expected, such as the replacement of seawater sodium in the Eocene sedimentary rock by freshwater calcium (Appelo, 1994): 1= Ca2z z Na{X ?1= Ca{X2 z Naz 2 2 where X represents the ion exchanger, such as clay minerals or other oxide surfaces in the sediments, i.e., desorption of Na from marine-inundated sediments by freshwater Ca. The mixing line in Figure 17 is drawn with seawater data from Hem (1985; Ca 5 410 mg/L) and from the Ellis and Lee (1919) results (Ca 5 57 mg/L). Data points that lie below the mixing line are indicative of

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 249–275

267

Sengebush, Heagle, and Jackson

Figure 14. Location of monitoring wells used for stable isotope sampling.

freshwater flushing, and data points that lie above the mixing line are indicative of marine inundation (Ravenscroft and McArthur, 2004) and brackish water flushing. Sample RWGC2, which contains ,5 percent modern seawater, represents a saline end member of the freshening process, while the Ellis and Lee (1919) sample represents the freshwater end member. Very generally, the freshening process of the Eocene sediments progresses from RWGC2 towards 18Q3, which represents saline drainage within the MVA. Similarly, Figure 18 illustrates the reaction involving the desorption of the borate anion in Eocene sediments by freshwater bicarbonate: { HCO{ 3 zB(OH)4 {X ~B(OH)4 zHCO3 {X

As Ravenscroft and McArthur (2004, p. 1428) noted of freshwater flushing of seawater from alluvium in coastal Bangladesh “desorption of B during freshwater flushing occurs in response to lowering of pH and

268

ionic strength, equilibrium re-adjustment, and, possibly, competitive exchange with HCO3/CO3.” They concluded that “enrichment of both Na and B results from desorption from mineral surfaces in response to flushing by fresh groundwater of previously saline aquifers.” A comparison of Figures 16 and 18 shows some similarity in Na and B desorption in the relationship of samples 17D1 and 18N1 to 18Q3. DISCUSSION: BACKGROUND, BASELINE, AND AMBIENT GWQ “Background” and “baseline” are adjectives used to describe the GWQ prior to anthropogenic development that might affect the chemical composition of groundwater. We believe that the following terms are consistent with North American usage: N Background describes the pristine GWQ derived from natural geological, biological, or atmospheric

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 249–275

History and Groundwater Quality of a Coastal Aquifer, San Diego, California

Figure 15. Relationship of chloride to bromide in groundwater samples from the Lower San Diego River Valley (city MWs, SDAQ, RWGC2, and D&B).

sources in the absence of identifiable anthropogenic influences (see Langmuir, 1997, p. 304), whereas N baseline describes the GWQ at the beginning of monitoring and prior to some anticipated event, e.g., “pre-drilling” before hydraulic fracturing (e.g., API, 2009, p. 20; Sloto, 2013).

Baseline GWQ results may include effects of human activities, e.g., coliform bacteria from septic tanks or nitrate from fertilizer applications, although these analytes may or may not be reported. It is noteworthy that European usage of “background” and “baseline” is exactly the opposite of North American usage (see Edmunds and Shand, 2008).

Figure 16. Relationship of chloride to sodium in groundwater samples from the Lower San Diego River Valley from prior to biodegradation-induced elevation of total dissolved solids.

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 249–275

269

Sengebush, Heagle, and Jackson

Figure 17. Calcium and chloride concentrations in the Lower San Diego River Valley prior to biodegradation-induced elevation of total dissolved solids. The solid line is a mixing line between the background sample of Ellis and Lee (1919) and seawater.

In addition, some institutions such as the USGS use a third term: N Ambient GWQ is that GWQ measured at some time and place without any assumption being made as to anthropogenic influences.

The USGS (2013, p. 1) states that the California Groundwater Ambient Monitoring and Assessment (GAMA) Program will not only “establish baseline groundwater quality for comparison with future conditions” but will also “identify emerging constitu-

Figure 18. Evidence of freshwater bicarbonate exchange for seawater boron in the DWR (1965) and SDAQ J7 (2010) data caused by aquifer “freshening,” i.e., displacement of brackish groundwater by freshwater recharge. The simple linear regression indicates that, if RWGC2 is excluded, 81 percent of the variance is explained by this ion-exchange reaction. DB boron values were not available for 2011.

270

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 249–275

History and Groundwater Quality of a Coastal Aquifer, San Diego, California

ents in groundwater.” Thus, “baseline” and “ambient” GWQ can indicate the same GWQ; the first term merely indicates that it is the GWQ measured prior to some event that may affect it. Our purpose is identify the particular meaning of each term with respect to the groundwater samples considered in Tables 5 through 9. The 1915 USGS analysis shown in Table 5 is that of a fresh groundwater with relatively low TDS (,400 mg/L) and trace quantities of iron and 1.0 mg/L of nitrate, indicating an oxygenated groundwater. The common occurrence of dissolved oxygen (DO) and nitrate is discussed by Appelo and Postma (2005, pp. 458–464) and by Langmuir (1997, p. 418); it is also well documented in the technical literature (e.g., Jackson et al., 1990). The data in Table 5 represent what is referred to for regulatory purposes as “background water quality conditions” (San Diego RWQCB, 2005). Tables 6 and 9 represent subsequent sampling of the MVA over approximately 100 years following the USGS sampling event of 1915. Table 6 data were collected during 1955–1965 by DWR, by which time (1959) the city well field shown in Figure 10 had been abandoned and the wells destroyed, although 18Q3 was sampled before abandonment (see DWR, 1959, p. B-20). Table 9 represents data collected during 2011 from the city’s pilot well field (DB-1 and DB-2), which at the time was beginning to show evidence of increased contamination (TDS, MTBE, and TBA) that we associate with the MVT gasoline release of 1987–1991. Therefore, these three sets of data as represented by Tables 5, 6, and 9 provide (1) background GWQ data (i.e., the 1915 USGS sampling in Table 5) and (2) two sets of supplemental analyses representing the evolution of GWQ in the MVA approximately 45 and 96 years later. The background GWQ of the MVA that emerges from these studies suggests that the MVA groundwater in 1915 was rather typical of alluvial aquifers found throughout the U.S. Southwest, i.e., aquifer sediments derived from plutonic rocks producing a sediment rich in feldspars and silica (“felsic”). In his study of the Southwestern alluvial basins, F.N. Robertson (1991, p. C-16) of the USGS commented that “The basin-fill sediments were transported into the basin and deposited under oxidizing conditions.” Robertson’s model identifies groundwater in the recharge area as a calcium-bicarbonate water with pH 5 7.2, DO ranging from 3 to 7 mg/L, and a mean and standard deviation of TDS 5 495 6 68 mg/L. The major geochemical reactions in the recharge areas according to Robertson (1991, p. C86) are: (1) generation of carbonic acid in the recharge area from the dissolution of soil-zone carbon dioxide, including plant respiration, in the recharging groundwater

(CO2 + H2O 5 H2CO3), producing an acidic groundwater (pH , 6); (2) weathering of feldspars and ferromagnesian minerals; (3) dissolution of carbonate minerals; and (4) formation of montmorillonite clay and iron oxides. Table 5, showing the USGS 1915 analysis, is an accurate representation of groundwater acquired through Robertson’s three processes and indicates a TDS , 400 mg/L and the presence of DO. These two parameters—TDS and DO—concisely define the major ion and redox state in the background GWQ of the MVA. Even after WWII, DWR (1967, p. 10) stated in its final report on the San Diego region that “[i]n general, ground water from the continental Pleistocene sediments has a TDS concentration falling within 200 to 600 ppm.” The background GWQ of the Friars Formation is unknown, but 17D1 and 17D2 had TDS , 2,000 mg/L in the 1960s (Table 4). The baseline GWQ in the MVA prior to urban development of the valley in the 1960s is represented by sample 18Q3 in Table 6, which is similar to the USGS GAMA data shown in Table 8. This highquality groundwater existed in the MVA until contamination from the MVT caused its deterioration, although some deterioration may have been caused by agricultural development in the area now occupied by the Qualcomm Stadium, e.g., the low nitrate concentrations in Table 6. Groundwater extraction by the city prior to WWII would have caused brackish water inflow into the MVA from the adjacent Friars Formation and is likely the reason for the increase in TDS from ,400 mg/L in 1915 to ,1,100 mg/L in 1955. In addition, the pavement surrounding Qualcomm Stadium would have decreased the infiltration of DO and low-TDS recharge to the MVA after the mid-1960s. Consequently, urban development of Mission Valley in the 1960s led to further deterioration of the MVA prior to the release of fuels from the MVT. Because knowledge of the MVA used by the city faded from memory after WWII, the perception of relatively high TDS concentrations in the Mission Valley groundwaters developed. This was likely the result of a review of the DWR reports that failed to discriminate between the Pleistocene MVA and Eocene formations, where TDS concentrations ranged up to 3,485 mg/L. The mean and standard deviation of 75 TDS samples reported by DWR (1965) were 1,694 6 723 mg/L, i.e., about the same as in the present city monitoring well DB-2 (Table 9). Of particular interest is that all three samples collected and analyzed by DWR (1959) from the present Qualcomm Stadium area shown in Table 6 (17D1, 17D2, and 18Q3) contained measurable dissolved nitrate. As noted above (Appelo and

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 249–275

271

Sengebush, Heagle, and Jackson

Postma, 2005, pp. 458–464; Langmuir, 1997, p. 418; Jackson et al., 1990), the presence of nitrate indicates the probable presence of DO in these groundwaters at that time prior to urban development. The USGS GAMA data (Table 8) indicate the ambient GWQ in coastal southern California alluvial aquifers that were not affected by the marine invasion during the last interglacial. All six USGS GAMA wells contained some DO, although their mean was only 1.6 mg/L. The 1955 DWR sample from 18Q3 also appears to have contained DO because of the presence of nitrate in the sample. The six Alluvial Basin wells sampled by the USGS during its GAMA survey shown in Table 8 had a mean TDS concentration of only 1,021 mg/L, which is similar to that reported for 18Q3 (1,105 mg/L) in 1955 (DWR, 1965) and displayed in Table 6. The ranges in Table 8 provide guidance as to what might be considered reasonable present-day concentrations within MVA groundwaters. That is, values beyond the ranges reported are likely due to either (1) poor sampling technique, such as incorporation of sub-micron fines, resulting in erroneous concentrations, or (2) contamination, e.g., the migration of the uncaptured plume of MTBE, TBA, and TDS from the MVT gasoline releases. In 2005, the San Diego Regional Water Quality Control Board (San Diego RWQCB, 2005, p. 3) ordered that the gasoline contamination from the gasoline tank farm (MVT; see Figure 1) that was contaminating the MVA should be remediated “to attain background water quality conditions” by the end of 2013. The board defined these background conditions as “the concentrations or measures of constituents or indicator parameters in water or soil that have not been affected by waste constituents/ pollutants from the Site”; i.e., from the MVT. However, it has been long established (Bennett et al., 1993; Borden et al., 1995; and McMahon et al., 1995) that all inorganic constituents are affected by the intrinsic bioremediation such as that which has occurred within that part of the MVA contaminated by the gasoline components. This is because the oxidation of hydrocarbons produces carbon dioxide, which causes a chain of hydrogeochemical processes: It dissolves in water and thus lowers the pH by forming carbonic acid, causing dissolution of aquifer minerals and raising TDS concentrations. Similarly, the dissolved hydrocarbons change the redox state by reducing DO, nitrate, and sulfate and causing the reductive dissolution of iron and manganese oxides on aquifer minerals. Thus, this definition of background GWQ by the regulator is based on a false assumption; i.e., there is no effect of intrinsic biodegradation of organic hydrocarbons on the

272

concentrations of inorganic groundwater constituents, and the Board’s definition of background conditions is therefore incorrect. The attainment of “background water quality conditions” could theoretically be associated with the attainment of the 1915 GWQ measured by the USGS (Ellis and Lee, 1919), i.e., TDS , 400 mg/L and presence of DO, which is the real background GWQ. However, this would likely require the recharge of distilled water to the aquifer but recharge of any water—treated or not—has been continually rejected by the owner of the MVT and by the board in recent years. The attainment of the baseline GWQ, represented by 18Q3 (TDS , 1,100 mg/L, DO present) measured in 1955, prior to the development of the MVT, is however practical given today’s desalinization technology. It should be noted that the DWR (1965, pp. 42–43) advised the San Diego RWQCB that the Eocene sediments of the valley were brackish and that “the most practical way to alleviate the problem of post-nate [i.e., brackish] water seepage is to increase the relative head of ground waters by means of ground water recharge of Mission San Diego Basin” (i.e., the Lower San Diego River Valley). SUMMARY AND CONCLUSIONS In order to explain the current pattern of GWQ in the Lower San Diego Valley, we have reconstructed the likely Quaternary evolution beginning with the last interglacial about 120,000 years ago. Pleistocene sea-level fluctuations have caused periodic incision of Lower San Diego River channels and inundation by rising seas. From water-well data, we have identified channels that developed during the LGM that cut into the Eocene and early Quaternary sediments. This channel probably extended far offshore relative to the present coastline as it graded to a lowered base (sea) level about 17,000 years ago, similar to other major coastal rivers in southern California. These gravels became the principal hydraulic unit in the city’s preWWII groundwater supply—the MVA. We hypothesize that the inundation of the Eocene sediments, such as the Friars Formation, by these shallow seas during the last interglacial is recorded in the brackish GWQ of water wells completed in the valley sediments, with the exception of those wells completed in the gravel paleochannel(s) that date to the LGM, during which the MVA was deposited. However, groundwater extraction from MVA wells has induced brackish groundwater flow into the MVA from the adjacent Friars Formation sediments, and biodegradation-induced natural attenuation of hydrocarbons has further raised TDS concentrations.

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 249–275

History and Groundwater Quality of a Coastal Aquifer, San Diego, California

Several lines of evidence support this hypothesis: N Abbott’s (1999, pp. 202–203) popular history of regional geology illustrates the marine invasion of 120 ka, confirms the geographic extent of the invasion, and outlines the same causes of sea-level rise since the LGM and tectonic uplift over ,120,000 years. N Similar gravel aquifers to the MVA were laid down in all the major southern Californian valleys following the LGM; recent drilling evidence from the MVA has confirmed this. N The USGS 1915 sampling of the GWQ in the MVA showed clearly that the aquifer had low TDS and was oxygenated prior to the beginning of large-scale municipal extraction. N Numerous water-supply wells sampled by DWR in the 1950s, i.e., prior to urbanization of the valley, indicate TDS concentrations varying from 700 to 3,500 mg/L; the lower TDS values were associated with the alluvial gravel aquifer. N DWR also concluded in the 1960s that marine waters explained the brackish GWQ of the Lower San Diego River Valley and that artificial recharge of water was necessary to keep the TDS concentrations low. N The USGS SDAQ (Aquaculture) multi-depth monitoring well in the valley records this brackish water (TDS , 2,000 mg/L) at the contact of the Quaternary and Eocene sediments. We conclude that the confusion concerning the GWQ in the valley is due to a loss of institutional memory of documentation and a failure to re-visit the excellent work of the DWR in the 1950s and 1960s conducted for the present San Diego Regional Water Quality Control Board. While new information is always helpful, it does not negate the bountiful data readily available in the published literature. Restoration of the MVA as a municipal water supply is being planned. Evidence of storm runoff recharge events is apparent in stable water isotope data from new monitoring wells; however, the nature of the recharge process is uncertain because of limited data at present. High residual TDS concentrations will require a significant watertreatment initiative to make MVA groundwaters suitable for public consumption in the area of the pre-WWII city well field. ACKNOWLEDGMENTS We are grateful to Anna Fyodorova, Rob Anders, and Roy Shlemon for their most helpful reviews of the original and revised drafts of this manuscript.

REFERENCES ABBOTT, P. L., 1999, The Rise and Fall of San Diego: 150 Million Years of History Recorded in Sedimentary Rocks: Sunbelt Publications: San Diego, CA, 231 p. ABBOTT, P. L. AND SMITH, T. E., 1989, Sonora, Mexico, source for the Eocene Poway Conglomerate of southern California: Geology, Vol. 17, No. 4, pp. 329–332. ALLEN, J. R. L., 1983, Studies in fluviatile sedimentation: Bars, bar complexes and sandstone sheets (low sinuosity braided streams) in the Brownstones (L. Devonian), Welsh Borders: Sedimentary Geology, Vol. 33, No. 4, pp. 237–293. AMERICAN PETROLEUM INSTITUTE (API), 2009, Hydraulic Fracturing Operations—Well Construction and Integrity Guidelines: API Guidance Document HF-1, First Edition, October 2009, American Petroleum Institute, Washington, D.C., 36 p. ANDERS, R.; DANSKIN, W. R.; AND MENDEZ, G. O., in review Geologic, Hydrologic, and Water Quality Data from Surface-Water and Groundwater Sites in San Diego County, California, 2002–2009. To be published in the USGS Data Series by the US Geological Survey, California Water Science Center, 4165 Spruance Rd., Suite 200, San Diego, CA 92101. ANDERS, R.; MENDEZ, G. O.; FUTA, K.; AND DANSKIN, W. R., 2014, A geochemical approach to determine sources and movement of saline groundwater in a coastal aquifer: Ground Water, Vol. 52, No. 5, pp. 756–768, doi: 10.1111/gwat.12108, 13 p. ANDERSEN, M. S.; NYVANG, V.; JAKOBSEN, R.; AND POSTMA, D., 2005, Geochemical processes and solute transport at the seawater/freshwater interface of a sandy aquifer: Geochimica et Cosmochimica Acta, Vol. 69, No. 16, pp. 3979–3994. APPELO, C. A. J., 1994, Cation and proton exchange, pH variations, and carbonate reactions in a freshening aquifer: Water Resources Research, Vol. 30, No. 10, pp. 2793–2805. APPELO, C. A. J. AND POSTMA, D., 2005, Geochemistry, Groundwater and Pollution, 2nd ed.: A. A. Balkema Publishers, New York, 649 p. BENNETT, P. C.; SIEGEL, D. E.; BAEDECKER, M. J.; AND HULT, M. F., 1993, Crude oil in a shallow sand and gravel aquifer—1: Hydrogeology and inorganic geochemistry: Applied Geochemistry, Vol. 8, No. 6, pp. 529–549. BLAIR, T. C. AND MCPHERSON, J. G., 1994, Alluvial fans and their distinction from rivers based on morphology, hydraulic processes, sedimentary process, and facies assemblages: Journal of Sedimentary Research, Vol. A64, No. 3, pp. 450–489. BORDEN, R.; GOMEZ, C. A.; AND BECKER, M. T., 1995, Geochemical indicators of intrinsic bioremediation: Ground Water, Vol. 33, No. 2, pp. 180–189. BULL, W. B., 1991, Geomorphic Responses to Climatic Change: Oxford University Press, New York, 326 p. CALIFORNIA, 2001, Groundwater Quality Monitoring Act: Water Code sections 10780-10782.3, State of California, Sacramento, CA. CLARK, I. D. AND FRITZ, P., 1997, Environmental Isotopes in Hydrogeology: Lewis Publishers, Boca Raton, FL, 328 p. DANIEL B. STEPHENS AND ASSOCIATES, INC., 2014, Laboratory Report, San Diego MVT: Unpublished consultant report, for INTERA, Inc., January 29, 2014. DAVIS, S. N.; WHITTEMORE, D. O.; AND FABRYKA-MARTIN, J., 1998, Uses of chloride/bromide ratios in studies of potable water: Ground Water, Vol. 36, No. 2, pp. 338–350. DEPANGHER, M., 2014, Petrographic Report #61B: Spectrum Petrographics Inc. Vancouver, Washington, June 22, 5 p. DEPARTMENT OF WATER RESOURCES (DWR), 1959, Ground Water Geology, San Diego River Valley: A Progress Report on

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 249–275

273

Sengebush, Heagle, and Jackson Investigation of Ground Water Conditions in San Diego River Valley, San Diego County: A report to San Diego Regional Water Pollution Control Board (No. 9), Department of Water Resources, State of California, December, 105 p. DWR, 1965, Ground Water Conditions in San Diego River Valley: A report to San Diego Regional Water Pollution Control Board (No. 9), Department of Water Resources, State of California, September, 51 pp. plus 5 appendices. DWR, 1967, Ground Water Occurrence and Quality: San Diego Region, Volume 1: Text: State of California, Department of Water Resources Bulletin 106-2, June, 235 p. EDMUNDS, W. M. AND SHAND, P. (Editors), 2008, Natural Groundwater Quality: Blackwell Publishing, Malden, MA, 469 p. EDWARDS, B. D.; HANSON, R. T.; REICHARD, E. G.; AND JOHNSON, T. A., 2009, Characteristics of southern Californian coastal aquifer systems. In Lee, H. J. and Normark, W. R., (Editors), Earth Science in the Urban Ocean: The Southern California Continental Borderland: The Geological Society of America Special Paper 454, pp. 319–344. EINARSON, M. AND CHERRY, J. A., 2002, A new multilevel ground water monitoring system using multichannel tubing: Ground Water Monitoring & Remediation, Vol. 22, No. 4, pp. 52–65. ELLIS, A. J. AND LEE, C. H., 1919, Geology and Ground Waters of the Western Part of San Diego County, California: U.S. Geological Survey Water-Supply Paper 446, 321 p. FAY, H. F., 1914, Sixth Annual Report of the Superintendent Depart of Water, Pueblo Lands and Forestry, City of San Diego, California: City of San Diego Water Department Archives. GEOFIRMA ENGINEERING LTD. AND INTERA INC., 2011, Hydrogeology of the Mission Valley Aquifer: Revision 0, prepared for the Office of the City Attorney and the Public Utilities Department, The City of San Diego, San Diego, CA, January 25, 2011, 143 p. GEOFIRMA ENGINEERING LTD. AND INTERA INC., 2013, Remediation of MTBE and TBA, Mission Valley Aquifer: Revision 0, prepared for the Office of the City Attorney and the Public Utilities Department, The City of San Diego, San Diego, CA, January 18, 2013, 51 p. HEM, J. D., 1985, Study and Interpretation of the Chemical Characteristics of Natural Water, 3rd ed.: U.S. Geological Survey Water Supply Paper 2254, 263 p. INTERA INC., 2014, Murphy Canyon Monitoring Well Installation and Sampling Report: Revision 1, prepared for the Water Department, City of San Diego, Water Resources and Planning Division, October, Albuquerque, NM, 9 p., 1 figure, 1 table, and 3 appendices. JACKSON, R. E.; MUTCH, J. P.; AND PRIDDLE, M. W., 1990, Persistence of aldicarb residues in the sandstone aquifer of Prince Edward Island, Canada: Journal of Contaminant Hydrology, Vol. 6, No. 1, pp. 21–35. KENNEDY, M. P. AND PETERSON, G. L., 2001, Geology of the San Diego Metropolitan Area, California: Del Mar, La Jolla, Point Loma, La Mesa, Poway, and SW 1/4 Escondido 7 1/2 Minute Quadrangles: California Division of Mines and Geology Bulletin 200, p. 56. KENNETT, J. P., 1995, Latest Quaternary benthic oxygen and carbon isotope stratigraphy: Hole 893A, Santa Barbara Basin, California. In Kennett, J. P.; Baldauf, J. G.; and Lyle, M. (Editors), Proceedings of the Ocean Drilling Program, Scientific Results, Vol. 146 (Pt 2): Ocean Drilling Program, College Station, TX, pp. 3–18.

274

KERN, J. P. AND ROCKWELL, T. K., 1992, Chronology and deformation of Quaternary marine shorelines, San Diego County, California. In Fletcher, C. H. and Wehmiller, J. F. (Editors), Quaternary Coasts of the United States: Marine and Lacustrine Systems: Society for Sedimentary Geology (SEPM) Special Publication 48, pp. 377–382. LANGMUIR, D., 1997, Aqueous Environmental Geochemistry: Prentice Hall, Upper Saddle River, NJ, 600 p. LEE, H. J. AND NORMARK, W. R. (Editors), 2009, Earth Science and the Urban Ocean, The Southern California Continental Borderland: Geological Society of America Special Paper 454, 481 p. MCMAHON, P. B.; VROBLESKY, D. A.; BRADLEY, P. M.; CHAPELLE, F. H.; AND GULLETT, C. D., 1995, Evidence for enhanced mineral dissolution in organic acid-rich shallow ground water: Ground Water, Vol. 33, No. 2, pp. 207–216. MIALL, A. D., 1985, Architectural-element analysis: A new method of facies analysis applied to fluvial deposits: Earth-Science Reviews, Vol. 22, No. 4, pp. 261–308. MIALL, A. D., 1992, Alluvial deposits. In Walker, R. G. and James, N. P. (Editors), Facies Models: Response to Sea Level Change: Geological Association of Canada, St. Johns, Newfoundland, Canada Chapter 7, pp. 119–142. MUHS, D. R.; SIMMONS, K. R.; KENNEDY, G. L.; AND ROCKWELL, T. K., 2002, The last interglacial period on the Pacific Coast of North America: Timing and paleoclimate: Geological Society of America Bulletin, Vol. 114, No. 5, pp. 569–592. PARKHURST, D. L. AND APPELO, C. A. J., 1999, User’s Guide to PHREEQC (Version 2)—A Computer Program for Speciation, Batch-Reaction, One-Dimensional Transport, and Inverse Geochemical Calculations: U.S. Geological Survey WaterResources Investigations Report 99-4259, 312 p. RAVENSCROFT, P. AND MCARTHUR, J. M., 2004, Mechanism of regional enrichment of groundwater by boron: The examples of Bangladesh and Michigan, USA: Applied Geochemistry, Vol. 19, No. 9, pp. 1413–1430. ROBERTSON, F. N., 1991, Geochemistry of Ground Water in Alluvial Basins of Arizona and Adjacent Parts of Nevada, New Mexico, and California: U.S. Geological Survey Professional Paper 1406-C, 90 p. SAN DIEGO REGIONAL WATER QUALITY CONTROL BOARD (RWQCB), 2005, Cleanup and Abatement Order No. 92-01: California Regional Water Quality Control Board, San Diego Region, April 13, Addendum No. 5, 8 p. SHACKLETON, N. J., 1969, The last interglacial in the marine and terrestrial record: Proceedings of the Royal Society of London B, Vol. 174, pp. 135–154. SHLEMON, R. J., 1979, Late Pleistocene channel of the lower Santa Margarita River, San Diego County, California. In Fife, D. L. (Editor), Geological Guide of the San Onofre Nuclear Generating Station and Adjacent Regions of Southern California: Pacific Sections, American Association of Petroleum Geologists, Society of Economic Mineralogists and Paleontologists, and Society of Exploration Geophysicists, Guidebook Number 46, pp. A63–A70. SLOTO, R. A., 2013, Baseline Groundwater Quality from 20 Domestic Wells in Sullivan County, Pennsylvania, 2012: U.S. Geological Survey Scientific Investigations Report 20135085, 27 p. TANAKA, H.; SMITH, T. E.; AND HUANG, C. H., 1984, The Santiago Peak volcanic rocks of the Peninsular Ranges Batholith, southern California: Volcanic rocks associated with coeval gabbros: Bulletin of Volcanology, Vol. 47, No. 1, pp. 153–171. U.S. GEOLOGICAL SURVEY (USGS), 2013, California Groundwater Ambient Monitoring and Assessment (GAMA) Program

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 249–275

History and Groundwater Quality of a Coastal Aquifer, San Diego, California Priority Basin Project: Shallow Aquifer Assessment: Fact Sheet 2012-3136, USGS, Sacramento, CA, February 2013, 2 p. U.S. GEOLOGICAL SURVEY (USGS), 2014, SDAQ Aquaculture Well: Electronic document, available at http://ca.water.usgs.gov/ projects/sandiego/wells/summary.html WILLIAMS, A. E. AND RODONI, D. P., 1997, Regional isotope effects and application to hydrologic investigations in southwestern California: Water Resources Research, Vol. 33, No. 7, pp. 1721–1729.

WRIGHT, M. T. AND BELITZ, K. 2011, Status and Understanding of Groundwater Quality in the San Diego Drainages Hydrogeologic Province, 2004: California GAMA Priority Basin Project: U.S. Geological Survey Scientific Investigations Report 2011-5154, 100 p. WRIGHT, M. T.; BELITZ, K.; AND BURTON, C. A., 2005, California GAMA Program—Ground-Water Quality in the San Diego Drainages Hydrogeologic Province, California, 2004: U.S. Geological Survey Data Series 129, 91 p.

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 249–275

275

The Timing of Susceptibility to Post-Fire Debris Flows in the Western United States JEROME V. DEGRAFF U.S. Department of Agriculture Forest Service, 1600 Tollhouse Road, Clovis, CA 93611

SUSAN H. CANNON U.S. Geological Survey, Box 25046, DFC, MS 966, Denver, CO 80225

JOSEPH E. GARTNER BGC Engineering, 1299 Washington Ave., Suite 280, Golden, Co 80401

Key Terms: Debris Flow, Timing, Wildfire, Forest Cover, Recovery

ABSTRACT Watersheds recently burned by wildfires can be susceptible to debris flow, although little is known about how long this susceptibility persists and how it changes over time. We use a compilation of 75 debrisflow response and fire-ignition dates, vegetation and bedrock class, rainfall regime, and initiation process from throughout the western United States to address these issues. The great majority (85 percent) of debris flows occurred within the first 12 months following wildfire, with 71 percent occurring within the first 6 months. Seven percent of the debris flows occurred between 1 and 1.5 years after a fire, or during the second rainy season to impact an area. Within the first 1.5 years following fires, all but one of the debris flows initiated through runoff-dominated processes, and debris flows occurred in similar proportions in forested and non-forested landscapes. Underlying geologic materials affected how long debris-flow activity persisted, and the timing of debris flows varied within different rainfall regimes. A second, later period of increased debris flow susceptibility between 2.2 and 10 years after fires is indicated by the remaining 8 percent of events, which occurred primarily in forested terrains and initiated largely through landslide processes. The short time period between fire and debris-flow response within the first 1.5 years after ignition and the longer-term response between 2.2 and 10 years after fire demonstrate the necessity of both rapid and long-term reactions by land managers and emergency-response agencies to mitigate hazards from debris flows from recently burned areas in the western United States.

INTRODUCTION There are few vegetative communities throughout the world in which wildfire does not occur periodically. Global fire images taken by the National Aeronautic and Space Administration (NASA) Terra satellite from March 2000 to April 2014 (Earth Observatory website, NASA, 2014) and forest ecology literature (e.g., Dwyer et al., 2000; Lloret and Mari, 2001; Weisberg and Swanson, 2003; and Floyd et al., 2004) illustrate this point. In the United States, vegetative communities affected by large wildfires in recent years include, for example, sagebrush-grasslands (Murphy Complex Fires, Idaho), spruce forest and woodlands (Taylor Complex Fires, Alaska), chaparral (Station Fire, California), Ponderosa pineâ&#x20AC;&#x201C;Douglas fir (Hayman fire, Colorado), and pine plantations and swamp lands (Big Turnaround Complex, Georgia) (NIFC, 2011). A broad spectrum of potential hydrologic responses can be triggered where wildfires burn on steep slopes that are subsequently affected by rainstorms. At the most destructive end of the post-fire runoff and erosion spectrum, debris flows pose a serious threat because they can move rapidly and deliver a significant destructive force along their flow paths (DeGraff et al., 2007; Cannon et al., 2011). Debris flows following a wildfire pose a particular hazard when burned watersheds are adjacent to densely populated areas (Cannon and DeGraff, 2009; Cannon et al., 2009), but they can also affect less populated, rural settings (DeGraff et al., 2011). Since the 1970s, debris flows have been widely recognized as a post-wildfire phenomenon within much of the western United States (Scott, 1971; Wells, 1987; Florsheim et al., 1991; Wohl and Pearthree, 1991; Cannon and Reneau, 2000; Cannon, 2001; Cannon et al., 2001, 2008, 2010; Meyer et al.,

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 277â&#x20AC;&#x201C;292

277

DeGraff, Cannon, and Gartner

2001; Wondzell and King, 2003; Gartner et al., 2005; Giraud and McDonald, 2007; and Wagner et al., 2013). Debris flows have also been documented from burned watersheds in Canada, Australia, Spain, France, the Swiss Alps, and parts of the Mediterranean Basin (Conedera et al., 2003; Jordan and Covert, 2009; Fox, 2011; Nyman et al., 2011; Parise and Cannon, 2012; Santi and Morandi, 2012; and Garc´ia-Ruiz et al., 2013). Debris flows initiated from burned watersheds have been reported to have been triggered by a variety of precipitation conditions including high-intensity rainfall during summer convective storms (DeGraff et al., 2011), long-duration winter and summer frontal storms (Cannon et al., 2008), cells of high-intensity rainfall within frontal storms (Cannon et al., 2011; Kean et al., 2011), rapid snowmelt (Schulz et al., 2006), and intense rainfall on melting snow (Meyer et al., 2001; Shaub, 2001). Moody and Martin (2009) used a synthesis of measured sediment yields following wildland fire (including some debris flows) to demonstrate that the rainfall that results in sediment movement will vary spatially with prevailing rainfall characteristics; measured post-fire sediment yields varied with the seasonality of storm rainfall as well as with the short– recurrence interval rainfall intensities typical of an area. Moody and Martin (2009) developed a map showing areas of similar rainfall conditions, or what they termed rainfall regimes, for the western United States. Debris flows following wildfires have been found to initiate by either progressive bulking of surface runoff with sediment eroded from hillslopes and channels or mobilization of a discrete landslide mass triggered by infiltration processes (Wells, 1987; Cannon et al., 2001; Meyer et al., 2001; Wondzell and King, 2003; and Parise and Cannon, 2012). A lack of discrete landslides bounded by shear surfaces at the heads of the great majority of post-wildfire debris flows indicates the prevalence of surface runoff-erosion processes in their initiation, rather than infiltrationdominated processes. Debris flows initiated through surface runoff processes can be particularly erosive, with material entrained from both hillslopes and channels (Santi et al., 2008; Parise and Cannon, 2012). Although there exists considerable study related to how vegetation recovers following wildfires (e.g., Cerda´ and Doerr, 2005; Lentile et al., 2007), little work has been done to evaluate how a fire-related increased susceptibility to debris flow changes over time. On a broad scale, Meyer et al. (2001) suggested that the timing of post-wildfire debris flows will be driven by the initiation process; runoff-initiated debris flows would be expected in the first few years after

278

a wildfire, while landslide-initiated failures would occur several years after the event, as roots decay and soil shear strengths decrease. With slightly more precision, Cannon et al. (2009) postulated that their emergency assessment of debris-flow hazards from recently burned areas is appropriate for the first 2 years following fires, and Santi and Morandi (2012) found that most post-fire sedimentation in southern California occurs in the first year following fires. However, a more precise understanding of when an increased likelihood of debris flows following wildfires exists, and of how this likelihood changes over time, is important to land managers and emergency responders tasked with minimizing risks to public safety and property (Santi et al., 2011; Moody et al., 2013). In this article we use a compilation of data on the timing of wildfires and post-fire debris-flow events throughout the western United States to examine the temporal distributions of post-fire debris flows in terms of vegetative cover, underlying bedrock, initiation process, and rainfall regime. This work serves to better identify the timeframes during which land managers and emergency-response personnel need to be aware of potential hazards in different settings throughout the western United States. DATA Starting with an existing database of post-fire debris flow events (Gartner et al., 2005), we reviewed published papers and our own field records, accessed the database of historically significant wildfires (NIFC, 2011), and interviewed other researchers to identify events for which both the date of the fire ignition and the date (to the day) of a debris flow– specific response were documented. We were able to build a database of 75 fire-ignition date–debris-flow response pairs from 52 fires in Arizona, California, Colorado, Idaho, Montana, New Mexico, and Utah (Figure 1 and Table 1). The database includes event pairs that we could confidently consider to have occurred as a result of the effects of wildfire and that were not the result of extreme rainfall events that would have triggered debris flows even without fire effects; we incorporated data only in cases in which abundant debris flows were not triggered from adjacent unburned terrain or in which the authors documented a higher density of debris flows within the burned area than within the unburned area. Although many fires had only one triggering event that resulted in debris flows and are thus represented by a single entry, fires in which multiple storms resulted in multiple debris flows are represented by an entry for each storm. For each entry in the database, we documented the dates of fire ignition and debris-flow

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 277–292

Post-Fire Debris Flow Timing

Figure 1. Map showing locations of wildfires from which the timing of debris-flow responses are known. Base is map of rainfall regimes in western United States from Moody and Martin (2009). See Table 2 for primary rainfall characteristics of each rainfall regime.

response and, where available, the vegetation class (forested, non-forested, mixed), bedrock class (granitic, metamorphic, volcanic [including meta-volcanic], sedimentary, mixed), initiation process (infiltration or runoff), and primary rainfall regimes (Moody and Martin, 2009). Table 2 includes the rainfall seasonalities and the range of 2-year recurrence, 30 minuteâ&#x20AC;&#x201C; duration rainfall intensities associated with each of the four primary rainfall regimes identified by Moody and

Martin (2009). Note that although more detailed information on vegetation configuration and underlying materials may be available, it was necessary to use the rather broad classes identified above to develop a consistent record. The data compiled here include primarily opportunistic documentation of post-fire debris-flow events from the literature and observations from our multi-year efforts maintaining rain gauge networks,

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 277â&#x20AC;&#x201C;292

279

280

Harvard Piute Mountain Trail Monument Willow Schultz Peak

Seeley

Aspen

Pickens Grand Prix and Old Wallow Baker South Canyon Mountain Schultz Coal Seam Missionary Ridge Monument Dome

Monument Gladiator Horseshoe 2 Aspen

Seeley

Poppet Unnamed Margarita

4 5 6 7 8 9 10

12 13

7 21 22 11

23 24 25

7 20

14 15 16 2 9 18 19

North Hills Mountain Seeley

Fire Name

1 2 3

Fire ID

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 277â&#x20AC;&#x201C;292

California California California

Utah

Arizona New Mexico Arizona Arizona Arizona Arizona

Arizona California Colorado California Arizona Colorado Colorado

California California

Arizona

Utah

California California California Arizona Arizona Arizona Arizona

Montana California Utah

State

9/23/1997 10/29/1930 9/4/1997

6/26/2012

6/12/2011 5/13/2012 5/8/2011 6/17/2003

6/12/2011 4/25/1996

5/29/2011 10/12/1997 7/11/1994 7/15/2013 6/20/2010 6/8/2002 6/9/2002

11/21/1933 11/14/2003

6/17/2003

6/26/2012

10/4/2005 6/29/2008 10/23/1978 6/12/2011 6/24/2004 6/20/2010 6/10/1988

8/27/1984 7/15/2013 6/26/2012

Fire Ignition Date1

12/1/1997 1/7/1931 11/15/1997

9/1/2012

8/13/2011 7/14/2012 7/11/2011 8/23/2003

8/11/2011 6/26/1996

7/11/2011 12/1/1997 9/1/1994 9/9/2013 8/16/2010 8/5/2002 8/8/2002

12/31/1933 12/25/2003

7/24/2003

7/31/2012

10/17/2005 7/12/2008 11/11/1978 7/10/2011 7/23/2004 7/20/2010 7/11/1988

8/31/1984 7/21/2013 7/7/2012

Debris Flow Date

69 70 72

62 62 64 67

60 62

43 50 52 56 57 58 60

40 41

13 13 19 28 29 30 31

4 6 11

Days

2 2 2

2 2 2 2

2 2

1 2 2 2 2 2 2

1 1

0 0 1 1 1 1 1

0 0 0

Months

Time Between Ignition and Debris Flow

Forested

Non-forested

Non-forested Forested Forested Forested Mixed Forested Forested

Forested Non-forested Forested

Vegetation Class2

Non-forested

Mixed Mixed Mixed Mixed

Mixed Forested

Forested Non-forested Non-forested Non-forested Forested Non-forested Forested

Runoff Non-forested Unknown Non-forested Runoff Forested

Runoff

Runoff Runoff Runoff Runoff

Runoff Runoff

Runoff Runoff Runoff Runoff Runoff Runoff Runoff

Unknown Non-forested Runoff Non-forested

Runoff

Runoff Runoff Runoff Runoff Runoff Runoff Runoff

Runoff Runoff Runoff

Debris Flow Initiation Process

Metamorphic Metamorphic Volcanic

Sedimentary

Mixed Granitic Volcanic Mixed

Mixed Volcanic

Volcanic Sedimentary Sedimentary Granitic Volcanic Sedimentary Volcanic

Mixed Mixed

Granitic

Sedimentary

Metamorphic Mixed Mixed Mixed Volcanic Volcanic Mixed

Metamorphic Granitic Sedimentary

Bedrock Class3

Youberg, 2014 Youberg, 2014 Youberg, 2014 A. Youberg, oral comm., 2014 R. Giraud, oral comm., 2013 Cannon, 2001 Eaton, 1936 Cannon, 2001

Youberg, 2014 Cannon et al., 1997

Youberg, 2014 Cannon, 2001 Cannon et al., 1998 This study Youberg, 2014 Cannon et al., 2003 Cannon et al., 2003

Parrett, 1987 This study R. Giraud, oral comm., 2013 Gardner et al., 2009 DeGraff et al., 2011 Wells, 1987 Youberg, 2014 Pearthree, 2004 Youberg, 2014 Wohl and Pearthree, 1991 R. Giraud, oral comm., 2013 A. Youberg, oral comm., 2014 Eaton, 1936 Brock et al., 2007

Source

Pacific Pacific Pacific

Sub-Pacific

Arizona Arizona Arizona Arizona

Arizona Arizona

Arizona Pacific Plains Pacific Arizona Plains Arizona

Pacific Pacific

Arizona

Sub-Pacific

Pacific Pacific Pacific Arizona Arizona Arizona Arizona

Plains Pacific Sub-Pacific

Seasonal Rainfall Regime (Moody and Martin, 2009)

Table 1. Data used to evaluate timing of post-wildfire debris flow occurrence, including identification number, fire name, state, date of fire ignition, date of debris-flow response, number of days and months between ignition and response, debris-flow initiation process, vegetation class, bedrock class, source of information, and seasonal rainfall regime (from Moody and Martin, 2009).

DeGraff, Cannon, and Gartner

Station Station Station Motor Station Rim Hopper Logan Unnamed Middle Arch Rock Overland Farmington

Bear Yellowstone Unnamed Yellowstone Inyo Complex Mollie

31 31 31 44 31 49 21 33 34 37 36 38 39

40 48 41 48 42 43

29 22 30 31 31 17 20

47 19

27 19

Montana Montana California Montana California Utah

California Arizona California California California California New Mexico California California California California California California California California California California California Colorado Utah

California Colorado

Sayre Missionary Ridge Sayre Missionary Ridge Hemlock Missionary Ridge Pauba Horseshoe 2 Molera Station Station Johnston Peak Dome

27 19

State

Cerro Grande New Mexico Bullock Arizona

Fire Name

Fire ID

Table 1. Continued.

9/1/2000 8/20/1988 12/3/1927 8/20/1988 7/15/2007 9/1/2001

8/26/2009 8/26/2009 8/26/2009 8/25/2011 8/26/2009 8/17/2013 8/5/1997 8/4/1997 6/16/1968 7/24/1977 8/7/1990 10/29/2003 7/15/2003

8/31/1997 5/8/2011 8/1/1972 8/26/2009 8/26/2009 7/20/1960 4/25/1996

7/5/1997 6/9/2002

11/20/2008 6/9/2002

5/20/2002

5/4/2000

Fire Ignition Date1

7/15/2001 7/9/1989 11/13/1928 8/10/1989 7/12/2008 9/12/2002

1/17/2010 1/19/2010 1/20/2010 1/21/2012 2/5/2010 2/26/2014 2/15/1998 2/15/1998 1/1/1969 2/8/1978 3/3/1991 6/24/2004 4/6/2004

12/1/1997 8/13/2011 11/12/1972 12/10/2009 12/11/2009 11/5/1960 8/19/1996

10/2/1997 9/7/2002

2/14/2009 9/5/2002

2/5/2009 8/29/2002

8/5/2002

7/16/2000

Debris Flow Date

317 323 346 355 363 376

144 146 147 149 163 193 194 195 199 199 208 239 266

92 97 103 106 107 108 116

89 90

86 88

77 81

Days

11 11 12 12 12 13

5 5 5 5 5 6 6 7 7 7 7 8 9

3 3 3 4 4 4 4

3 3

Months

Time Between Ignition and Debris Flow

Runoff Runoff Unknown Runoff Runoff Runoff

Runoff Runoff Runoff Runoff Runoff Runoff Runoff Runoff Landslide Runoff Runoff Runoff Runoff

Runoff Runoff Runoff Runoff Runoff Runoff Runoff

Runoff Runoff

Runoff

Debris Flow Initiation Process

Non-forested Forested Non-forested Forested Non-forested Forested

Non-forested Non-forested Non-forested Non-forested Non-forested Forested Non-forested Non-forested Non-forested Non-forested Non-forested Forested Forested

Non-forested Mixed Non-forested Non-forested Non-forested Forested Forested

Non-forested Forested

Forested

Vegetation Class2

Metamorphic Volcanic Mixed Volcanic Metamorphic Metamorphic

Mixed Mixed Mixed Metamorphic Mixed Granitic Sedimentary Sedimentary Sedimentary Sedimentary Mixed Granitic Metamorphic

Metamorphic Volcanic Metamorphic Mixed Mixed Volcanic Metamorphic

Sedimentary Granitic

Sedimentary Sedimentary

Sedimentary Metamorphic

Granitic

Volcanic

Bedrock Class3

Cannon et al., 2011 Cannon et al., 2011 Cannon et al., 2011 This study Cannon et al., 2011 This study Cannon, 2001 Cannon, 2001 Morton, 1989 Wells, 1987 DeGraff, 1994 Gardner et al., 2008 Giraud and McDonald, 2004 Parrett et al., 2003 Meyer and Wells, 1997 Eaton, 1936 Meyer and Wells, 1997 DeGraff et al., 2011 McDonald and Giraud, 2002

Cannon, 2001 Youberg, 2014 Cleveland, 1973 Cannon et al., 2011 Cannon et al., 2011 Doehring, 1968 Cannon et al., 1997

Cannon, 2001 Cannon et al., 2003

This study Cannon et al., 2003

A. Youberg, oral comm., 2014 This study Cannon et al., 2003

Cannon et al., 2001

Source

Plains Plains Pacific Plains Sub-Pacific Sub-Pacific

Pacific Pacific Pacific Pacific Pacific Pacific Pacific Pacific Pacific Pacific Sub-Pacific Plains Sub-Pacific

Pacific Arizona Pacific Pacific Pacific Pacific Arizona

Pacific Arizona

Arizona

Seasonal Rainfall Regime (Moody and Martin, 2009)

Post-Fire Debris Flow Timing

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 277â&#x20AC;&#x201C;292

281

282

Motor Gaviota

Williams

Clover Mill Creek Missionary Ridge Lowman Stanislaus Complex Overland

44 45

46 35 19

Colorado

Idaho California

California California Colorado

California

California California

Utah

State

10/12/2003

7/1/1989 8/1/1987

5/28/2008 11/2/1975 6/9/2002

9/2/2012

8/25/2011 6/12/2004

6/26/2012

Fire Ignition Date1

820 831 1,034

544

450 515

385

Days

9/11/2013

3,622

12/20/1996 2,729 1/2/1997 3,442

8/26/2010 2/10/1978 4/8/2005

2/28/2014

11/17/2012 11/9/2005

7/16/2013

Debris Flow Date

121

91 115

27 28 34

15 17

Months

Time Between Ignition and Debris Flow

Non-forested

Non-forested Non-forested

Non-forested

Vegetation Class2

Landslide Forested

Landslide Forested Landslide Forested

Runoff Non-forested Unknown Non-forested Landslide Forested

Runoff

Runoff Runoff

Runoff

Debris Flow Initiation Process

Granitic

Volcanic Granitic

Granitic Metamorphic Sedimentary

Metamorphic

Metamorphic Granitic

Sedimentary

Bedrock Class3

Coe et al., 2014

Meyer et al., 2001 This study

R. Giraud, oral comm., 2013 This study Santa Barbara County Public Works Dept, 2005 D. Staley, oral comm., 2015 Wagner et al., 2013 Bruington, 1982 Schulz et al., 2006

Source

Plains

Sub-Pacific Pacific

Sub-Pacific Pacific Arizona

Pacific

Pacific Pacific

Sub-Pacific

Seasonal Rainfall Regime (Moody and Martin, 2009)

Date of fire ignition was used as baseline, rather than date of fire containment, because containment date is not always consistently identified. Forested vegetation classes include those reported as conifer, spruce, fir, pine, oak, or combinations thereof; non-forested includes vegetation reported as chaparral, scrub oak, oak brush, and oak sage; and mixed includes both forested and non-forested terrains. 3 Bedrock classes include volcanic and meta-volcanic, metamorphic, sedimentary, granitic, and mixed.

39 50

Seeley

Fire Name

Fire ID

Table 1. Continued.

DeGraff, Cannon, and Gartner

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 277â&#x20AC;&#x201C;292

Post-Fire Debris Flow Timing Table 2. Rainfall characteristics of rainfall regimes, as defined by Moody and Martin (2009). 2-Year Recurrence, 30-Minute Duration Rainfall Intensity (mm/hr) Rainfall Regime

Rainfall Seasonality

Pacific

Winter maximum Summer minimum

Arizona

Winter and summer wet Spring dry Fall moist Winter wet Spring moist Summer and fall dry Winter minimum Summer maximum

Sub-Pacific

Plains

Rainfall Intensity Classes

Lower

Upper

High Medium Low Extreme High Medium Low

.36 .20 .15 .52 .36 .20 .10

52 36 20 100 52 36 20

Extreme High Medium

.52 .36 .19

100 52 36

operating a warning system for post-fire debris flows in southern California with the National Weather Service, and documenting the effects of debris flows within recently burned areas throughout the western United States. We cannot be certain that the database includes all events from a given burned area, particularly those events reported in the literature. However, we are more confident that most debris flows were documented during our own efforts. APPROACH AND METHODS To first identify the time period following fires during which an increased likelihood of debris flow exists, we examine frequency distributions of our 75entry database, providing information on the timing of debris flows relative to initiation mechanism, vegetative cover, and underlying bedrock. We also examine the timing of debris flows within each of the four primary rainfall regimes defined by Moody and Martin (2009) to account for the spatial variability of prevailing rainfall conditions in different settings in the western United States. We also consider the relative timing of fire ignitions, storm rainfall, and debris-flow occurrence to identify the seasons and months during which most debris flows can be expected in each rainfall regime. To further characterize how the increased susceptibility to debris flows following wildfires may change over time within areas with similar rainfall conditions, we calculate cumulative probabilities of a day with debris flows for each of the primary rainfall regimes defined by Moody and Martin (2009). We calculated probabilities as the ratio of the number of events divided by the number of possible results (Helsel and Hirsch, 2002) or, in this case, the number of days during which debris flows occurred within each rainfall regime area divided by the number of days

leading up to, and including, the debris flows. For example, if debris flows are first triggered on day 4 following fire ignition, this day is characterized by a 1/4, or 25 percent, probability of a day with debris flows. On a cumulative basis, then, if after 100 days there have been a total of 10 days with debris flows within a given rainfall regime area, there will be a 10/100, or 10 percent, probability of a day with debris flows for that area. Cumulative probabilities are calculated for each rainfall regime area for each day in which debris flows were documented, starting with the first day with debris flows following ignition, through the first 18 months following ignition or less, depending on the length of the record. If more than one entry exists for a given day, the duplicate values are not included in the cumulative calculation because we are evaluating the number of days with debris flows, and not the total number of debris flows. A best-fit analysis identifies the functions that best characterize the cumulative probability functions for each rainfall regime. Note that the probabilities calculated here simply characterize the relative chances of a given day experiencing debris flows based on the numbers of days with debris flows within each rainfall regime area. The probabilities do not describe the immediate post-fire debris-flow susceptibility of individual drainage basins based on a given set of physical and rainfall conditions, as addressed by Cannon et al. (2009). RESULTS Debris-Flow Timing The first debris flows from a given fire are documented as occurring as soon as 4 days after fire ignition (Table 1). The great majority (85 percent) of debris flows occurred within the first 12 months following wildfire, with 71 percent occurring within

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 277â&#x20AC;&#x201C;292

283

DeGraff, Cannon, and Gartner

Figure 2. Histograms showing number of months between fire ignition and debris-flow occurrence by (a) vegetation class and initiation process, (b) bedrock class, and (c) rainfall regime. Data timeframe is from 0 to 10 years.

the first 6 months (Figure 2a). The importance of the first year following fire is in keeping with the findings of Santi and Morandi (2012) in southern California. Seven percent of the debris flows occurred between 1 and 1.5 years after a fire. Following 18 months, there is an 8-month period over which we did not identify any fire ignition dateâ&#x20AC;&#x201C;debris-flow response pairs. The final 8 percent of events occurred between 2.2 and 10 years (26 to 120 months on Figure 2a) after the fire.

Of the six events that occurred between 2.2 to 10 years following wildfire, four were reported to have initiated through landslide processes, one through runoff-dominated processes, and the initiation process of one was unknown. Although the sample number is small, these proportions may suggest the importance of infiltration-dominated processes in the triggering of debris flows over these longer timeframes.

Initiation Processes and Debris-Flow Timing

Vegetation Class, Initiation Process, and DebrisFlow Timing

All but 1 of the 69 fire-debris flow pairs documented in the first 18 months following a fire were reported to have been triggered by runoff-dominated processes (Figure 2a and Table 1), indicating the prevalence of such processes within this timeframe.

284

In the first 18 months following a wildfire, the similar proportions of number of debris flows in forested and non-forested terrains (Figure 2a) suggests that vegetation characterized by these broad

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 277â&#x20AC;&#x201C;292

Post-Fire Debris Flow Timing

Figure 3. Monthly occurrences of fire ignitions and debris flows in the Pacific, Sub-Pacific, Arizona, and Plains rainfall regimes.

classes exerts little influence on post-wildfire debrisflow susceptibility within this timeframe. Over longer timeframes (2.2 to 10 years), four landslide-initiated events occurred in forested terrains, one runofftriggered event occurred in non-forested terrain, and the initiation mechanism of one event is unknown. Again, although the sample size is small, the proportions indicate the possibility that over long timeframes in forested terrains, post-fire debris-flow occurrences may be dominated by landslide processes. Bedrock Class and Debris-Flow Timing Our data show that the type of bedrock underlying burned areas may influence how long debris flows will continue to occur after fires (Figure 2b). Debris flows were triggered in areas underlain by granites and metamorphic rocks through the first 18 months following ignition, while those underlain by sedimentary materials produced debris flows through 14

months and those underlain by volcanic materials through 12 months, with most of these occurring in the first 4 months. Each of the four bedrock classes is represented in the longer timeframes (2.2 to 10 years), but no clear effect of bedrock on debris-flow occurrence is apparent. Rainfall Regime, Fire Ignition, and DebrisFlow Timing The timing of debris flows varied within the four rainfall regimes during the first 18 months following fire ignition (Figure 2c), seasonally (Figure 3) and intra-seasonally (Figure 4). Note that the probability of a day with debris flows (as shown in Figure 4) cannot be calculated until debris flows actually occur; probabilities before the first occurrence are not known. Within the area of the Pacific rainfall regime, fires started during a 7-month period from June through

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 277â&#x20AC;&#x201C;292

285

DeGraff, Cannon, and Gartner

Figure 4. Probability of a day with debris flow over time for the Pacific, Sub-Pacific, Arizona, and Plains rainfall regimes through the first 18 months following ignition or less, depending on length of record. Day 0 is the first fire ignition in each rainfall regime area, and data start at the day of the first documented debris flow. Best-fit lines to the probability functions and their equations are shown as colored lines, with x as time and y as the probability of a day with debris flows.

December, and most debris flows were triggered in November through February (Figure 3). Debris flows were documented as soon as 6 days after fire ignition (Table 1). Although most debris flows occurred within the first 6 months following ignition, they continued to be triggered through an 18-month period, with some occurring during the second winter after a fire (Figures 2c and 4). The debris flows produced during the second winter are the only events documented from the areas burned by the Motor (44), Gaviota (45), and Unnamed (41) fires. Debris flows were also reported in July and August, presumably in response to summer thunderstorms, and the remainder occurred during the winter months, when rain is reported to fall at a range of low to high intensities (Figure 3 and Table 2). The cumulative probability of a day with debris flow (Figure 4) shows a gradual exponential decrease over the 18-month period of record, but with periods of increasing probability that reflect approximately week-long periods of increased debris-flow activity superimposed on the general

286

declining trend as sequential storms move through the area. Within the area of the Arizona rainfall regime, fires ignited in April, May, and June, and most debris flows followed in July and August (Figure 3), the occurrence of which corresponds with the monsoon season, which is characterized by medium- to extreme-intensity rainfall (Table 2). Debris flows were not reported until 28 days following ignition (Table 1), but all of the documented events occurred within 4 months of the fire, with the great majority occurring in the first 2 months (Figure 2c). The cumulative probability of a day with debris flow calculation for the Arizona rainfall regime area is distinct from the three other rainfall regimes (Figure 4). No debris-flow activity was documented until 28 days after the onset of ignitions, resulting in a low initial probability of a day with debris flows relative to the other rainfall regimes. Once storms capable of initiating debris flows affected susceptible areas, however, the probability of a day with debris

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 277â&#x20AC;&#x201C;292

Post-Fire Debris Flow Timing

flow increased rapidly over a little more than a month’s time. The two nearly vertical segments on Figure 4 for the Arizona rainfall regime are the result of 1- to 2-week periods over which debris flows occur on a near-daily basis, and the two more horizontal segments reflect slight decreases in the frequency of debris-flow events. The probability of a day with debris flow starts to decline after more than 2 months of debris-flow activity, presumably as the monsoon season wanes. Within the area of the Sub-Pacific rainfall regime, fires were ignited over a 4-month period from May through September (Figure 3), and the first debris flow occurred 11 days following ignition (Table 1). Even though the area is typified by low-intensity rainfall (Table 2), which is not generally considered sufficient to generate debris flows, debris flows were documented in this area. Debris flows were triggered up to 13 months after ignition (Figure 2c), during March and April, in July and September within days of fire ignition as well as a year later, and in August and December (Figure 3). The three debris flows that occurred more than a year after fire ignition were produced from the Seeley (3), Inyo (42), and Mollie (43) fires. This was the fourth debris flow produced from the area burned by the Seeley fire and the first from the Inyo and the Mollie fire areas. Within the Sub-Pacific rainfall regime area, the probability of a day with debris flow decreases as a power law, with a higher rate at the beginning that slows over time but that extends into the year following ignition (Figure 4). Within the area of the Plains rainfall regime, fires started during a 5-month period from June through October, and debris flows were triggered from June through September (Figure 3), a period typified by summer rainstorms ranging in intensity from medium to extreme (Table 2). Debris flows first occurred only 4 days following fire ignition (Table 1), and activity continued for an additional 12 months (Figure 2c). The debris flow produced nearly a year after ignition was generated from the area burned by the Yellowstone fire (48) and was the second debris flow reported from this area. The probability of a day with debris flow decreases as a power law similar to that of the Sub-Pacific rainfall regime area, the difference being that debris flows start as soon as 4 days after fire ignition, rather than after the 11 days reported in the Sub-Pacific area. DISCUSSION The data examined here indicate two periods of increased debris flow susceptibility in burned drainage basins—the first immediately following the fire

and lasting up through 18 months and the later, second period beginning after 2.2 years and extending up to 10 years. Debris flows are considerably more frequent during the initial period of susceptibility, occur with similar proportions in non-forested and forested terrains, and initiate primarily through the process of progressive entrainment of material eroded from hillslopes and channels by surface runoff. Those few debris flows that occurred during the second period of susceptibility are most common in forested terrains and for the most part occur through mobilization of discrete landslide masses. The time interval of 18 months to 2.2 years separating the two periods of increased debris-flow susceptibility in burned drainage basins is seen as representing recovery from fire-induced conditions, favoring erosion by the process of progressive entrainment of material. Wildfire can change the hydrologic response of a drainage basin by (1) removing soil-mantling vegetation and litter, (2) depositing ash, (3) altering of the physical properties of soil and rock, and (4) enhancing, generating, or destroying water-repellent soils (Parise and Cannon, 2012). In general, these changes reduce infiltration of water and make the burned surface soils more prone to entrainment by overland flow. In addition, Moody and Ebel (2012) found that hyper-dry soil conditions affecting infiltration are an important factor responsible for the extreme runoff events that frequently occur during the first rainstorm after a wildfire. Natural restoration of soil-mantling vegetation is a key factor in reducing the amount of erosion by overland flow. Vegetation regrowth can occur quickly, regardless of burn severity and ecosystem, based on a study of post-fire vegetation burn severity at eight large western U.S. wildfires (Lentile et al., 2007). Recovery plant cover was found to be predominantly grasses in chaparral ecosystems, forbs in mixed conifer forests, and shrubs in boreal forests. The composition of post-fire vegetation was influenced by the presence of burned vegetation types capable of resprouting and the condition of ungerminated seeds in the soil (Lentile et al., 2007). Re-vegetation by herbs and shrubs in burned areas under the influence of Mediterranean climatic conditions reduced erosion to negligible levels within 2 years after the wildfire (Cerda´ and Doerr, 2005). This finding is consistent with the debris flow susceptibility due to progressive entrainment of material effectively ceasing by 18 months. Similarly, the second period of debris-flow susceptibility due to infiltration-triggered landslides starting at 2.2 years and persisting for up to 10 years in forested drainage basins is consistent with research on the influence of tree root decay on shallow soil

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 277–292

287

DeGraff, Cannon, and Gartner

movement. Two years coincides with the earliest time period at which tree root decay would reach a level that significantly weakens the strength of the soil mass (Sidle, 2005; Sidle and Ochai, 2006), and Schulz et al. (2006) described both a loss of root strength and a decline in evapotranspiration rates as contributing to landslides generated from an area that had burned 3 years prior. In addition, the senior author has examined a number of debris flows initiated by infiltration-type landslides in both burned and unburned forested watersheds in the southern Sierra Nevada (DeGraff, 1994). In failures from unburned forested slopes, there are commonly numerous intact, flexible roots dangling within the exposed slide plane and its margins. These are remnants of what were live roots placed in tension during initial landslide movement and broken off by the moving mass. In contrast, roots exposed in landslide-initiated debrisflow scars in burned areas were generally sheared close to or at the slide plane and were noticeably dry and brittle, indicating a decline in strength attributable to fire-induced mortality and decay (DeGraff, 1997). Understanding the timing of debris flows following wildfire has implications for successful protection of life and property. The current attention and research devoted to debris flows that occur in the few months to first year following a wildfire is well placed. The threat is widespread and largely independent of the type of vegetative community burned. The immediacy of this threat calls for prompt implementation of mitigation measures. The design and placement of these measures needs to be effective for debris flows resulting from progressive entrainment of material eroded from hillslopes and channels by surface runoff. The likelihood of shortâ&#x20AC;&#x201C;recurrence interval storms with periods of high-intensity rainfall will need to be determined as part of this effort. When wildfire burns a forested area, the mitigation of any imminent debris-flow threat must also be coupled with similarly prompt action to limit the impact of later debris flows. The later debris-flow threat in burned forest area provides time for mitigation within the debris-flow source area. Countering the decreasing root strength provided by firekilled trees with the increasing strength of re-planted trees requires action within 1 to 2 years of the wildfire. Because the later threat of debris flows in forested drainage basins is associated with infiltration-type landslides, there is sufficient time to prioritize reforestation to basins in which the soil burn severity is high, tree mortality is great, and infiltration-type landslide activity has resulted in debris-flow activity in the past. Some mitigation measures implemented to address immediate life and property concerns for debris flows due to progressive bulking might be

288

maintained for later mitigation of infiltration-type debris flows in forested basins. In this study we found that the cumulative probability of a day with debris flows varies between the Pacific and Arizona rainfall regimes, but the responses of the Sub-Pacific and Plains rainfall regime areas were remarkably similar to each other, following similar power laws. The similarities and differences may reflect the timing of fire ignitions and rainfall and rainfall conditions in each area. In the Sub-Pacific area, most rain is reported to fall in the spring and winter months and typically at relatively low intensities, although most debris flows were documented as occurring in the summer and fall. In contrast, within the Plains rainfall regime, most rain is reported to fall in summer at intensities that can range from medium to extreme, most fires occur in the summer, and most debris flows occur in the summer. The similarities in the timing of the fires and debris flows but contrasts in the described rainfall conditions could indicate that the rainfall characterization for the Sub-Pacific rainfall regime is missing the important effect of summer and fall convective storms, which could produce rainfall at high intensities. The trend of the calculated probabilities of a day with debris flows within the Arizona rainfall regime differs from that seen in other rainfall regimes, with an overall increase in probabilities that starts with the onset of the monsoon season and starts to decrease only after the season ends. This trend may be due to extreme rainfall intensities throughout the monsoon season that trigger debris flows on a neardaily basis, but which may also hinder vegetative and soil property recovery at the same time. Within the Pacific rainfall regime, the prolonged and nearly linear decline in probabilities of a day with debris flows superimposed with approximately week-long periods of increased probabilities can perhaps be attributed to the fact that most fires are nearly immediately followed by repeated bands of debrisflow triggering winter storms, but any vegetative recovery that occurs in the wet winter months will slow during the sequential dry springs and summers. Within the Pacific, Plains, and Sub-Pacific rainfall regimes, the asymptotic decline in the probability of a day with debris flows after about 8 months may be considered to indicate a background susceptibility to post-fire debris flows in these areas, given no additional fires in the area. Note, however, that once a new fire burns in debris-flowâ&#x20AC;&#x201C;susceptible terrain in either of these two rainfall regime areas, one would expect to re-set and return to the high end of each of the curves. That these curves do not return to zero over time indicates that a slight susceptibility to post-fire debris flows remains after the initial

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 277â&#x20AC;&#x201C;292

Post-Fire Debris Flow Timing

18-month period for which post-fire debris flows were documented in this study. The rainfall regimes used here to characterize postfire debris-flow timing and susceptibility are broad spatial categories. Moody and Martin (2009) also define secondary categories, but the spatial distribution of our data was not sufficiently dense to warrant evaluation for each of these secondary classes. It is also possible that some of the variability in storm conditions specific to the initiation of post-fire debris flows is not completely captured by the divisions used here. As discussed above, characterization of the SubPacific rainfall regime lacks the high-intensity summer thunderstorm activity that produces debris flows in this area. In addition, both the Pacific Northwest of the western United States and southern California are included in the Pacific rainfall regime, while the post-fire debris-flow response can be quite different between these two areas. Although runoff-initiated debris flows are the norm in southern California, we have not observed evidence of this process in the Pacific Northwest, and this difference could be attributed to differing prevailing rainfall conditions. And finally, although summers are considered to be dry within the Pacific rainfall regime, we have records of debris flows being triggered during this season. Definition of additional rainfall regimes may be necessary to capture this variability and better characterize conditions that lead, specifically, to post-fire debris flows. Localized weather will be a primary source of variability in the timing of debris flows and of some of the uncertainty associated with the findings of this study. For example, fewer storms capable of triggering debris flows during drought periods, or storms lacking periods of intense rainfall, will prolong the time between burning and when debris flows occur. We see that within the Pacific and Sub-Pacific rainfall regime areas, some of the first known debris flows from given burned areas were produced during the second rainy season to affect the area, indicating that the initial rainy season was perhaps lacking in sufficient rain to trigger debris flows. Additional data and more complex analyses will be necessary to characterize this variability. Additionally, although we were not able to locate records of post-fire debris flows beyond 18 months and before 2.3 years following fire ignition, this does not mean that they would not occur during this time period; the probability curves developed in this study do not return to zero over time, indicating a persistence of post-fire debris-flow susceptibility beyond the initial 18-month period. However, because 2 years coincides with the earliest time at which tree root decay would reach a level that significantly weakens

the strength of the soil mass, we would expect debris flows that might occur during this period to initiate through the process of progressive entrainment of material from hillslopes and channels by runoff, rather than through landslide initiation. Finally, the data used in this study do not reflect the results of a spatially extensive, long-term, instrumental monitoring of the hydrologic response of burned areas, which could conceivably produce a more complete representation of all debris-flow activity in a given burned area. In addition, the calculated changes in the probability of a day with debris flows do not reflect changes in the expected volumes of potential debris flows, an important element in assessing hazards and risk. We thus suggest that this analysis presents a first look at changes in post-fire debris-flow susceptibility that can be refined and improved with additional data. SUMMARY AND CONCLUSIONS In summary, both forested and non-forested landscapes are likely to experience an immediate increased susceptibility to debris-flow occurrence in drainage basins recently burned by wildfires. This period of greater debris flow susceptibility can last up to 18 months but will vary depending on underlying bedrock materials and prevailing climatic conditions or rainfall regimes. The great majority of debris flows that occur during this period initiate through progressive entrainment of material eroded from hillslopes and channels by surface runoff. The data in this study indicate that debris flows will occur in areas underlain by granites and metamorphic rocks through the first 18 months following wildfires, while those underlain by sedimentary materials can produce debris flows for up to 15 months, and those underlain by volcanic materials for up to 12 months. Within the area included in the Pacific rainfall regime, debris flows can be expected just 6 days after ignition, and although the great majority of events will occur in the first 6 months, they can continue to be triggered through an 18-month period, with some occurring during the second winter after a fire. Although most debris flows will be triggered during the months of November through February in response to winter rainfall, some can occur in July and August, presumably in response to summer thunderstorms. The probability of a day with debris flows in the area of the Pacific rainfall regime declines as a gradual exponential function with superimposed approximately week-long periods of increasing probabilities as sequential winter storms move through the area.

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 277â&#x20AC;&#x201C;292

289

DeGraff, Cannon, and Gartner

Within the Arizona rainfall regime, debris flows were not reported until 28 days following ignition, but all events can be expected within 4 months of the fire, and the great majority can be expected in the first 3 months, all in response to monsoonal storms. July and August will host the majority of the debris flows. The probability of a day with debris flows increases rapidly once storms capable of triggering debris flows affect susceptible areas, with debris flows possible on a near-daily basis during this time. Within the Sub-Pacific rainfall regime, the first debris flow can ensue as soon as 11 days following ignition, and most debris flows can be expected within the first 13 months. Debris flows can be triggered in March, April, July, August, and September in response to spring, summer, and fall rainstorms. The probability of a day with debris flow decreases as a power law, with a higher rate at the beginning that slows over time but that extends into the year following ignition. In the area of the Plains rainfall regime, debris flows can first occur 4 days following fire ignition, and activity can continue for an additional 12 months, into the following summer and fall rainy seasons. Debris flows can be expected from June through September. The probability of a day with debris flow decreases as a power law very similar to that of the Sub-Pacific rainfall regime area, the difference being that debris flows start as soon as 4 days following ignition, rather than after the 11 days noted in the Sub-Pacific area. The decreasing likelihood of debris flows over time described here is thought to reflect the gradual restoration of hydrologic function as vegetative cover and soil infiltration rates return to pre-fire conditions, the temporal sequence of fires and storms, and prevailing storm-rainfall conditions in a given rainfall regime. Forested landscapes are recognized as having a second, later period of increased debris flow susceptibility in burned drainage basins. Although previous workers have reported that this period of susceptibility typically does not begin until 4 to 5 years following the wildfire (Meyer et al., 2001; Wondzell and King, 2003), our database included events that occurred 2.3 to 10 years following fire ignition, with 2 years coinciding with the earliest time at which tree root decay would reach a level that significantly weakens the strength of the soil mass. Debris flows that occur during these longer time frames most frequently are attributable to infiltration-triggered landslides, which mobilize into debris flow. The short time period between fire and debris-flow response within the first 1.5 years after ignition demonstrates the necessity of a rapid response by

290

land managers and emergency response agencies to mitigate potential hazards from potential runofftriggered debris flows from recently burned areas in the western United States. The presence of a second period of susceptibility to landslide-triggered debris flows between 2.3 and 10 years after a fire in forested terrains indicates the need to consider additional long-term mitigation solutions specific to this initiation process. ACKNOWLEDGMENTS The authors wish to express their appreciation to Richard Giraud and other researchers who provided either unpublished data or their time for interviews during compilation of our database. The insightful and detailed comments provided by an unnamed reviewer, Jeffrey Coe, Paul Santi, Dennis Staley, and Jason Kean materially improved the final manuscript. REFERENCES BROCK, R. J.; CANNON, S. H.; GARTNER, J. E.; SANTI, P. M.; HIGGINS, J. P.; AND BERNARD, D. R., 2007, An ordinary storm with an extraordinary response: Mapping the debris-flow response to the December 25, 2003 storm on the 2003 Old and Grand Prix fire areas in southern California: Geological Society America Abstracts Programs Vol. 39, No. 6, p. 180. BRUINGTON, A. E., 1982, Fire-loosened sediment menaces the city. In Proceedings of the Symposium on Dynamics and Management of Mediterranean-Type Ecosystems: U.S. Department of Agriculture Forest Service, Pacific Southwest Forest and Range Experimental Station, General Technical Report PSW-58, pp. 420–422. CANNON, S. H., 1997, Evaluation of the Potential for Debris and Hyperconcentrated Flows in Capulin Canyon as a Result of the 1996 Dome Fire, Bandelier National Monument, New Mexico: U.S. Geological Survey Open-file Report 97-136, 20 p. CANNON, S. H., 2001, Debris-flow generation from recently burned watersheds: Environmental Engineering Geoscience Vol. 7, pp. 321–341. CANNON, S. H.; BIGIO, E. R.; AND MINE, E., 2001, A process for fire-related debris-flow initiation, Cerro Grande Fire, New Mexico: 15, pp. 3011–3023. CANNON, S. H.; BOLDT, E. M.; KEAN, J. W.; LABER, J.; AND STALEY, D. M., 2011, Rainfall intensity-duration thresholds for postfire debris flow emergency-response planning: Natural Hazards: Vol. 9, pp. 209–236, 10.1007/s11069-011-9747-2. CANNON, S. H. AND DEGRAFF, J. V., 2009, Incorporating spatial, temporal, and climate variability into tools for assessing post wildfire debris-flow hazards. In Sassa, K. and Canuti, P. (Editors), Landslides: Disaster Risk Reduction: SpringerVerlag, Berlin, Germany, pp. 177–190. CANNON, S. H.; GARTNER, J. E.; HOLLAND-SEARS, A.; THURSTON, B. M.; AND GLEASON, J. A., 2003, Debris-flow response of basins burned by the 2002 Coal Seam and Missionary Ridge fires, Colorado. In Boyer, D. D.; Santi, P. M.; and Rogers, W. P. (Editors), Engineering Geology in Colorado–Contributions, Trends, and Case Histories: AEG Special Publication 15, Colorado Geological Survey Special Publication 55, 31 p.

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 277–292

Post-Fire Debris Flow Timing CANNON, S. H.; GARTNER, J. E.; RUPERT, M. G.; MICHAEL, J. A.; REA, A. H.; AND PARRETT, C., 2010, Predicting the probability and volume of post-wildfire debris flows in the intermountain west, USA: Geological Society America Bulletin: Vol. 122, pp. 127–144, doi:10.1130/B26459.1. CANNON, S. H.; GARTNER, J. E.; RUPERT, M. G.; MICHAEL, J. A.; STALEY, D. B.; AND WORSTELL, B. B., 2009, Emergency Assessment of Postfire Debris-Flow Hazards for the 2009 Station Fire, San Gabriel Mountains, Southern California: U.S. Geological Survey Open-File Report 2009-1227, 27 p. CANNON, S. H.; GARTNER, J. E.; WILSON, R. C.; AND LABER, J. L., 2008, Storm rainfall conditions for floods and debris flows from recently burned areas in southwestern Colorado and southern California: Geomorphology Vol. 96, pp. 250–269. CANNON, S. H.; POWERS, P. S.; AND SAVAGE, W. Z., 1998, Firerelated debris flows on Storm King Mountain, Glenwood Springs, Colorado, USA: Environmental Geology Vol. 35, pp. 210–218. CANNON, S. H. AND RENEAU, S. L., 2000, Conditions for generation of fire-related debris flows, Capulin Canyon, New Mexico: Earth Surface Processes Landforms Vol. 25, pp. 1103–1121. CERDA´, A. AND DOERR, S. H., 2005, Influence of vegetation recovery on soil hydrology and erodibility following fire: An 11-year investigation: International Journal Wildland Fire, Vol. 14, No. 4, pp. 423–437. CLEVELAND, G. B., 1973, Fire + rain 5 mudflows: California Geology, Vol. 26, pp. 127–135. COE, J. A.; KEAN, J. W.; GODT, J. W.; BAUM, R. L.; JONES, E. S.; GOCHIS, D. J.; AND ANDERSON, G. S., 2014, New insights into debris-flow hazards from an extraordinary event in the Colorado Front Range: GSA Today, Geological Society America Vol. 24, pp. 4–10. CONEDERA, M.; PETER, L.; MARXER, P.; FORSTER, F.; RICKENMANN, D.; AND RE, L., 2003, Consequences of forest fires on the hydrogeological response of mountain catchments: A case study of the Riale Buffaga, Ticino, Switzerland: Earth Surface Processes Landforms, Vol. 28, pp. 117–129. DEGRAFF, J. V., 1994, The geomorphology of some debris flows in the southern Sierra Nevada, California: Geomorphology, Vol. 10, pp. 231–252. DEGRAFF, J. V., 1997, Geologic Investigation of the Pilot Ridge Debris Flow, Groveland Ranger District, Stanislaus National Forest: U.S. Department of Agriculture Forest Service, Sonora, CA, 20 p. DEGRAFF, J. V.; CANNON, S. H.; AND GALLEGOS, A. J., 2007, Reducing post-wildfire debris flow risk through the Burned Area Emergency Response (BAER) process. In Schaefer, V. R.; Schuster, R. L.; and Turner, A. (Editors), Conference Presentations, 1st North American Landslide Conference: Vail, CO, AEG Special Publication 23, pp. 1440–1447. DEGRAFF, J. V.; WAGNER, D.; GALLEGOS, A. J.; DEROSE, M.; SHANNON, C.; AND ELLSWORTH, T., 2011, The remarkable occurrence of large rainfall-induced debris flows at two different locations on July 12, 2008, Sierra Nevada, CA: Landslides, Vol. 8, No. 2, pp. 343–353. DOEHRING, D. O., 1968, The effect of fire on geomorphic processes in the San Gabriel Mountains, California: Contributions Geology Vol. 7, pp. 43–65. DWYER, E.; PINNOCK, S.; GREGOIRE, J. M.; AND PEREIRA, J. M. C., 2000, Global spatial and temporal distribution of vegetation fire as determined from satellite observations: International Journal Remote Sensing Vol. 21, Nos. 6 and 7, pp. 1289–1302. EATON, E. C., 1936, Flood and erosion control problems and their solution.: Proceedings American Society Civil Engineers Vol. 62, pp. 1302–1362.

FLORSHEIM, J. L.; KELLER, E. A.; AND BEST, D. W., 1991, Fluvial sediment transport in response to moderate storm flows following chaparral wildfire, Ventura County, southern California: Geological Society America Bulletin Vol. 103, pp. 504–511. FLOYD, M. L.; HANNA, D. D.; AND ROMME, W. H., 2004, Historical and recent fire regimes in Pinõn-Juniper woodlands on Mesa Verde, Colorado, USA: Forest Ecology Management Vol. 198, pp. 269–289. FOX, D. M., 2011, Evaluation of the efficiency of some sediment trapping methods after a Mediterranean forest fire: Journal Environmental Management, Vol. 92, pp. 258–265. GARCÍA-RUIZ, J. M.; ARNAÉZ, J.; GO´MEZ-VILLAR, A.; ORTIGOSA, L.; AND LANA-RENAULT, N., 2013, Fire-related debris flows in the Iberian Range, Spain: Geomorphology, Vol. 196, pp. 221–230. GARTNER, J. E.; CANNON, S. H.; BIGIO, E. R.; DAVIS, N. K.; PARRETT, C.; PIERCE, K. L.; RUPERT, M. G.; THURSTON, B. L.; TREBISH, M. J.; GARCIA, S. P.; AND REA, A. H., 2005, Compilation of Data Relating to the Erosive Response of 606 Recently Burned Watersheds: U.S. Geological Survey OpenFile Report 2005-1218. GARTNER, J. E.; CANNON, S. H.; HELSEL, D. R.; AND BANDURRAGA, M., 2009, Multivariate Statistical Models for Predicting Sediment Yields from Southern California Watersheds: U.S. Geological Survey Open-File Report 2009-1200, 42 p. GARTNER, J. E.; CANNON, S. H.; SANTI, P. M.; AND DEWOLFE, V. G., 2008, Empirical models to predict the volumes of debris flows generated by recently burned basins in the Western U.S.: Geomorphology Vol. 96, pp. 339–354. GIRAUD, R. E. AND MCDONALD, N. M., 2004, April 6, 2004, Fire-Related Flooding and Debris Flows in Farmington: Electronic document, available at http://geology.utah.gov/ utahgeo/hazards/landslide/farmington040604.htm GRAY, D. H. AND MEGAHAN, W. F., 1981, Forest Vegetation Removal and Slope Stability in the Idaho Batholith: U.S. Department of Agriculture Forest Service Research Paper INT-271, Intermountain Forest and Range Experiment Station, Ogden, UT, 23 p. HELSEL, D. R. AND HIRSCH, R. M., 2002, Statistical Methods in Water Resources Investigations: Book 4, Chapter A3, 510 pp. Electronic document, available at http://pubs.usgs.gov/twri/ twri4a3/html/pdf_new.html JORDAN, P. AND COVERT, S. A., 2009, Debris flows and floods following the 2003 wildfires in southern British Columbia: Environmental Engineering Geoscience Vol. 15, No. 4, pp. 217–234. KEAN, J. W.; STALEY, D. M.; AND CANNON, S. H., 2011, In situ measurements of post-fire debris flows in southern California: Comparisons of the timing and magnitude of 24 debrisflow events with rainfall and soil moisture conditions: Journal Geophysical Research: 116:F04019, doi:10.1029/2011JF00 2005 LENTILE, L. B.; MORGAN, P.; HUDAK, A. T.; BOBBITT, M. J.; LEWIS, S. A.; SMITH, A. M. S.; AND ROBICHAUD, P. R., 2007, Post-fire burn severity and vegetation response following eight large wildfires across the western United States: Fire Ecology Vol. 3, No. 1, pp. 91–108. LLORET, F. AND MARI, G., 2001, A comparison of the medieval and the current fire regimes in managed pine forests of Catalonia (NE Spain): Forest Ecology Management Vol. 141, pp. 155–163. MCDONALD, G. N. AND GIRAUD, R. E., 2002, Fire-Related Debris Flows East of Santaquin and Spring Lake, Utah County, Utah: Utah Geological Survey Technical Report 02-09, 15 p.

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 277–292

291

DeGraff, Cannon, and Gartner MEYER, G. A.; PIERCE, J. L.; WOOD, S. H.; AND JULL, A. J. T., 2001, Fire, storms, and erosional events in the Idaho Batholith: Hydrological Processes Vol. 15, pp. 3025–3038. MEYER, G. A. AND WELLS, S. G., 1997, Fire-related sedimentation events on alluvial fans, Yellowstone National Park, U.S.A.: Journal Sedimentary Research Vol. 67, No. 5, pp. 776–791. MOODY, J. A. AND EBEL, B. A., 2012, Hyper-dry conditions provide new insights into the cause of extreme floods after wildfire: Catena Vol. 93, pp. 58–63. MOODY, J. A. AND MARTIN, D. A., 2009, Synthesis of sediment yields after wildland fire in different rainfall regimes in the western United States: International Journal Wildland Fire Vol. 18, pp. 96–115. MOODY, J. A.; SHAKESBY, R. A.; ROBICHAUD, P. R.; CANNON, S. H.; AND MARTIN, D. A., 2013, Current research issues related to post-wildfire runoff and erosion processes: Earth-Science Reviews Vol. 122, pp. 10–37. MORTON, D. M., 1989, Distribution and frequency of stormgenerated soil slips on burned and unburned slopes, San Timoteo Badlands, Southern California. In Sadler, P. M. and Morton, D. M. (Editors), Publications Inland Geological Society Vol. 2, pp. 279–284. NASA, 2014, Global Maps-Fire. Earth Observatory, National Aeronautic and Space Administrations: Electronic document, available at http://earthobservatory.nasa.gov/GlobalMaps/ view.php?d15MOD14A1_M_FIRE NIFC, 2011, Historically Significant Wildland Fires. National Interagency Fire Center, Boise, Idaho: Electronic document, available at http://www.nifc.gov/fireInfo/fireInfo_stats_histSigFires.html NYMAN, P.; SHERIDAN, G. J.; SMITH, H. G.; AND LANE, P. N. J., 2011, Evidence of debris flow occurrence after wildfire in upland catchments of south-east Australia: Geomorphology Vol. 125, pp. 383–401. PARISE, M. AND CANNON, S. H., 2012, Wildfire impacts on the processes that generate debris flows in burned watersheds: Natural Hazards, Vol. 61, pp. 217–227. PARRETT, C.,1987, Fire-Related Debris Flows in the Beaver Creek Drainage, Lewis and Clark County, Montana: U.S. Geological Survey Water-Supply Paper 2330, pp. 57–67. PARRETT, C.; CANNON, S. H.; AND PIERCE, K. L., 2003, WildfireRelated Floods and Debris Flows in Montana in 2001: U.S. Geological Survey Water-Resources Investigations Report 03-4319. PEARTHREE, P. A., 2004, Fire and sediment deposition: Arizona Geology, Vol. 34, pp. 1–2. SANTA BARBARA COUNTY PUBLIC WORKS DEPARTMENT, Santa Barbara County Flood Control and Water Conservation District, 2005, Santa Barbara County Public Works Department 2004–2005 Winter Report: Electronic document, available at http://www.worldcat.org/title/santa-barbara-county-publicworks-department-2004-05-winter-report/oclc/213938602

292

SANTI, P. M.; DEWOLFE, V. G.; HIGGINS, J. D.; CANNON, S. H.; AND GARTNER, J. E., 2008, Sources of debris flow material in burned areas: Geomorphology Vol. 96, pp. 310–321. SANTI, P. M.; HEWITT, K.; VANDINE, D. F.; AND BARILLAS CRUZ, E., 2011, Debris-flow impact, vulnerability and response: Natural Hazards Vol. 56, pp. 371–402. SANTI, P. M. AND MORANDI, L., 2013, Comparison of debris-flow volumes from burned and unburned areas: Landslides Vol. 10, No. 6, pp. 757–769. SCHULZ, W. H.; COE, J. A.; ELLIS, W. L.; AND KIBLER, J. D., 2006, Preliminary Assessment of Landslides Along the Florida River Downstream from Lemon Reservoir, La Plata County, Colorado: U.S. Geological Survey Open-file Report 20061341, 29 p. SCOTT, K. M., 1971, Erosion and Sediment Yields in the Transverse Ranges, Southern California: U.S. Geological Survey Professional Paper 1030, Pp. C242–C247 SHAUB, S., 2001, Landslides and Wildfire: An Example from the Boise National Forest: M.S. Thesis, Boise State University, ID. SIDLE, R. C., 2005, Influence of forest harvesting activities on debris avalanches and flows. In Jakob, M. and Hungr, O. (Editors), Debris-Flow Hazards and Related Phenomena: Springer, Berlin, Germany, pp. 387–409. SIDLE, R. C. AND OCHIAI, H., 2006, Landslides: Processes, Prediction, and Land Use: American Geophysical Union, Washington, DC. 312 p. WAGNER, D. L.; DEGRAFF, J. V.; AND LANCASTER, J. T., 2013, Debris flows in the southern and eastern Sierra Nevada region, California. In Putirka K. (Editor), Geologic Excursions from Fresno, California, and the Central Valley: A Tour of California’s Iconic Geology: Geological Society of America Field Guide 32, pp. 99–128. doi 10.1130/2013.0032(06) WEISBERG, P. J. AND SWANSON, F. J., 2003, Regional synchroneity in fire regimes of western Oregon and Washington, USA: Forest Ecology Management, Vol. 172, pp. 17–28. WELLS, W., 1987, The effects of fire on the generation of debris flows in southern California. In Costa, J. E. and Wieczorek, G. F. (Editors), Debris Flows/Avalanches, Process, Recognition, and Mitigation: Geological Society America Reviews Engineering Geology, No. 7, pp. 105–114. WOHL, E. E. AND PEARTHREE, P. P., 1991, Debris flows as geomorphic agents in the Huachuca Mountains of southeastern Arizona: Geomorphology, Vol. 4, pp. 273–292. WONDZELL, S. M. AND KING, J. G., 2003, Postfire erosional processes in the Pacific Northwest and rocky Mountain regions: Forest Ecology Management, Vol. 178, pp. 75–87. YOUBERG, A. M., 2014, Prehistoric and Modern Debris Flows in Semi-Arid Watersheds: Implications for Hazard Assessments in a Changing Climate: Ph.D. Dissertation, University of Arizona, Tucson, p. 235.

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 277–292

Measuring Orientations of Individual Concealed Sub-Vertical Discontinuities in Sandstone Rock Cuts Integrating Ground Penetrating Radar and Terrestrial LIDAR NORBERT H. MAERZ1 Missouri University of Science and Technology, 1006 Kingshighway, Rolla, MO 65409-0660

ADNAN M. AQEEL Taiba University, P.O. Box 30002 Madinah, Saudi Arabia 41477 email: ben-aqeel.2005@yahoo.com

NEIL ANDERSON Missouri University of Science and Technology, 1006 Kingshighway, Rolla, MO, 65409-0660, email: nanders@mst.edu

Key Terms: Engineering Geology, Terrestrial Lidar, Discontinuities, Orientation, Ground Penetrating Radar

discontinuities. Using the three-point method, the orientation of the hidden discontinuity is calculated. INTRODUCTION

ABSTRACT Vertical or sub-vertical discontinuities striking parallel to rock cuts are dangerous because toppling and spontaneous raveling failures can initiate from these surfaces, creating hazards below. At the same time, the surfaces of these discontinuities are often concealed because they do not “daylight” into the rock, and any trace of the discontinuity that might be seen at the top of the rock cut is obscured by overburden. These hidden discontinuities can often be detected by ground penetrating radar (GPR). Our new method uses GPR in conjunction with terrestrial LIDAR (light detection and ranging) to accurately measure the orientation of these hidden discontinuities. The method presented in this article establishes three control points on the surface of the rock cut. At each control point the global coordinates are remotely measured using LIDAR. GPR soundings at each control point are used to measure “the perpendicular horizontal distance” (depth) from each control point on the rock cut face to any discontinuities hidden behind the rock cut face. The true perpendicular distance is added to the GPR coordinates at each control point to form three new control points on the surface of each of the hidden

Corresponding author email: norbert@mst.edu.

Rock cut stability in most rock hard masses depends on the nature and orientation of the discontinuities (joints) within the rock mass. When they happen, failures typically start and propagate from and along the discontinuities. Failures can result in property damage, lane or highway closure, and even serious injury or death. If the orientation of all discontinuities within the rock mass behind the cut can be measured, a deterministic analysis can be undertaken to assess the likelihood of failure. This can be as simple as a kinematic analysis on a stereonet, as a limiting equilibrium analysis, or as complex as a numerical model that supports analysis of discontinuous rock. Discontinuities can be measured on rock cuts using the traditional manual scanline method (Otoo et al., 2011), which is very common, inexpensive, and easy to use but is time consuming and risky when measurements are carried out at the base of potential failure slopes. Alternatively, advanced in situ geometrical data collection methods are used, such as photogrammetric methods, total station surveying methods, or, more recently, light detection and ranging (LIDAR) methods (Post, 2001; Slob and Hack, 2004; Donovan et al., 2005; Haneberg, 2008; Sturzeinegger and Stead, 2009; and Otoo et al., 2011). Ground-based LIDAR scanners, which are sometimes called Terrestrial Laser Scanners (TLS), are geodetic instruments that have become very

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 293–309

293

Maerz, Aqeel, and Anderson

popular for engineering and geology surveys in recent years (Gonza´lez-Jorge et al., 2011; Otoo et al., 2011). These can provide accurate point cloud data of the scanned slope within minutes, and the geometry of rock discontinuities can be extracted in an automated and objective way (Pernito, 2008). Only discontinuities that can be seen in the rock cut can possibly be measured in this manner. Vertical or sub-vertical discontinuities that strike parallel to the rock cut often are not exposed in the rock cut because they do not “daylight” into the face of the excavation. Although they might be seen at the top of the rock cut, more often than not they are obscured by overburden and vegetation. In addition, if vertical borings are done from the top of the slope the vertical or subvertical discontinuities will not be encountered. Horizontal drilling can always be used to identify hidden discontinuities, but this drilling is a timeconsuming, expensive process that in some cases will require one or more lanes of the road to be closed. In addition to obtaining orientation measurements, costly oriented core drilling would also need to be done. Ground penetrating radar (GPR) has been shown to have the ability to both detect the hidden discontinuities and measure their angular trend (Maerz and Kim, 2000; Soel et al., 2001; Porsani et al., 2006; Pernito, 2008; and Torres, 2008). GPR is an active geophysical method for nondestructive subsurface imaging based on the propagation of electromagnetic (EM) waves in the subsurface (Reynolds, 1997; Conyers, 2004; Daniels, 2004; Otto and Sass, 2006; and Sass, 2007). Once the EM waves contact an interface plane whose electrical properties differ from those electrical properties of the surrounding subsurface materials (such as a discontinuity plane), a portion of this energy is reflected back to and recorded in the GPR system at the ground surface or the rock face. The strength of the GPR pulses reflecting from a rock discontinuity depends mainly on the aperture of the discontinuity and the infilling materials, both of which control the reflection coefficient (Gre´goire, 2001). This will give a distinctive linear interface, reflector, or event, with a high amplitude compared to background reflection, in the radiogram image. This distinctive reflection signature can be used as criteria with which to objectively delineate the discontinuities in rock masses (Pernito, 2008). METHODOLOGY Overview The key focus in this research is to appropriately combine both LIDAR and GPR data for the purpose

294

Figure 1. The colored circles are the locations of the three marked control points. The location of the GPR survey profile (the two dashed lines are the two index GPR profiles passing through the control points) on the rock cut face of Station 2 (St. 2).

of measuring the geometry of concealed sub-vertical discontinuities. Terrestrial LIDAR can accurately and easily measure the position of any point visible to the LIDAR scan, while GPR can measure the distance (perpendicular horizontal depth) between any point on an exposed surface and the projection of that point onto the plane of a detectable hidden subvertical discontinuity. Consequently, if only three points that are co-planar, but not co-linear, on the exposed surface are projected onto the plane of the hidden discontinuity, the true orientation of this hidden discontinuity can be determined. Measurement Protocols Briefly, the procedure to measure hidden discontinuities starts with identification of a nearly flat rock face. This does not have to be a discontinuity face, but a near-vertical surface is optimum for identifying and measuring hidden sub-vertical discontinuities. Then, three none colinear control points are marked on the rock face such that their positions in space in terms of coordinate triplets (x, y, z) can be measured using LIDAR (Figure 1). These control points could also be measured using other survey techniques. At a minimum, one GPR sounding must be done at each control point; however, obtaining multiple GPR traverses will be more effective in isolating the geometry of the concealed structures behind the rock slope face (Figure 1). Once GPR data are acquired, processing these data is required to enhance the quality of the data and to improve the clarity of the resulting radiogram images. After that process is complete, the apparent horizontal distance (depth) to the hidden discontinuities can be measured on the radiograms.

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 293–309

Measuring Orientations of Sub-Vertical Sandstone Discontinuities

As a final step in GPR processing data, a GPR data migration method is used to calculate the true horizontal distance to the hidden structure in a direction perpendicular to the rock face. By adding these perpendicular distances to each of the coordinate triples (x, y, z; control points) for each of the identified hidden structures (transpose the coordinate triplets to the hidden structure), the orientation of the hidden structures using the transposed coordinate triples can be calculated. Field Trials The measurements for this research were conducted on small rocks cut along an outer road paralleling the Interstate 44 highway about just north of Rolla, Missouri, USA. The measurements were made in the Rubidoux Sandstone, hard, competent sandstone with near-horizontal bedding and two sub-vertical joint sets that are approximately mutually orthogonal to each other and to the bedding orientation. The rock cut itself consist of abundant sub-vertical fracture surfaces stained red, with occasional whitecolored, blast-induced fractures. For this research, the testing was conducted only on sub-vertical discontinuity surfaces exposed in the rock cut. The rock cut (the study area) was divided into five stations. Each station was treated as an individual station on which both LIDAR and GPR measurements were conducted. LIDAR MEASUREMENT Overview The idea of mapping discontinuities on rock mass faces using remote sensing techniques is conventional. Terrestrial stereo-photogrammetric techniques have been recently used through improvements in imaging and digital processing data techniques. These and LIDAR measurements can be used for many applications in different disciplines, especially in the fields of rock engineering, rock mechanics, and landslides (Roberts and Propat, 2000; Fasching et al., 2001; Gaich et al., 2006; Kemeny et al., 2006; Whitworth et al., 2006; Dunning et al., 2009; Sturzeinegger and Stead, 2009; Abella´n et al., 2010; Asahina and Taylor, 2011; Garcı´a-Selle´s et al., 2011; and Gonza´lez-Jorge et al., 2011). Slob and Hack (2004) successfully used the semiautomated and automated approaches of three-dimensional (3D) laser scanning survey to map a rock slope face composed of carboniferous meta-siltstone and slate with well-developed discontinuity sets along a secondary road in Catalonia, Spain. They found

that even though the two approaches can produce high-resolution data suitable for mapping discontinuities and any other purpose in rock engineering, the full-automatic method is capable of capturing more data than are required for further statistical and/or modeling analysis. Moreover, another one of the most recent applications of the LIDAR is in the art of forecasting possible rock falls and rock mass slides, which are mainly controlled by the presence of discontinuities and their orientations and geometry (Abella´n et al., 2010). Alba and Scaioni (2010) have described how to extract change and rock mass deformation detection based on a LIDAR survey for the same rock face at two different times or periods. Their analysis was conducted by taking into account the multi-temporal point cloud georeferences and is built on three main steps: (1) vegetation filtering based on near infrared imagery; (2) detection of major changes such as loss materials; and (3) deformation analysis and testing. Moreover, the prediction of slope failures by monitoring and understanding of ongoing, even millimeter, deformation, which is mainly controlled by the geometrical and orientation characteristics of discontinuities, has been conducted by utilizing LIDAR technology (Abella´n et al., 2010). It is not difficult to carry out a 3D laser scan survey; however, it is quite a challenge to convert the LIDAR data to useful information that can directly be used for the purpose of slope stability analysis or any other purpose in the rock engineering practice. Different methods or approaches have been used to handle this issue, such as semi-automated or automated methods, in some cases using a geological compass for calibration (Slob et al., 2007; Aqeel, 2012; Maerz et al., 2012; and Otoo et al., 2012). Thus, the uncertainty and error are limited to that of manual compass measurements. However, traditional and LIDAR measurements are limited to exposed discontinuities on the rock slope, which excludes detection of hidden discontinuities that may have a significant effect on the rock slope stability analysis. For these reasons, it is desirable to employ a geophysical tool that will be able to detect and delineate or map hidden discontinuities or fractures at shallow depths inside the slope. Lidar Scanning Lidar scans were taken using a Leica ScanStation II scanner. The ScanStationII scanner is a time-of-flight, static, tripod-mounted system that deploys front and top windows with an oscillating mirror design to cover the full field of view of 360u horizontally and

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 293–309

295

Maerz, Aqeel, and Anderson

270u vertically. It has a detection range of 90 m at 90 percent reflectivity. It can scan 50,000 points per second with an accuracy on the order of 3 t0 5 mm. A laptop connected by Ethernet cable records data on range, angles, and degree of reflectivity of returning laser signals. The scanning system also collects optical images that are registered to the LIDAR data automatically using a built-in digital camera. The measurement process is initiated by selecting a rock cut face to image. In preparation for imaging, non-colinear control points are manually marked using chalk on near-planar face at each station. These points need to be enumerated and recorded either on the face or in the field notes (Figure 1). For the LIDAR measurement calibration process, the strike and dip angle of the rock face (or any planar vertical or sub-vertical orientation control element in the scan, such as a clipboard) is measured using a compass and the result recorded. The LIDAR scanner is then set up across from the rock face. No survey control is needed, but the scanner must be correctly leveled. Scanning at a low resolution (5-mm spacing) is more than adequate for most purposes. Only a single scan is required. Lidar Data Processing Processing of the LIDAR data is simple. A LIDAR viewer is used to view the resulting point cloud. The coordinate triplets (x, y, z) for each of the control points are identified by mouse click on the point cloud and their values recorded. Coordinate triples must also be identified and recorded for the orientation control element. These can be any arbitrary non-colinear points on the control element surface. In some cases, as shown in Figure 1, the rock cut face itself is used as an orientation control element, then the coordinate triples of the control points are used. No further processing of LIDAR data is required. GPR MEASUREMENT Overview GPR has previously been used to identify and measure the horizontal depth to discontinuities in a rock mass. Maerz and Kim (2000) conducted a field investigation in a sandstone rock formation in Missouri using GPR with 400-MHz antenna for the purpose of identifying the hidden vertical and/or subvertical discontinuities in a rock cut. The results showed the ability of the GPR to detect and depict the vertical discontinuities up to 2.5 m deep horizontally into the rock mass under dry conditions.

296

However, the strength of the reflected GPR pulses increases as the difference between relative dielectric permittivities increases (Aqeel, 2012). Open discontinuities that are filled with water and/or clay are clearly more visible in the GPR radiogram than are those discontinuities that are closed or that have no filling material (Toshioka et al., 1995; Pernito, 2008; and Otoo et al., 2011). However, the GPR penetration depth with appropriate resolution is less and the background noise is greater when water occurs in discontinuity apertures, causing, to some extent, more difficulty in identifying the hidden discontinuities. GPR Soundings Equipment In this research and for rock slope engineering purposes, the main point behind using GPR is to detect and map discontinuities in rock cuts within depths up to 4 m, since this is usually the range of depths in which discontinuities can play a major role in causing rock failures. The deeper (rather than shallower) discontinuities are not as likely to contribute to slope failure as they are less likely to contribute to toppling-type failures. The impulse GPR equipment used was manufactured by Geophysical Survey Systems, Inc. (GSSI) and utilized a 400-MHz monostatic antenna. A distance measuring wheel was attached to the antenna to acquire the GPR data in distance mode in order to produce 3D images for the detected discontinuities using both Radar Data Analyzer software package (RADANTM), which is manufactured by Geophysical Survey Systems, Inc., and ArcGIS 9.3, produced by Environmental Systems Research Institute, Inc. The 400-MHz antenna was selected as a compromise between depth of penetration and minimum resolution (minimum aperture of discontinuity that can be detected). Higher frequency antennas will detect smaller fractures at shallower depths, while lower frequency antenna will penetrate to deeper depths but detect only fractures with greater apertures. Using the 400-MHz antenna revealed measureable discontinuities at depths of at least 2.5 m. The theoretical minimum resolution (minimum resolvable thickness) of a layer is considered to be one quarter of the antenna wavelength (GonzaÂ´lez-Jorge et al., 2011). Measurement of GPR Pulse Velocity The GPR pulse velocity can be measured either in the field or in the lab; however, we preferred to estimate the velocity in the lab to ensure repeatability and accuracy. Therefore, a large rock sample was

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 293â&#x20AC;&#x201C;309

Measuring Orientations of Sub-Vertical Sandstone Discontinuities

collected from the study area to measure the velocity of GPR pulses that transmit through the sandstone of the study area. The rock sample was then trimmed into two rectangular blocks. To simulate natural conditions, the two rock blocks were positioned on top of each other with a separation of 1.80 cm between them to act as a discontinuity plane (Figure 2). The top block had a thickness of 10.60 cm, which means that the true perpendicular depth from the surface of the top block to that created discontinuity had to be 10.60 cm in the resulting radiograms of this test (Figure 3). Subsequently, the velocity of GPR pulses was measured in the laboratory utilizing a 1,500-MHz GPR monostatic antenna on the two rock blocks. The velocity of GPR signals is fixed for each earth material regardless of the frequency of the utilized GPR antenna. The reason a 1,500-MHz antenna was used is that it is more compatible with the small sample in terms of the sheer size of the 400-MHz antenna and its penetration depth. The resulting radiogram image showed that the two-way travel time (t) of the GPR pulses was 2 nanoseconds (ns), as illustrated in Figure 3. Since the measured perpendicular (vertical) depth is 10.60 cm, the velocity (V) can be calculated as follows: V ~2d=t

a detected hidden discontinuity in the study area of this research should be about 6 cm. However, Kovin (2010) showed that with a 400-MHz antenna much smaller parallel fracture apertures can be detected. In this study, discontinuities with apertures that were apparently less than 1 cm were resolved in the processed GPR data.

ð1Þ

where t 5 the two-way travel time 5 2 ns 5 t; d 5 the perpendicular (vertical) depth 5 10.60 cm 5 0.106 m, which was measured manually in the lab. Accordingly, V 5 0.106 ns/m. Once GPR pulse velocity is estimated, relative dielectric constant (e) can be accurately calculated as follows: V ~c=(e) =2

Figure 2. The two sandstone blocks, with an artificial separation acting as a discontinuity plane, to measure the velocity of the sandstone. The thickness of the top block is 10.6 cm.

ð2Þ

where e 5 the relative dielectric constant (e), which is dimensionless; and c 5 the speed of light in meters per nanosecond (0.3 m/ns). Accordingly, the relative dielectric constant (e) of the sandstone is 8. As a result, the relative dielectric constant in the GPR system during data acquisition can be set to 8. To test our results for this part of the research, the vertical axis that corresponds to the travel time or the penetration depth of GPR signals of the radiogram was set in a depth mode in the monitor of the GPR system; the radiogram image showed that the perpendicular (vertical) depth to the discontinuity was 10.60 cm, which reflects the accuracy and the precision of our lab work (Figure 3). Since the 400-MHz GPR antenna has a pulse period of 2.5 ns, the pulse wave length in this research is the product of 0.106 m/ns and 2.5 ns, which is 0.265 m. Consequently, the minimum resolvable aperture of

GPR Field Measurement Methodology Several parallel horizontal GPR profiles were acquired on each rock cut face (station) (Figure 1). These profiles were conducted in a horizontal direction. Two of these profiles were located to pass through the three fixed co-planar control points earlier identified for the LIDAR scanning. These two GPR profiles are called the index GPR profiles. Additional profiles were measured simply to increase the definition of the hidden discontinuities. The spacing between consecutive GPR profiles was between 10 and 20 cm. The spacing between and the trend of the GPR profiles were recorded in the field. A 400-MHz GPR monostatic antenna was moved along these profiles on the rock cut faces. Three persons were used for the GPR data acquisition (Figure 4). The locations of the control points on each rock cut were identified on the radiogram images utilizing the option of creating tick marks in the GPR display. Next, the collected GPR data were processed. GPR Data Processing Visual Enhancement GPR data processing was done for all of the acquired GPR data to enhance the visual quality of

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 293–309

297

Maerz, Aqeel, and Anderson

Figure 3. The radiogram of the sandstone blocks; the vertical axis is in time mode (above), showing that the two-way travel time (TWTT) from the surface of the top block (block 1) to the discontinuity plane (yellow dashed line) is 2 ns, and the vertical axis is in distance mode (below), showing that the true perpendicular (vertical) depth to the artificial discontinuity plane is 10.6 cm, which matches the distance measured manually in the lab; e 5 8. The horizontal axis is the distance along the rock surface in 1025 m.

the resulting two-dimensional (2D) and 3D radiogram images. The GPR product radiogram is not only an image but is also the recorded response of the subsurface materials and structures to the GPR EM waves. In most cases, radiograms are processed using specialized software to enhance their visual quality, and, thus, interpretation of the radiograms becomes easier and more reliable. In this research, the Radar Data Analyzer software package RADANTM was used for this purpose. After applying a zero correction (position or time-offset correction) for all GPR data, high-pass and low-pass filtering were used to

298

remove instrument noise from data to improve the quality of the data (Reynolds, 1997; Annan, 2009; and Cassidy, 2009a). Infinite impulse response filtering was used. A 100-MHz vertical pass filter was used to remove potential flat-laying ringing system noise, while an 800-MHz vertical low pass filter was used to remove high frequency. Generally, filtering the GPR data processing method is sufficient to locate subsurface features in many GPR applications (Reynolds, 1997; Annan, 2009). However, the main goal in this research was to detect concealed discontinuities, map them, and

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 293â&#x20AC;&#x201C;309

Measuring Orientations of Sub-Vertical Sandstone Discontinuities

Figure 4. Acquiring GPR data utilizing the 400-MHz GPR antenna on the rock cut face of St. 2. At least three persons are required to conduct such work.

measure their orientations, up to of 4 m deep. Consequently, additional GPR data processing methods were applied. Color transformation of value 17 was applied to hide what little noise may remain in the GPR radiograms. Moreover, automated range and display gain and deconvolution were applied to maximize resolution and improve the visual quality. The main purpose behind using the deconvolution method is to maximize bandwidth and reduce GPR pulse dispersion in order to ultimately maximize resolution. In deconvolution, an operator length of 31, a prediction lag of 5, pre-whitening of 10 percent, and an overall gain of 1 were used as parameters. Figures 5 through 7 display raw and processed acquired GPR data from the study area. In the study areas, 13 hidden sub-vertical discontinuities were manually identified on the radiograms at five stations. Figures 5 through 7 display the resulting processed radiograms of the two index GPR profiles (of the first three stations) showing the detected discontinuities at each rock cut (station). Three

hidden sub-vertical discontinuities were identified at each of both stations 1 and 3, two hidden sub-vertical discontinuities were identified at station 2, four hidden sub-vertical discontinuities were identified at station 5, and only one hidden sub-vertical discontinuity was identified at station 4. Apparent Horizontal Depths Once the vertical axis of the processed GPR radiogram image is set into depth mode instead of time mode on the GPR monitor, apparent depth (d) to any detected target in the radiogram image can be estimated manually from this radiogram image. Apparent depth can be defined here as the distance measured manually from a specific point (control point) on the rock surface to a discontinuity detected on the radiogram image in a direction perpendicular to the plane of the discontinuity. A migration process, as illustrated in Figure 8, is used to determine the true horizontal depth (d9)

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 293â&#x20AC;&#x201C;309

299

Maerz, Aqeel, and Anderson

Figure 5. The apparent perpendicular depths (d) to the detected hidden sub-vertical joints in station 1 (St. 1) after GPR data processing for the two GPR index profiles (P2 and P4); e 5 8.

perpendicular to the measurement plane. The apparent horizontal perpendicular depths (d), which were measured from the three control points to each detected hidden sub-vertical discontinuity the 2D radiograms of each of the five stations (Figures 5 to 7), are listed in Table 1. Migration and True Horizontal Depths Migration is a mathematical process and commonly the final step in GPR data processing used to relocate and reconstruct detected targets to their true locations and, thus, to their true geometry. Migration was applied in this research to determine the position of the control points projected onto the hidden discontinuity to reconstruct the true geometry of the detected discontinuities and, thus, the true perpendicular horizontal depths. Migration can be done

300

using specialized software or manually. GPR data migration using software is usually successful in relatively homogeneous environments, such as pavements and glacial environments. However, it tends to be less successful for complex and heterogeneous sites (Cassidy, 2009b). Therefore, manual migration was used in this research to avoid uncertainty resulting from variability in the inherent properties of the hidden discontinuities. Manual migration is explained in many geophysical references (Kleyn, 1983; Jenyon and Fitch, 1985; Conyers, 2004; Lines and Newrick, 2004; and Aqeel, 2012). The GPR pulses along each profile can be imagined as a perpendicular horizontal plane penetrating the rock cut face and intersecting the planes of the detected discontinuities. As a result, linear features (reflectors, interfaces, or events) will be recorded and shown in radiogram images (Figures 5 through 7).

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 293â&#x20AC;&#x201C;309

Measuring Orientations of Sub-Vertical Sandstone Discontinuities Table 1. The apparent and true perpendicular horizontal depths measured from the three control points on the rock cut face to the detected hidden sub-vertical discontinuities at each station in the study area. Apparent Perpendicular Horizontal Depths (z) Measured from the Control Points at St. 1 (cm) St. No. 1

2 3

4 5

Apparent (a) and True (b) Declination Angles of the Strike (u)

True Perpendicular Horizontal Depths (d) Measured from the Control Points at St. 1 (cm)

Discontinuity No.

Point 1 (Blue)

Point 2 (Red)

Point 3 (Yellow)

Point 1 (Blue)

Point 2 (Red)

Point 3 (Yellow)

1 2 3 1 2 1 2 3 1 1 2 3 4

148 261 325 188 287 165 208 364 348 187 224 248 336

133 234 344 198 285 172 223 396 355 189 227 249 323

170 282 314 177 318 206 290 404 379 197 240 236 320

11 16 10 16 22 17 35 03 11 7 7 9 5

11.2 16.7 10.2 16.7 23.8 17.8 44.4 03 11.2 7.1 7.1 9.1 5.1

151 272 330 196 314 173 291 364 355 188 225 251 338

136 244 350 207 311 181 312 396 362 190 229 252 324

173 294 319 185 348 216 406 404 386 198 242 239 321

Since the strike of a discontinuity can be defined as the angle of the intersection of the discontinuity plane and a horizontal plane, the resulting linear feature in the radiogram can be considered as “the strike line” of the detected discontinuity. For a dipping discontinuity, the migration process results in the apparent dip angle of the discontinuity being corrected to a steeper angle (Cassidy, 2009a, 2009b). Consequently, migration will reconstruct the apparent “strike line” of the detected discontinuity to a steeper, “deeper” declination angle of “the strike line.” Manual migration was done using the 2D GPR radiogram images and based on the following equation from Aqeel (2012): sin b~ tan a

ð3Þ

where b 5 the true declination angle of the strike line of the detected hidden sub-vertical discontinuity and a 5 the apparent declination angle of the strike line of the detected hidden subvertical discontinuity. Both the apparent perpendicular horizontal depths (distances) (d) and apparent declination angles (a) can be estimated from the 2D radiograms (Figures 5 through 7 and Table 1). Figure 8 explains how to estimate a and then calculate b based on Eq. 3. The true horizontal depths (d9) can then be estimated using the following equation (Table 1): d’~d= cos b

ð4Þ

Once the true depths (d9) are estimated, migration can be done, and 3D images showing the apparent and the true locations of the detected hidden sub-

vertical discontinuities can be generated (Figures 9 through 11). COMBINING LIDAR AND GPR MEASUREMENTS

Introduction Geospatial coordinate triplets (x, y, z) of any point on a rock cut are determined by simply identifying any particular point in the data set. The z-axis is in the vertical direction, while y and z axes are arbitrary horizontal axes, as defined by the LIDAR scanner. Selecting three control points on a planar object such as a discontinuity surface defines the attitude of that plane. Projecting these three control points horizontally and in a perpendicular direction from the rock cut face onto the hidden discontinuity plane will result in three new coordinate triplets that define the orientation of the hidden discontinuity (x9, x9, z9). Methodology The dip direction (w) and dip angle (h) of each rock cut face were measured using a Brunton compass. On each rock cut face, the Cartesian coordinates of the three control points were measured on the LIDAR viewer and recorded (Table 2). Using the three-point method, here the Cartesian coordinates converted to spherical (geographical) coordinates and, thus, the geometry of the rock cut face measured using the LIDAR (Table 2).

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 293–309

301

Maerz, Aqeel, and Anderson

Figure 6. The apparent perpendicular depths (d) to the detected hidden sub-vertical joints in station 1 (St. 2) after GPR data processing for the two GPR index profiles (P2 and P4); e 5 8.

The true depth vector (d) is always perpendicular to the strike direction at each rock cut face. Subsequently, the true depth could be resolved to two components x9 and y9. For instance, Figure 12 shows that the true depth vector at station 2 was resolved to x9 5 x 2 d cos 63u and y9 5 y 2 d cos 27u. Sometimes the true depth vector is in the same direction of the dip of the rock cut face. By substituting d values from Table 1 into the resulting two components for each rock cut (station), the Cartesian coordinates (x9, y9, z9) for each corresponding point on each detected

302

discontinuity plane can be mathematically calculated (Tables 3 through 5). Using the three-point method (Maerz et al., 2012), the resulting Cartesian coordinates of the detected hidden sub-vertical discontinuities were converted to spherical (geographical) coordinates, and, thus, their orientations were measured (Tables 3 through 5). Field verification measurements were conducted for those detected hidden discontinuities that had linear traces appearing on the rock cuts (Tables 3 through 5).

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 293â&#x20AC;&#x201C;309

Measuring Orientations of Sub-Vertical Sandstone Discontinuities

Figure 7. The apparent perpendicular depths (d) to the detected hidden sub-vertical joints in station 1 (St. 3) after GPR data processing for the two GPR index profiles (P2 and P4); e 5 8.

RESULTS AND VERIFICATION Verification Field verification (using a Brunton compass) of the LIDAR/GPR orientation measurements was possible only on those discontinuities that were exposed on the cut slope, and these were manually measured using a Brunton compass (Figure 13). The “Results” section below shows that about one half of the discontinuities measured by the LIDAR/GPR tech-

nique could be identified and measured elsewhere on the rock cut because they projected to an open face. The “Results” section also shows that typically the verification dip angle/dip direction measurements were between 3u and 7u. Typically, rock discontinuity surfaces have irregular and/or undulated surfaces that cause a variation in the value of the measured dip angle/dip direction of the slope using a geological compass from one part to another part on the same surface. The LIDAR/GPR measurements can give more reasonable and reliable measurements because

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 293–309

303

Maerz, Aqeel, and Anderson

Figure 8. The linear features in the radiogram represent the detected hidden sub-vertical discontinuities (yellow lines) at station 3 (e 5 8). The apparent perpendicular depth (d) (red vertical raw) measured from control point 1 (blue circle) to the detected discontinuity 2 at station 3 is 208 cm. The strike of discontinuity No. 1 has an apparent declination angle (a) of 17u, and so it has a true declination angle (b) of 17.8u.

they measure the orientation of the discontinuity over a wider base than can be obtained when using a compass, bridging over the irregularities. Results At station 1, the GPR technique was able to identify three hidden sub-vertical discontinuities to

a depth of 320 cm (Figures 5 and 8). Both of discontinuities 1 and 3 have a dip angle of 69u, with dip directions of 025u and 027u, respectively, while discontinuity 3 has a dip angle of 68u and a dip direction of 024u. Manual measured field verification measurements were conducted by measuring the orientation of those discontinuities that extend (daylight) out of the rock mass off to the side or at the top of the cut, where they can be measured (Table 3). The difference between the measured orientation using the LIDAR and geological compass was within 3u for dip direction and within 4u for dip angle, which can be attributed to human and/or device errors, or it can simply be a function of measuring the discontinuity in a different place. Similarly, at station 2, the GPR technique was able to identify two hidden sub-vertical discontinuities to a depth of 360 cm (Figures 5 through 7). The dip angle of discontinuity 1 is 78u, and the dip angle is 82u for discontinuity 2, while the dip direction for these two discontinuities is 029u and 035u, respectively (Table 4). These two mapped discontinuities have exposed linear traces, which made field verifications easy and accurate (Figure 13). The difference between the measured geometry using the LIDAR and that obtained using a geological compass was within 6u for dip direction and within 7u for dip angle. At station 3, the GPR technique was able to identify three hidden sub-vertical joints within a perpendicular depth of 400 cm (Figures 6 and 10). The dip angles for those three detected discontinuities are between 86u and 89u, while the dip direction is between 193u and 197u (Table 5). Verification was

Figure 9. A 3D image displays both of the apparent and true locations of the deteted sub-vertical discontinuities at station 1 (St. 1).

304

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 293â&#x20AC;&#x201C;309

Measuring Orientations of Sub-Vertical Sandstone Discontinuities

Figure 10. A 3D image displays both of the apparent and true locations of the deteted sub-vertical discontinuities at station 2 (St. 2).

possible for only one hidden discontinuity, the linear trace of which was exposed (Figure 13). The difference between the measured geometry using the LIDAR and that obtained using a geological compass was within 3u for dip direction and within 3u for dip angle (Tables 2 and 5). Only one hidden sub-vertical discontinuity was identified at a depth of 360 cm by the GPR instrument at station 4. The dip direction and dip angle of this discontinuity are 021u and 89u, respectively. Verification was not possible for this discontinuity, as it does not “daylight” anywhere.

Verification was also not possible for the four detected hidden discontinuities at station 5. The dip angles of these four detected discontinuities are 88u, 89u, 88u, and 82u, respectively. Their dip directions are 015u, 014u, 016u, and 015u, respectively. Limitations and Application A common limitation of using GPR technology is the GPR data processing and interpretation, which is still subjective and depends mainly on the interpreter’s skills and experience in interpreting radiograms. In

Figure 11. A 3D image displays both of the apparent and true locations of the deteted sub-vertical discontinuities at station 3 (St. 3).

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 293–309

305

Maerz, Aqeel, and Anderson Table 2. The coordinates of the three control points and, thus, the geometry of the rock cut faces. Station No.

Geometry of the Rock Slope Face (u)

Cartesian Coordinates of the Three Control Points on the Rock Slope Face

In the Field Point 1

1 2 3 1 2 3 1 2 3 1 2 3 1 2 3

(Blue) (Red) (Yellow) (Blue) (Red) (Yellow) (Blue) (Red) (Yellow) (Blue) (Red) (Yellow) (Blue) (Red) (Yellow)

6458.09 7422.82 5639.71 2991.82 2979.83 2194.30 8966.93 8977.10 8170.74 4506.50 4485.41 2804.01 2644.17 2599.15 1629.43

16235.77 15878.50 16751.02 15447.26 15417.44 15803.41 15322.53 15359.48 15708.86 13003.10 13012.93 13626.68 13969.41 13986.50 14195.73

2324.57 2122.79 2003.56 464.68 260.11 248.70 2936.98 2486.48 2394.12 21057.25 2821.73 2670.18 588.63 741.64 775.75

Dip Direction

By LIDAR

Dip Angle

Dip Direction

Dip Angle

028

026

069

022

027

202

199

195

202

010

015

Table 3. Lidar geometrical measurements for the detected hidden sub-vertical joints at St. 1. Cartesian Coordinates and Orientations of the Detected Discontinuities Cartesian Coordinates of the Three Corresponding Control Points Discontinuity No.

Corresponding Point No.

1 2 3 1 2 3 1 2 3

6391.89 7363.20 5563.87 6338.85 7315.85 5510.82 6313.42 7269.38 5499.86

16100.05 15756.26 16595.53 15991.30 15659.19 16486.77 15939.17 15563.92 16464.30

2324.57 2122.79 2003.56 2324.57 2122.79 2003.56 2324.57 2122.79 2003.56

LIDAR (u) Dip Direction

Field Verification (u)

Dip Angle

Dip Direction

Dip Angle

025

024

027

026

*NA 5 Not Applicable

Table 4. LIDAR geometrical measurements for the detected hidden sub-vertical joints at St. 2. Cartesian Coordinates and Orientations of the Detected Discontinuities Cartesian Coordinates of the Three Corresponding Control Points Discontinuity No. 1

306

Corresponding Point No.

1 2 3 1 2 3

2903.42 2886.47 2110.87 2850.21 2839.57 2307.35

15272.62 15233.00 15638.58 15167.49 15140.34 15493.34

464.68 260.11 248.70 464.68 260.11 248.70

LIDAR (u) Dip Direction

Dip Angle

Field Verification (u) Dip Direction

Dip Angle

029

035

030

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 293â&#x20AC;&#x201C;309

Measuring Orientations of Sub-Vertical Sandstone Discontinuities Table 5. LIDAR geometrical measurements for the detected hidden sub-vertical joints at St. 3. Cartesian Coordinates and Orientations of the Detected Discontinuities Cartesian Coordinates of the Three Corresponding Control Points Discontinuity No. 1

Corresponding Point No.

1 2 3 1 2 3 1 2 3

8910.60 8918.17 8100.41 8872.18 8875.51 8038.55 8848.41 8848.16 8039.20

15158.96 15188.34 15504.63 15047.38 15064.48 15324.98 14978.36 14985.05 15326.87

2936.98 2486.48 2394.12 2936.98 2486.48 2394.12 2936.98 2486.48 2394.12

addition, this entire methodology is somewhat time consuming and requires lab testing to determine the velocity of the rock, although this is likely to be constant over large areas of the same rock type. Another practical limitation of the method is the necessity for a relatively flat near-vertical face. While this seems to disqualify a large number of rock cuts, it is also true that because discontinuities tend to occur in parallel to sub-parallel sets, where, consequently, there is sub-vertical concealed discontinuity striking parallel to a rock cut, often the rock cut itself is a product of a parallel or sub-parallel discontinuity. A second practical limitation is that the GPS operation requires physical contact between the GPS unit and the face. This means that equipment like a man-lift would be needed for access to anything that cannot be reached from a safe standing spot below

Figure 12. Resolved depth vector at St.2 to x9 and y9; x9 5 x 2 d cos 63u; and y9 5 y 2 d cos 27u.

LIDAR (u)

Field Verification (u)

Dip Direction

Dip Angle

Dip Direction

Dip Angle

197

194

193

199

the cut. Consequently, the height limitation would involve the limitations of the man-lift. CONCLUSIONS The results of this investigation show how accurately joint orientation data can be obtained on hidden sub-vertical discontinuities using a combination of LIDAR and GPR. Comparison of the orientation results shows that the manually (compass-) measured results are very close to the LIDAR measurements. Differences are likely to be caused by the fact that discontinuities are not perfectly planar, and variable measurements can be expected depending on the part of the discontinuity in which the measurements are made. Furthermore, the extension of linear traces of some detected discontinuities measured using a geological compass may differ from those measurements resulting from GPR and/or LIDAR. This explains why the differences between the measured dip direction using a geological compass and those measured using LIDAR can be up to 7u. Migration of the GPR data was a necessary step in order to accurately estimate the true depths (true perpendicular horizontal depths) to the detected hidden discontinuities. Once this step has been accomplished,, the true geometry of those discontinuities can be measured using our new method. Not only can hidden discontinuity orientations be measured, but any other specific hidden object can be detected and mapped and then measured in terms of its orientation in space using this method, whether for civil or military purposes. Moreover, the results show that three GPR survey lines conducted on a rock slope face can be enough to obtain GPR data, 2D radiograms, which can be integrated with LIDAR data for hidden sub-vertical discontinuity orientation

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 293â&#x20AC;&#x201C;309

307

Maerz, Aqeel, and Anderson

Figure 13. Conducting field verifications at station 2 (left) and station 3 (right) in the study area. Two linear traces of discontinuities appear on the side of station 2, and only one linear trace of a discontinuity appears on the side of station 3.

measurements. Consequently, not only can accuracy be increased, but the labor cost and time-consuming factors can be significantly reduced to a much greater extent when 3D GPR radiograms are produced and used for such purposes. ACKNOWLEDGMENTS We would like to deeply thank both the Geological Engineering Program and Rock Mechanics & Explosive Research Center at Missouri University of Science and Technology for their technical assistance. REFERENCES ABELLA´N, A.; CALVET, J.; VILAPLANA, J. M.; AND BLANCHARD, J., 2010, Detection and spatial prediction of rockfalls by means of terrestrial laser scanner monitoring: Geomorphology, Vol. 119, pp. 162–171. ALBA, M. AND SCAIONI, M., 2010, Automatic detection of changes and deformations in rock faces by terrestrial laser scanning. In Commission V Symposium of Newcastle upon Tyne: International Archives Photogrammetry Remote Sensing Spatial Information Sciences Vol. 38, No. 5, pp. 11–16. ANNAN, A. P., 2009, Electromagnetic principles of ground penetrating radar. In Jol. and H. M. (Editors), Ground Penetrating Radar: Theory and Applications: Elsevier B. V, Slovenia, 524 p. AQEEL, A. M., 2012, Measuring the Orientations of Hidden Subvertical Joints in Highway Rock Cuts Using Ground Penetrating Radar in Combination with Lidar: Unpublished Ph.D. Dissertation, Missouri University of Science and Technology, p. 282. ASAHINA, D. AND TAYLOR, M. A., 2011, Geometry of irregular particles: Direct surface measurements by 3-D laser scanner: Powder Technology, Vol. 213, pp. 70–78.

308

CASSIDY, N. J., 2009a, Ground penetrating radar data processing, modeling, and analysis. In Jol. and H. M. (Editors), Ground Penetrating Radar: Theory and Applications: Elsevier B. V, Slovenia. 524 p. CASSIDY, N. J., 2009b, Electrical magnetic properties of rocks, soils and fluids. In Jol, H. M. (Editor), Ground Penetrating Radar: Theory and Applications: Elsevier B. V, Slovenia. 524 p. CONYERS, L. B., 2004, Ground-Penetrating Radar for Archaeology: ALTAMIRA Press, Lanham, MD, 209 p. DANIELS, D. J., 2004, Ground penetrating radar: Institution of Electrical Engineers, London, U.K, 726 p. DONOVAN, J.; KEMENY, J.; AND HANDY, J., 2005, The application of three-dimensional imaging to rock discontinuity characterization, Alaska rocks. In Proceedings of the 40th U.S. Rock Mechanics Symposium: Anchorage AK, 25–29 June 2005, 7 p. DUNNING, S. A.; MASSEY, C. I.; AND ROSSER, N. J., 2009, Structural and geomorphological features of landslides in the Bhutan Himalaya derived from terrestrial laser scanning: Geomorphology, Vol. 103, pp. 17–29. FASCHING, A.; GAICH, A.; AND SCHUBERT, W., 2001, Data acquisition in engineering geology: An improvement of acquisition methods for geotechnical rock mass parameters: Felsbau, Vol. 19, No. 5, pp. 93–101. GAICH, A.; PO¨TSCH, M.; AND SCHUBERT, W., 2006, Acquisition and assessment of geometric rock mass features by true 3D images. In Proceedings of ARMA Golden Rocks 2006—The 41st U.S. Symposium on Rock Mechanics (USRMS): “50 Years of Rock Mechanics—Landmarks and Future Challenges: Golden, CO, 17–21 June 2006, 10 pp. GARCI´A-SELLE´S, D.; FALIVENE, O.; ARBUE´S, P.; GRATACOS, O.; TAVANI, S.; AND MUN˜OZ, J. A., 2011, Supervised identification and reconstruction of near-planar geological surfaces from terrestrial laser scanning: Computers Geosciences, Vol. 37, pp. 1584–1594. GONZA´LEZ-JORGE, H.; RIVERIO, B.; ARMESTO, J.; AND ARIAS, P., 2011, Standard artifact of the geometric verification of terrestrial laser scanning systems: Optics Laser Technology, Vol. 43, pp. 1249–1256.

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 293–309

Measuring Orientations of Sub-Vertical Sandstone Discontinuities GRE´GOIRE, C., 2001, Fracture Characterization by Ground-Penetrating Radar: Unpublished Ph.D. Dissertation, Katholieke University of Leuven, Belgium. HANEBERG, W., 2008, Using close range terrestrial digital photogrammetry for 3-D rock slope modeling and discontinuity mapping in the United States: Bulletin Engineering Geology Environment, Vol. 67, No. 4, pp. 457–469. JENYON, M. K. AND FITCH, A. A., 1985, Seismic Reflection Interpretation: Stuttgart- Gebru¨der-Borntraeger, Berlin, Germany. KEMENY, J.; NORTON, B.; AND TURNER, K., 2006, Rock slope stability analysis utilizing ground-based lidar and digital image processing: Felsbau, Vol. 24, No. 3, pp. 8–15. KLEYN, A. H., 1983, Seismic Reflection Interpretation: Elsevier Applied Science Publishers LTD, London, England, 269 p. KOVIN, O. N., 2010, Ground Penetrating Radar Investigations in Upper Kama Potash Mines: Unpublished Ph.D. Dissertation, Missouri University of Science and Technology, 182 p. LINES, L. R. AND NEWRICK, R. T., 2004, Fundamentals of Geophysical Interpretation: In Geophysical Monograph Series, No. 13: Society of Exploration Geophysicists, Tulsa, OK. MAERZ, N. AND KIM, W., 2000, Potential use of ground penetrating radar in highway rock cut stability: Geophysics: 2000, St. Louis, MO, USA, Dec. 11–15, 2000, 9 p. MAERZ, N. H.; YOUSSEF, A.; OTOO, J. N.; KASSEBAUM, T. J.; AND DUAN, Y., 2012, A simple method for measuring discontinuity orientations from terrestrial lidar images: Environmental Engineering Geoscience, Vol. 19, pp. 185–195. OTOO, J.; MAERZ, N.; XIAOLING, L.; AND DUAN, Y., 2011, 3-D discontinuity orientations using combined optical imaging and Lidar techniques. In Proceedings of the 45th U.S. Rock Mechanics Symposium: San Francisco CA, 26–29 June 2011, 9 p. OTOO, J. N.; MAERZ, N. H.; DUAN, Y.; AND XIAOLING, L., 2012, Verification of a 3-D lidar point cloud viewer for measuring discontinuity orientations. In The 46th U.S. Rock Mechanics/ Geomechanics Symposium: San Francisco, CA, 24–27 June 2012. OTTO, J. C. AND SASS, O., 2006, Comparing geophysical methods for talus slope investigations in the Turtmann Valley (Swiss Alps): Geomorphology, Vol. 76, pp. 257–277. PERNITO, M., 2008, Rock Mass Slope Stability Analysis Based on 3D Terrestrial Laser Scanner and Ground Penetrating Radar: Unpublished M.S. Thesis, ITC: International Institute for Geo-Information Science and Earth Observation, 86 p.

PORSANI, J.; SAUCK, W.; AND JUNIOR, A., 2006, GPR for mapping fractures and as a guide for the extraction of ornamental granite from a quarry: A case study from southern Brazil: Journal Applied Geophysics, Vol. 58, pp. 177–187. POST, R., 2001, Characterizing of Joints and Fractures in a Rock Mass Using Digital Image Processing: Unpublished M.S. Thesis, University of Arizona, Tucson, 105 p. Reynolds, J. M., 1997, An Introduction to Applied and Environmental Geophysics: John Wiley & Sons, New York. 796 p. ROBERTS, G. AND PROPAT, G., 2000, Highwall joint mapping in 3D at the Moura mine using SIROJOINT. In: Bowen Basin Symposium of Coal and Mining: The New Millennium, Rockhampton, October 2000. SASS, O., 2007, Bedrock detection and talus thickness assessment in the European Alps using geophysical methods: Journal Applied Geophysics, Vol. 62, pp. 254–269. SLOB, S. AND HACK, R., 2004, 3D terrestrial laser scanning as a new field measurement and monitoring technique: Engineering Geology Infrastructure Planning Europe, Vol. 103, pp. 179–189. SLOB, S.; HACK, H.; FENG, Q.; ROSHOFF, K.; AND TURNER, A. K., 2007, Fracture mapping using 3D laser scanning techniques. In Proceedings of the 11th Congress of the International Society for Rock Mechanics, Lisbon, Portugal, 9–13 July 2007, pp. 299–302. SOEL, S.; KIM, J.; SONG, Y.; AND CHUNG, S., 2001, Finding the strike direction of fractures using GPR: Geophysical Prospecting, Vol. 49, pp. 300–308. STURZEINEGGER, M. AND STEAD, D., 2009, Close-range terrestrial digital photogrammetry and terrestrial laser scanning for discontinuity characterization on rock cuts: Engineering Geology, Vol. 106, pp. 163–182. TORRES, C., 2008, Geometric Characterization of Rock Mass Discontinuities Using Terrestrial Laser Scanner and Ground Penetrating Radar: Unpublished M.S. Thesis, ITC: International Institute for Geo-Information Science and Earth Observation, 74 p. TOSHIOKA, T.; TSUCHIDA, T.; AND SASAHARA, K., 1995, Application of GPR to detecting and mapping cracks in rock slopes: Journal Applied Geophysics, Vol. 33, No. 1–3, pp. 119–124. WHITWORTH, M. Z.; GILES, D.; ANDERSON, I.; AND CLEWETT, M., 2006, Terrestrial laser scanning for applied geosciences studies in the urban environment: JAEG, Geological Society London, Vol. 252, pp. 1–9.

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 293–309

309

Cut Slope Design for Stratigraphic Sequences Subject to Differential Weathering: A Case Study from Ohio YONATHAN ADMASSU Department of Geology and Environmental Science, James Madison University, Harrisonburg, VA 22807

ABDUL SHAKOOR1 Department of Geology, Kent State University, Kent, OH 44242

Key Terms: Cut Slope Design, Stratigraphy, Interlayered Rock Units, Differential Weathering, Rockfalls

ABSTRACT Designing cut slopes along Ohio highways depends on local stratigraphy and slope stability problems. Based on stratigraphy and modes of failure, cut slopes in Ohio were divided into three types: 1) those consisting of strong rock units (sandstones and limestones) that exhibit discontinuity-related failures; 2) those consisting of weak rock units (shales, claystones, and mudstones) that exhibit raveling, gully erosion, and rotational sliding; and 3) those consisting of interlayered strong and weak rock units where differential weathering causes undercutting-induced failures. Data regarding geological, geotechnical, and geometrical parameters were collected for 26 sites representing the three types of slopes and were used to perform kinematic analysis, rockfall trajectory simulations, and global stability analysis. This article focuses on the design of cut slopes in interlayered stratigraphy where differential weathering is the primary cause of slope instability. Based on stratigraphic variations, we categorized cut slopes in the interlayered units into four types: Type I—thick sandstone underlain by thick shale or claystone/ mudstone; Type II—sandstone interlayered with shale or claystone/mudstone in nearly equal proportions; Type III—limestone interlayered with claystone/mudstone in nearly equal proportions; and Type IV— claystone/mudstone interlayered with minor, thin limestone layers. Based on stability analyses and rockfall simulations, we recommend cut slope designs for each stratigraphic sequence that consider slope angles for undercut units to reduce rockfall potential, slope angles for undercutting units that are close to naturally stable angles, benches to reduce undercutting and contain

Corresponding author email: ashakoor@kent.edu.

rockfalls, drainage to reduce erosion, and catchment ditches to contain rockfalls. INTRODUCTION Most cut slopes in Ohio consist of interlayered sequences of strong and weak rock units of varying thicknesses. These slopes are highly subject to differential weathering and undercutting-induced failures (Shakoor and Weber, 1988; Shakoor, 1995; and Admassu and Shakoor, 2012). Currently available methods of rock slope design can be used for designing cut slopes in uniformly strong rocks or uniformly weak rocks but they cannot be directly applied for designing cut slopes in interlayered sequences of strong and weak rocks. The purpose of this study, which was funded by the Ohio Department of Transportation (ODOT), is to develop a rational approach for designing cut slopes in sub-horizontal, interlayered sequences of sedimentary rocks subject to differential weathering. The approach is based on a detailed investigation of 26 cut slope sites representative of varying stratigraphic conditions present in Ohio.

GEOLOGIC SETTING OF OHIO The geologic setting of Ohio is mainly a result of Paleozoic sedimentation and Pleistocene glaciation. The oldest rocks in Ohio are Ordovician-age limestones deposited in a shallow, warm sea (Camp, 2006). The Acadian mountains, resulting from the collision of the Baltica plate and the North American plate (,375 mega-annum [Ma]), supplied sediments for the Devonian-Mississippian-age sandstones and shales of Ohio. The latest Alleghenian orogeny (,318 Ma) resulted in the rise of the Appalachian Mountains that provided the source of Ohio’s Pennsylvanian-Permianage sedimentary rocks. There is, however, little to no sedimentary record present in Ohio from the late Paleozoic, Mesozoic, and most of Cenozoic time

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 311–324

311

Admassu and Shakoor

(between 290 million and 300,000 years) present in Ohio. As a result of the tectonic history, the bedrock geology of Ohio consists of nearly flat-lying carbonate and siliciclastic sedimentary rocks from Upper Ordovician to Lower Permian. In southwestern Ohio, bedrock predominantly consists of Upper Ordovician interlayered limestones and shales or claystones/ mudstones. Dolomites and shales of Silurian age underlie the western and west-central parts of the state. The northwestern and central parts of Ohio are underlain by Devonian marine carbonate rocks (limestones and dolomites interlayered with shales) as well as siliciclastic rocks (sandstones and siltstones interlayered with shales). Mississippian-age rocks, including sandstones, siltstones, conglomerates, and shales, with minor proportions of limestone, cover the east-central part and northwestern corner of the state. The largest part of eastern Ohio is covered by Pennsylvanian-age rocks, including sandstones, siltstones, limestones, shales, claystones/mudstones, and some coals. Lower Permian/Upper Pennsylvanian-age sandstones, siltstones, shales, claystones/mudstones, and minor coal seams cover southeastern Ohio. These rocks were deposited as cyclothems in non-marine, deltaic, or estuarine environments (Chesnut, 1981; Bennington, 2002). Most cut slopes in Ohio are located in the eastern and southeastern parts of Ohio, which are characterized by interlayered sandstones, limestones, shales, claystones, mudstones, and minor amounts of coal. Slopes in the southwestern part are characterized by interlayered limestones, shales, claystones, and mudstones. Only a few rock slopes are present in central Ohio. Types of Rock Slopes and Modes of Failure in Ohio Based on stratigraphy, cut slopes in Ohio can be divided into three broad types: 1) those that comprise mostly (.90 percent) strong rock units (sandstones, limestones); 2) those that comprise mostly (.90 percent) weak rock units (shales, claystones, mudstones); and 3) those that comprise interlayered strong and weak rock units. Cut slopes belonging to type 1 are usually less than 30 ft (9 m) high and make up approximately 20–25 percent of all cut slopes. Those belonging to type 2 are much less common (,10 percent). The majority of the cut slopes in Ohio belong to type 3. These cut slopes are highly variable in height and may contain from a few to many interlayered strong and weak rock units of varying thicknesses. For cut slope design purposes, these three types of slopes can be considered to have three distinctly different design requirements.

312

Figure 1. Example of ravelling of a shale slope (FRA-270-23).

The failures that may affect type 1 slopes include plane, wedge, and toppling failures that are triggered by an unfavorable orientation of discontinuities with respect to orientation of the slope face (Admassu and Shakoor, 2013a). The common modes of failure affecting type 2 cut slopes are general degradation and raveling of weak rock due to weathering with raveled material accumulating at the bottom of the slope (Figure 1), gully erosion and mudflows (Figure 2), and, less frequently, rotational slides. Type 3 slopes experience failures that are typical of both strong rock units and weak rock units. Differences in durability of interlayered rock units causes differential weathering, resulting in undercutting of the stronger rock layers, creating overhangs. Once the depth of undercutting exceeds the joint spacing within the undercut layer, rockfalls begin to occur (Figure 3). Undercutting also promotes plane, wedge, and toppling failures along discontinuities that do not “daylight” on the original cut slope (Shakoor and Weber, 1988; Shakoor, 1995). Depending on their location on the slope face, these failures become rockfalls as they descend the slope (Shakoor and Weber, 1988). Thus, rockfalls are the dominant mode of failure affecting cut slopes subject to differential weathering. Kinematics of Undercutting-Induced Failures Undercutting-induced failures are kinematically possible when at least three sets of intersecting

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 311–324

Cut Slope Design

Figure 3. Rockfalls resulting from the undercutting of a stronger rock unit by a weaker rock unit (WAS-7-18.2).

that develop perpendicular to each other. Valley stress relief joints result from horizontal extension of valley walls as stream erosion removes lateral support (Ferguson and Hamel, 1981). The rock blocks released as rockfalls can be bounded either by the bedding and two sets of orthogonal joints or by bedding, a set of orthogonal joints, and a set of stress relief joints. When the undercut blocks are first released, the initial movement could be either in the form of a plane failure, a wedge failure, or a toppling failure (Shakoor and Weber, 1988; Shakoor, 1995). Toppling failures, associated with undercutting, occur when the depth of undercutting extends beyond the blockâ&#x20AC;&#x2122;s center of gravity (Neimen, 2009). As a result of the dominance of near-vertical discontinuities in Ohio, toppling is a common mode of failure. Regardless of the initial mode of failure, all undercutting-induced failures become rockfalls. RESEARCH METHODS Site Selection

Figure 2. (a) Gully erosion of a slope consisting of redbeds (ATH-33-23); (b) mudflow on a shale slope caused by groundwater seepage (CLE-275-5.2).

discontinuities are present so that a rock block can move freely when the depth of undercutting exceeds the spacing between the discontinuities. The three common types of discontinuities present in Ohio are bedding planes, orthogonal joints, and valley stress relief joints. Orthogonal joints are sub-vertical joints

We performed a reconnaissance survey of 113 cut slope sites across the state of Ohio to select 26 sites (Appendix 1 and Figure 4) for detailed investigations for this study. Site designation in Appendix 1 and Figures 1 through 4 follows ODOT standard notation, which uses the three-letter county code, the numerical name of the road, and the mile marker from the county line. The sites were selected to ensure that they are representative of different stratigraphic scenarios and geologic ages. Appendix 1 shows that of the 26 sites, six consist mostly of strong rock units, two mostly of weak rock units, and 18 mostly of interlayered rock units. The selected sites have

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 311â&#x20AC;&#x201C;324

313

Admassu and Shakoor

Figure 4. Location map of the 26 study sites.

a greater representation of slopes consisting of interlayered strong and weak rock units because such slopes are the most common and most problematic in terms of performance and hazard potential. This study focuses on design of cut slopes for the interlayered stratigraphy based on the 18 sites representative of this type of stratigraphy.

314

Field Investigations Field investigations consisted of collecting data regarding slope geometry (slope angle, slope height, slope aspect, slope profile, bench width), slope stratigraphy, discontinuity characteristics (orientation, continuity, spacing, nature of infilling material,

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 311â&#x20AC;&#x201C;324

Cut Slope Design

surface irregularities, groundwater conditions), depth of undercutting, and catchment ditch dimensions width and depth). A laser range finder was used to prepare the slope profiles, which were then used to prepare stratigraphic cross sections for each site. Fifteen of the sites were drilled, and borehole logs from drilling were used to refine the stratigraphic details. Field values of rock quality designation (RQD) for stronger rock units were estimated using the method proposed by Palmstrom (1982), which is based on the number of joints within a cubic meter of rock outcrop. The detailed line survey method (Piteau and Marin, 1977), the window mapping method (Wyllie and Mah, 2004), and random measurements were used to collect the discontinuity data. A measuring tape and laser range finder (for inaccessible layers) were used to measure the depth of undercutting. The presence of pre-split blast-hole markings on the undercut rock unit provided a reference to ensure that the undercut unit had remained in place since the time of construction. Where pre-split markings were absent, original design plans for the cut were used for this purpose. Laboratory Investigations Laboratory tests were performed to determine unconfined compressive strength, second-cycle slake durability index (Id2), friction angle, and density values for both strong and weak rock units. Unconfined compressive strength and slake durability index were determined using American Society for Testing and Materials (ASTM) methods D2968 and D4644, respectively (ASTM, 1996); density was determined by measuring weights and volumes of oven-dried core samples; and Stimpson’s method (Stimpson, 1981) was used for determining friction angle. Laboratory data were used for various types of stability analyses. Unconfined compressive strength values ranged from 1,179 to 21,507 psi (8.13–148.32 MPa), with a mean of 9,001 psi (62.10 MPa), for sandstone core samples (n 5 34); from 4,148 to 25,699 psi (28.61–177.23 MPa), with a mean of 15331 psi (105.73 MPa), for limestone core samples (n 5 23); from 322 to 10,646 psi (2.22–73.40 MPa), with a mean of 2399 psi (16.54 MPa), for shale core samples (n 5 43); and from 222 to 3,109 psi (1.53–21.44 MPa), with a mean of 1,557 psi (10.73 MPa), for claystone/ mudstone core samples (n 5 28). The mean secondcycle slake durability index values for sandstones (n 5 34), limestones (n 5 18), shales (n 5 30), and claystone/mudstone (n 5 26) samples were found to be 93 percent, 98 percent, 91 percent, and 35 percent, respectively. The mean density values for sandstones (n 5 7), limestones (n 5 4), shales (n 5 13), and

claystones/mudstones (n 5 8) were 145 lb/ft3 (2.32 Mg/m3), 158 lb/ft3 (2.53 Mg/m3), 166 lb/ft3 (2.66 Mg/m3), and 166 lb/ft3 (2.66 Mg/m3), respectively. The mean basic friction angle along discontinuities was 36u for sandstones and 43u for limestones. Stability Analysis Undercutting-induced rockfalls, promoted by differential weathering, represent the most common type of failure affecting cut slopes with interlayered stratigraphy. A rational method for analysis and design of cut slopes subject to differential weathering cannot be developed without understanding the factors that cause undercutting. We used bivariate and multivariate statistical methods to identify the geological and geotechnical factors that influence the depth of undercutting. SPSS (Statistical Package for the Social Sciences) and Microsoft Excel software were used for statistical analysis. A detailed discussion of the factors affecting the depth of undercutting can be found in Admassu et al. (2012). The fate of the rockfalls with respect to the effect of slope height, slope angle, and catchment ditch dimensions was evaluated using rockfall simulation software, RocFall (Rocscience, 2006). RocFall determines the trajectory and the landing space of a rockfall generated from any point on the slope face. We performed kinematic analysis, using DIPS software (Rocscience, 2006), to evaluate the potential for plane, wedge, and toppling failures for thick (.10 ft/3 m), strong rock units within the interlayered sequences. Kinematic analysis did not show any significant potential for plane or wedge failures due to persistently steep dips of the discontinuities. In addition, we used the SLIDE software (Rocscience, 2006) to determine factor-of-safety values against rotational failures due to low rock mass strength for slopes consisting of interlayered rock units. Compressive strength and geologic strength index data (GSI) were used for this analysis. The GSI data were obtained using the method and charts developed by Marinos and Hoek (2001) and Hoek et al. (2005). The resulting factor-of-safety values ranged from 2.7 to 40. We also observed naturally stable slope angles for weak rock units at numerous sites to aid in selecting appropriate cut slope angles for weak rock units. Information about the factors that have the most influence on the depth of undercutting (obtained from statistical analyses), the expected frequency, sizes, and fate of rockfalls (obtained from RocFall analysis), was used to suggest an appropriate cut slope design (slope angle, bench width and location, catchment ditch width, stabilization techniques) for the interlayered strata.

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 311–324

315

Admassu and Shakoor

DESIGN APPROACH Based on our detailed investigations of the 18 cut slopes in interlayered stratigraphic sequences in Ohio, we propose the following approach for designing cut slopes subject to differential weathering: 1. Categorize the site stratigraphy into one of four categories, as discussed later. 2. Identify the factors that influence the depth of undercutting at the site and prepare a list of methods that may be used to reduce the effect of these factors. 3. Evaluate the appropriate slope angles for strong rock units, taking into account the number of undercutting-induced rockfalls, using the twodimensional (2D) numerical simulation software UDEC (UDEC, 2014). 4. Evaluate the ultimate stable angles for weak, undercutting rock units. 5. Evaluate the fate of rockfalls including bounce heights and roll out distances, using rockfall trajectory simulation software (RocFall), for designing catchment ditches. 6. Design benches that take into account the sitespecific stratigraphic conditions so that the potential for undercutting is reduced and some of the rockfalls may land on the benches, considering rockfall roll out distances.

Categorizing Site Stratigraphy for Design Purposes On the basis of variations observed within the interlayered stratigraphic sequences in the Appalachian plateau of Ohio, we categorized stratigraphy into four types: Type I—thick sandstone layer underlain by thick shale or claystone/mudstone layer; Type II—sandstone interlayered with shale or claystone/mudstone in nearly equal proportions; Type III—limestone interlayered with claystone/mudstone in nearly equal proportions; and Type IV—claystone/ mudstone interlayered with minor limestone. Figure 5 shows examples of the four types of stratigraphy. Factors Influencing Undercutting Based on observations of the 18 cut slopes chosen, and the subsequent statistical analysis, the following factors were studied for their possible contribution to the depth of undercutting: 1) vertical distance of the undercut unit from the slope crest; 2) relative position of the undercut unit from the slope crest, defined as the vertical distance of the undercut unit from the slope crest divided by the total slope height.; 3) total

316

Figure 5. Example of four types of stratigraphic configurations: (a) Type I; (b) Type II; (c) Type III; and (d) Type IV.

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 311–324

Cut Slope Design

contributes to undercutting; Figure 6). Shakoor and Rogers (1992) showed that the rate of undercutting is dependent on the durability of the undercutting rock unit. Admassu et al. (2012) and Neimen (2009) suggest that the rate of undercutting is not necessarily linear and might slow down over time as the undercutting unit reaches a stable angle. The above discussion suggests that during cut slope design, special considerations should be given to undercut layers close to the slope crest and to the highly jointed strong rock layers because they will likely be deeply undercut. Effect of Slope Angles for Strong Rock Units on the Number of Rockfalls

Figure 6. (a) Undercutting caused by groundwater flowing out of joints and (b) undercutting caused by surface runoff over a slope face.

thickness of the undercut unit; 4) spacing of orthogonal joints within the undercut unit; 5) slake durability index value of the undercutting unit; 6) initial slope angle; and 7) age of the road cut. Multiple regression analyses conducted by Admassu et al. (2012) showed that the seven factors identified above explained 61 percent of the variation in the depth of undercutting, with position of the undercut unit and spacing of joints in the undercut unit explaining the most variation. Undercut units closer to slope crest and highly jointed undercut units have the highest susceptibility to deeper undercutting. Admassu et al. (2012) attribute the unexplained 39 percent of the variation to the amount of water seeping along the contact between the undercut and the undercutting units (i.e., surface runoff also

The main discontinuities within strong rock units (sandstones, limestones) of interlayered stratigraphic sequences are invariably sub-horizontal bedding planes and orthogonal joints (Admassu et al., 2012). The orthogonal joints do not exhibit any preferred orientation but consistently show steep dips (average 79u) and steep lines of intersection (.70u). Kinematic analysis, using DIPS software (Rocscience, 2006), shows that plane and wedge failures are not possible because of the steepness of discontinuities. Therefore, undercutting-induced rockfalls, including toppling, is the primary mode of failure in interlayered stratigraphy. Using a 2D numerical modeling software, UDEC (UDEC, 2014), we investigated the effect of varying slope angles on the number of undercuttinginduced rockfalls (Admassu and Shakoor, 2013a). The results of our investigation showed that strong rock units cut at 1H:1V (45u) angles resulted in the smallest number of rockfalls. However, considering the large amount of excavation required to cut slopes at 1H:1V (45u), we recommend an angle of 0.5H:1V (63u), which would significantly reduce the number of rockfalls without requiring too much excavation. Stable Slope Angles for Weak, Undercutting Rock Units One important question that needs to be addressed is whether the weak rock units tend to reach a final stable angle beyond which further undercutting will not occur. Angles of undercutting rock units that appeared stable, as indicated by the presence of vegetation, were measured for the 18 study sites as well as for some additional sites. These angles show a normal distribution, with an average value of 38u (Figure 7), which may be considered as the final stable angle for undercutting rock units. Therefore, weak units cut at 1.5H:1V (35u) are expected to stabilize sooner, consequently causing fewer rockfalls over time. However, where more

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 311â&#x20AC;&#x201C;324

317

Admassu and Shakoor

Figure 7. Frequency distribution of stable slope angles for the undercutting rock units. The upper bound of each class in the histogram is labeled in the middle of each bar.

durable silty shales form the weaker rock units, 1H:1V (45u) slope angles may be acceptable. Fate of Rockfalls and Design of Benches and Catchment Ditches An important consideration when designing cut slopes prone to undercutting-induced failures is to study the trajectories and volume of released rockfalls. This information is vital for designing benches and rockfall catchment ditches. We investigated the following aspects of rockfalls to address this issue: 1. Effect of rockfall shape and type of rock comprising the rockfalls on the fate of rockfalls. 2. Effect of slope angle, slope height, and slope stratigraphy on the roll out distance of rockfalls (i.e., how far rockfalls travel from the toe of the slope). This was evaluated using the rockfall simulation software RocFall (Rocscience, 2006). This information is essential for designing benches and width, depth, and slope angle of catchment ditches. Field observation and measurement of rockfall dimensions revealed that the travel behavior of

rockfalls is governed mainly by the ratio of bedding thickness to joint spacing for the undercut units. Equi-dimensional rockfalls with a ratio of 1 tend to roll, whereas rock units for which this ratio is less than 1 result in flat rockfalls that tend to slide and may remain on the slope instead of landing in the catchment ditch (Admassu and Shakoor, 2013b). Rockfalls generated from undercut limestone units, with an average bedding thickness to joint spacing ratio of 0.96, tend to roll and, therefore, have greater trajectories. On the other hand, rockfalls comprised of sandstone, with an average bedding thickness to joint spacing ratio of 0.53, have flatter shapes and are more likely to stay on the slope face. The sizes of undercutting-induced rockfalls also vary substantially, ranging from less than 0.81 ft3 (0.03 m3) to 110.70 ft3 (4.1 m3). In addition to investigating rockfall shape–controlled trajectories, we performed simulations, using RocFall software (Rocscience, 2006), for the different rockfall weights (average weight for sandstone rockfalls 5 764 lb/361 kg; average weight of limestone rockfalls 5 68 lb/31 kg), stratigraphic configurations, slope angles, and catchment ditch slopes. The simulation considered pre-split blasting that is used to create a smooth slope surface during slope excavation. Pre-split blast-holes can be drilled to a maximum depth of 40 ft (12 m) before providing a 1.5 ft (0.5 m) offset for the next phase of drilling. The maximum roll out distances of rockfalls released from different heights, different slope angles, and different catchment ditch slopes were recorded for each type of stratigraphic configuration (Appendix 2). The widths of benches and catchment ditches are based on the rockfall roll out distances. In order to recommend catchment ditch width and bench width for different slope angles, a relationship between rockfall roll out distances to slope height was established using regression equation. For example, bench width (based on roll out distance) 5 0.5 H + 2, where H is slope height in feet (Table 1).

Table 1. Summary of catchment ditch design recommendations based on rockfall simulations (modified from Admassu and Shakoor, 2013b). Type I

Type II

Option 1 Slope Flat 6:1 3:1

Option 2

Option 1

Option 2

Width (ft)

Slope

Width (ft)

Slope

Width (ft)

Slope

Width (ft)

0.4 3 H + 3 0.4 3 H 2 2 0.1 3 H + 2

Flat 6:1 3:1

0.3 3 H 0.3 3 H 2 1 0.15 3 H + 1

Flat 6:1 3:1

0.35 3 H + 5 0.3 3 H + 5 0.2 3 H + 5

Flat 6:1 3:1

0.4 3 H + 7 0.25 3 H + 9 0.25 3 H + 6

Note: H is slope height above the catchment ditch, in feet; 1 ft 5 0.303 m. 1 “Slope” under various options refers to slope of the catchment ditch, expressed as the ratio of the horizontal to the vertical. Catchment ditches in Ohio are usually designed with slope angles of 6:1, 3:1, or zero (flat).

318

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 311–324

Cut Slope Design

DESIGN RECOMMENDATIONS Our recommendations for designing cut slopes in interlayered stratigraphic sequences subject to differential weathering take into account the four types of stratigraphic configurations, the associated slope instability problems, and the anticipated trajectories of rockfalls. The recommendations are aimed at 1) reducing the number of undercutting-induced rockfalls by selecting appropriate cut slope angles for thick, strong rock units, based on modelling results using UDEC software; 2) providing stabilization measures for undercut rock units that have the highest potential for the deepest undercutting, as suggested by the previously discussed factors influencing the depth of undercutting; 3) reducing the rate of undercutting by selecting slopes for undercutting rock units that are close to their naturally stable angles, providing benches along the contacts between undercut and undercutting units when feasible, and providing drainage to intercept water seeping from the fractures in the undercut, strong rock units onto the undercutting units; and 4) reducing the rockfall hazard to the roadways by providing adequate catchment ditches based on rockfall trajectories. In the following sections, we provide specific recommendations for each of the four types of stratigraphic scenarios. Table 2 provides a summary of the design recommendations.

reduce rockfalls due to toppling failures and cutting shale or claystone/mudstone at 38u or less. Based on rockfall simulations, a bench should be provided along the contact between the two units to delay the process of undercutting and to retain fallen rock (Figure 8). The bench width should be 0.5 3 slope height in feet (above the bench) + 2. Pre-split blasting should be used for the sandstone layer, and the maximum allowable single slope should not exceed 40 ft (12 m), which is the maximum depth for an offset to be provided during pre-split blasting. Catchment ditch width should be based on the guidelines provided in Table 2. In order to reduce surface runoff on the slope of the undercut unit, a drainage ditch, filled with rip rap, should be provided behind the crest of the sandstone slope (Figure 8). In addition, the seepage along the sandstone-shale contact should be collected by a rip rap–filled ditch constructed on the bench. Both drainage ditches should be connected to the drainage at the toe of the cut slope. With option 2, the sandstone can be cut at an 0.25H:1V (76u) angle, which will be prone to toppling/rockfalls, but rock bolts can be used to minimize such failures. A narrower bench (0.45 3 slope height 2 2) can be provided, as rockfalls will have shorter trajectories from the steeper slope. The underlying unit can be cut at 1.5H:1V (34u), gentler than 38u, so that degradation of the underlying slope will not compromise the width of the bench. The catchment ditch design should be based on the guidelines provided in Table 2.

Type I Stratigraphy Type I stratigraphy consists of thick (.7–10 ft/2–3 m) sandstone units underlain by shale or claystone/ mudstone. The main concern with such slope configuration is undercutting of the sandstone layer by the underlying weak layer, causing rockfalls and toppling failures. The presence of thin, friable sandstone layers within the thick, harder sandstone can also cause undercutting-induced failures. Two design options are recommended for Type I stratigraphy. Option 1 consists of cutting the sandstone at 0.5H:1V (63u) to

Type II Stratigraphy Type II stratigraphy consists of thin (,3 ft/1 m) sandstone interlayered with shale or claystone/mudstone in variable proportions. We recommend using a uniform slope angle for this type of stratigraphy that would result in fewer rockfalls. Pre-split blasting should be used, and the maximum slope height should not exceed 40 ft (12 m). Benches should be provided at a maximum slope height of 40 ft (14 m). Two design options are shown in Figure 9. Option 1

Table 1. Extended. Type III

Type IV

Option 1

Option 2

Option 1

Option 2

Slope

Width (ft)

Slope

Width (ft)

Slope

Width (ft)

Slope

Width (ft)

Flat 6:1 3:1

0.8 3 H + 15 0.6 3 H + 4 0.35 3 H + 5

Flat 6:1 3:1

1.1 3 H + 7 0.55 3 H + 6 0.35 3 H + 5

Flat 6:1 3:1

1.25 3 H + 4 0.55 3 H + 4 0.35 3 H + 3

Flat 6:1 3:1

1.1 3 H + 8 0.55 3 H 0.25 3 H + 2

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 311–324

319

Admassu and Shakoor Table 2. Summary of design recommendations. (Note: The recommendations in this table represent guidelines based upon the study of a series of generalized rock slope models. The reader is advised that these guidelines are intended strictly for the preliminary design of rock slopes in Ohio and other localities having similar stratigraphy. Other, more detailed analyses may be required for actual slope applications). Stratigraphic Group Type I (Figure 8)

Slope Angle Option 1

Option 2

Type II (Figure 9)

Type III (Figure 10)

Type IV (Figure 11)

0.5H:1V for sandstone/ 1H:1V for shale 0.25H:1V for sandstone/ 1.5H:1V for shale

Bench Design: Width (ft)

Catchment Ditch Design Slope1

Width (ft)

0.5 3 H + 2 (H . 40 ft)

Flat 6:1 3:1

0.4 3 H + 3 0.4 3 H 2 2 0.1 3 H + 2

0.45 3 H 2 2 (H . 40 ft)

Flat 6:1 3:1

0.3 3 H 0.3 3 H 2 1 0.15 3 H + 1

Drainage2

Stabilization

Option 1

Uniform angle of 1H:1V

0.35 3 H + 5 (H . 40 ft)

Flat 6:1 3:1

0.35 3 H + 5 0.3 3 H + 5 0.2 3 H + 5

Option 2

Uniform angle of 0.5H:1V

0.4 3 H + 7 (H . 40 ft)

Flat 6:1 3:1

0.4 3 H + 7 0.25 3 H + 9 0.25 3 H + 6

Option 1

Uniform angle of 1H:1V

0.8 3 H + 15 (H . 40 ft)

Flat 6:1 3:1

0.8 3 H + 15 0.6 3 H + 4 0.35 3 H + 5

Option 2

Uniform angle of 0.5H:1V

1.1 3 H + 7 (H . 40 ft)

Flat 6:1 3:1

1.1 3 H + 7 0.55 3 H + 6 0.35 3 H + 5

Option 1

Uniform angle of 1.5H:1V

1.25 3 H + 4 (H . 40 ft)

Flat 6:1 3:1

1.25 H + 4 0.55 3 H + 4 0.35 3 H + 3

Option 2

Uniform angle of 1H:1V

1.1 3 H + 8 (H . 40 ft)

Flat 6:1 3:1

1.1 3 H + 8 0.55 3 H 0.35 3 H + 2

Rock bolts should be used to stabilize the upper half of the slope

Shotcrete (concrete spraying) the top half of the slope

Shotcrete (concrete spraying) the top half of the slope Erosion-control matting

Slope crest, bench, and catchment ditch drainage should be provided Slope crest, bench, and catchment ditch drainage should be provided Slope crest and catchment ditch drainage should be provided Slope crest and catchment ditch drainage should be provided Slope crest and catchment ditch drainage should be provided Slope crest and catchment ditch drainage should be provided Slope crest, mid-slope, and catchment ditch drainage should be provided

Erosion-control matting

Note: H is slope height above the bench or the catchment ditch, in feet; 1 ft 5 0.303 m. 1 “Slope,” in the Catchment Ditch Design column, refers to the slope of the catchment ditch, expressed as the ratio of the horizontal to the vertical. Catchment ditches in Ohio are usually designed with slope angles of 6:1, 3:1, or zero (flat). 2 Slope crest drainage refers to the provision of a drainage ditch on top of the backslope of a bench. Bench drainage refers to the provision of a drainage ditch on the bench. Mid-slope drainage refers to the provision of a drainage ditch in the middle of a slope. Catchment ditch drainage refers to the provision of a drainage ditch within the catchment ditch at the toe of the slope (Figures 8–11).

involves using a uniform slope of 1H:1V (45u), which is close to the average natural slope angle and not too gentle for pre-split blasting. Catchment ditches or benches can be designed based on the rockfall simulation results shown in Table 2. A drainage ditch should be provided on the slope crest to reduce surface runoff. For option 2, in order to reduce the amount of excavation, the slope can be cut at a steeper angle of 0.5H:1V (63u), which can significantly reduce the number of undercutting-induced rockfalls. The top half of the slope should be stabilized using shotcrete to prevent the weathering of weak layers. Perforated drain pipes should be installed through the shotcrete so that groundwater seepage is not blocked. Catchment ditches and benches should be designed according to the results of rockfall simulation, as summarized in Table 2.

320

Type III Stratigraphy Type III stratigraphy involves limestone interlayered with equal proportions of claystone/mudstone. Such slopes generate a high number of cubical, limestone rockfalls that have long trajectories because of their shapes. The design approach for this type of stratigraphy should include a uniform slope angle, which will reduce undercutting-induced failures. Benches should be provided at a maximum height of every 40 ft (12 m). Stabilization techniques such as rock bolts will not be effective because of the close spacing of joints in thin limestone layers. As for slope angles, we recommend two design options (Figure 10), similar to Type II stratigraphy. In option 1, the slope is cut at 1H:1V (45u), which will be gentle enough to reduce undercutting-induced rockfalls. Slope crest drainage

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 311–324

Cut Slope Design

Figure 10. Design options for Type III stratigraphy. Figure 8. Recommended slope design for Type I stratigraphy.

Type IV stratigraphy is characterized by claystone/ mudstone units with minor layers of limestone. Presplit blasting is not necessary, as these slopes consist mainly of weak rocks. Slopes cut in Type IV stratigraphy also generate cubical rockfalls that can have long trajectories. Since catchment ditches to contain such rockfalls can be excessively wide, catchment fences should be considered. The best

design approach for such slopes is to use a gentle slope angle that is close to the natural stable angle of 38u and to provide slope crest drainage to reduce surface erosion. Since such slopes are highly prone to surface erosion, a mid-slope drain, lined with rip rap and connected to the back slope drain, may be necessary for high cut slopes. These highly erodible slopes should also be covered by erosion-controlling mats made from biodegradable materials. Figure 11 provides two options for selecting cut slope angles. Option 1 consists of cutting the slope at 1.5H:1V (34u), which is closer to the natural stable angle. If the slope is higher than 20 ft (6 m), a mid-slope drainage ditch connected to the slope toe drain should be provided. If there is a layer of limestone close to a mid-slope section, the mid-slope drain should follow the limestone layer to siphon away seeping groundwater as well as to collect surface water (Figure 11). Erosion-control matting should be installed. Catchment ditch/bench design should be based on rockfall simulation, as summarized in Table 2. Rockfall catchment fences should be used as part of the ditch design to reduce the width of catchment ditches. If the limestone layers represent , 20 percent of the slope face, a steeper slope angle of 1H:1V (45u) can be used as option 2. Mid-slope and slope crest drainage ditches and erosion-control matting should be

Figure 9. Design options for Type II stratigraphy.

Figure 11. Design options for Type IV stratigraphy.

should be provided. Benches, when necessary (slopes . 40 ft/12 m high), and catchment ditches should be designed using Table 2. The catchment ditch widths, based on the trajectories of simulated rockfalls (Table 2), tend to be extremely large. Therefore, narrower catchment ditches with catchment fences should be considered. Option 2 utilizes a cut slope of 0.5H:1V (63u), which will require less excavation than option 1 but at the same time will reduce undercuttinginduced failures to a large extent. The upper part of the slope can be stabilized with well-drained shotcrete. Slope crest drainage and bench drainage should be provided (Figure 10). Catchment ditch/bench widths are summarized in Table 2. Type IV Stratigraphy

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 311â&#x20AC;&#x201C;324

321

Admassu and Shakoor

provided, similar to option 1. Catchment ditch/bench design should be based on rockfall simulation (Table 2). Rockfall catchment fences should be used as part of the ditch design to reduce the width of catchment ditches. CONCLUSIONS

stratigraphic variations II and III, cut slopes can generally be designed at 1H:1V (45u) or 0.5H:1V (63u), and for variation IV they can be designed at 1H:1V (45u). Rockfall simulation is required for design of drainage ditch widths and bench widths. Table 2 in the text provides a summary of design recommendations.

Based on the results of this study, the following conclusions can be drawn: ACKNOWLEDGMENT 1. Slope stability problems in Ohio depend on the prevalent stratigraphy, which often consists of stronger, durable rock units (sandstones, limestones) alternating with weaker, non-durable rock units (shales, claystones, mudstones). This type of stratigraphy is highly prone to differential weathering, which results in undercutting and formation of unsupported overhangs. Undercutting leads to various types of slope failures, such as rockfalls, toppling failures, plane failures, and wedge failures. 2. Cut slopes in Ohio can be divided into three distinctly different types: 1) those consisting mostly (.90 percent) of strong rock units; 2) those consisting mostly (.90 percent) of weak rock units; and 3) those consisting of interlayered strong and weak rock units, each ranging in proportion from more than 10 percent to 90 percent. Most cut slopes in Ohio are located in the interlayered stratigraphic sequences. 3. Slope stability problems affecting the interlayered rock units are primarily undercutting-induced failures (plane failure, wedge failure, toppling failure). Regardless of the mode of failure, all undercutting-induced failures become rockfalls. 4. Cut slope design for interlayered sequences is complex and depends on stratigraphic variations. Four stratigraphic variations, designated as Types I through IV, are recognized within the interlayered sequences, as follows: Type I—thick (.7–10 ft/ 2–3 m) sandstone or limestone underlain by shale or claystone/mudstone; Type II—thin to mediumthick (,3 ft/1 m) sandstone units interlayered with shale or claystone/mudstone units in variable proportions; Type III—thin to medium-thick limestone (,3 ft/1 m) units interlayered with claystone/mudstone units in variable proportions; and Type IV—thin to medium-thick (,3 ft/1 m) limestone units interlayered with claystone/mudstone units in usually minor proportions. The cut slope design for Type I stratigraphy combines design principles for strong and weak rocks, with the provision for a bench along the contact between the two rock types. The sandstone can be cut at 0.5H:1V (63u) and the shale at 1H:1V (45u). For

322

The authors would like to thank the Ohio Department of Transportation (ODOT) and the Federal Highway Administration (FHWA) for the financial support of this research project. DISCLAIMER The authors are solely responsible for the contents of this this article, including the accuracy of the data. The contents do not reflect the official views or policies of the Ohio Department of Transportation or the Federal Highway Administration. REFERENCES ADMASSU, Y. AND SHAKOOR, A., 2013a, Cut slope design recommendations for sub-horizontal hard sedimentary rock units in Ohio, USA: Geotechnical Geological Engineering, Vol. 31, pp. 1207–1219. ADMASSU, Y. AND SHAKOOR, A., 2013b, Computer simulationbased evaluation of rock fall roll-out distances for catchment ditch design in Ohio, USA: Georisk: Assessment Management Risk Engineered Systems Geohazards, Vol. 7, No. 3, pp. 198–208. AMASSU, Y.; SHAKOOR, A.; AND WELLS, N. A., 2012, Evaluating selected factors affecting the depth of undercutting in rocks subject to differential weathering: Engineering Geology, Vol. 124, pp. 1–11. AMERICAN SOCIETY FOR TESTING and MATERIALS (ASTM), 1996, Annual Book of ASTM Standards, Soil and Rock (1): Vol. 4.08, Section 4: ASTM, West Conshohocken, PA. 1000 p. BENNINGTON, J. B., 2002, Eustacy in cyclothems is masked by loss of marine biofacies with increasing proximity to detrital source: An example of central Appalachian Basin, U.S.A. In Hills, L. V.; Henderson, C. M.; and Bamber, E. W. (Editors), Carboniferous and Permian of the World: Canadian Society of Petroleum Geologists, Memoir 19, Ontario, Canada, pp. 12–21. CAMP, M. J., 2006, Roadside Geology of Ohio: Mountain Press Publishing Company, Missoula, MT, 412 p. CHESNUT, D. R., 1981, Marine zones of the Upper Carboniferous of eastern Kentucky. In Cobb, J. C.; Chesnut, D. R.; Hester, N.; and Howard, J. C. (Editors), Coal and Coal Bearing Rocks of Eastern Kentucky: Geological Society of America Coal Division Field Trip, Kentucky Geological Survey, Lexington, KY, pp. 57–66. FERGUSON, H. F. AND HAMEL, J. V., 1981, Valley stress relief in flat lying sedimentary rocks: Proceedings International

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 311–324

Cut Slope Design Symposium Weak Rock, September 21–24, Tokyo, Japan, pp. 1235–1240. HOEK, E.; MARINOS, P. G.; AND MARINOS, V. P., 2005, Characterization and engineering properties of tectonically undisturbed but lithologically varied sedimentary rock masses: International Journal Rock Mechanics Mining Sciences, Vol. 42, pp. 277–285. MARINOS, P. AND HOEK, E., 2001, GSI: A geologically friendly tool for rock mass strength estimation: International Conference on Geotechnical and Geological Engineering (Geoeng 2000), Technomic Publishing Co., Inc. Melbourne, Australia, pp. 1422–1442. NEIMEN, W., 2009, Lessons learned from rates of mudrock undercutting measured over two time periods: Environmental Engineering Geoscience, Vol. 15, No. 3, pp. 117–131. PALMSTROM, A., 1982, The volumetric joint count—A useful and simple measure of the degree of rock jointing: Proceedings, Fourth International Congress of the International Association of Engineering Geology, Delhi, pp. 221–228. PITEAU, D. R. AND MARIN, D. C., 1977, Description of detailed line engineering mapping method: Rock Slope Engineering, Part G. In Federal Highway Administration, Reference Manual FHWA-13-97-208, Portland, OR, 29 p.

ROCSCIENCE, 2006, Determining Input Parameters for Rocfall Analysis: RocFall Software, University of Toronto, Ontario, Canada. SHAKOOR, A., 1995, Slope stability considerations in differentially weathered mudrocks: Reviews Engineering Geology, Vol. 10, pp. 131–138. SHAKOOR, A. AND RODGERS, J. P., 1992, Predicting the rate of shale undercutting along highway cuts: Bulletin Association Engineering Geologists, Vol. 29, No. 1, pp. 61–75. SHAKOOR, A. AND WEBER, M. W., 1988, Role of shale undercutting in promoting rockfalls and wedge failures along Interstate 77: Bulletin Association Engineering Geologists, Vol. 25, No. 2, pp. 219–234. STIMPSON, B., 1981, A suggested technique for determining the basic friction angle of rock surfaces using core: International Journal Rock Mechanics Mining Sciences Geomechanics Abstracts, Vol. 18, pp. 63–65. UDEC, 2014, UDEC Universal Distinct Element Code User’s Guide—Section 2: Itasca Consulting Group, Inc., Minneapolis, MN, 90 p. WYLLIE, D. C. AND MAH, C. W., 2004, Rock Slope Engineering, 4th ed.: Spon Press, London, U.K. 432 p.

APPENDIX 1 Geologic Summary of the 26 Project Sites Site ADA-32-12 ADA-41-15 ATH-33-14 ATH-50-22 BEL-470-6 BEL-70-22 BEL-7-10

Lithology

Slope Type

Limestone underlain by claystone/ mudstone Limestone interlayered with clays tone/mudstone Sandstone Red claystone/mudstone (redbeds) interlayered with limestone Limestone and sandstone interlayered with green shale Sandstone interlayered with shale

Interlayered competent/ incompetent rock Interlayered competent/ incompetent rock Mostly competent rock Interlayered competent/ incompetent rock Interlayered competent/ incompetent rock Interlayered competent/ incompetent rock Interlayered competent/ incompetent rock Mostly competent rock Mostly competent rock Interlayered competent/ incompetent rock Interlayered competent/ incompetent rock Mostly incompetent rock Interlayered competent/ incompetent rock Mostly competent rock

COL-7-5

Limestone and sandstone interlayered with green shale Limestone Limestone Limestone interlayered with claystone/mudstone Sandstone interlayered with shale

FRA-270-23 GUE-22-6.9

Shale Sandstone interlayered with shale

GUE-77-8.2

Sandstone underlain by coal with minor interlayers with siltstone/ shale Claystone/mudstone interlayered Interlayered competent/ with minor limestone incompetent rock

CLA-4-8 CLA-68-6.9 CLE-275-5.2

HAM-74-6.4 HAM-126-12

Claystone/mudstone interlayered with minor limestone

JEF-CR77-0.38 Sandstone interlayered with shale LAW52-11

Sandstone interlayered with shale

LAW-52-12

Sandstone interlayered with shale

LIC-16-28

Sandstone

Geologic Age

Formation or Group Name

Upper and Lower Silurian

Peebles Dolomite

Lower Silurian

Drowning Creek Formation

Upper Pennsylvanian Upper Pennsylvanian

Conemaugh Group Conemaugh Group

Upper Pennsylvanian

Monongahela Group

Lower Permian/Upper Pennsylvanian Upper Pennsylvanian

Dunkard Group Monongahela Group

Upper and Lower Silurian Upper and Lower Silurian Upper Ordovician

Cedarville, Springfield Formation Cedarville, Springfield Formation Kope Formation

Middle/Lower Pennsylvanian

Allegheny and Pottsville Groups

Upper Devonian Middle/Lower Pennsylvanian

Ohio Shale Allegheny and Pottsville Groups

Middle/Lower Pennsylvanian

Allegheny and Pottsville Groups

Upper Ordovician

Interlayered competent/ incompetent rock

Upper Ordovician

Interlayered competent/ incompetent rock Interlayered competent/ incompetent rock Interlayered competent/ incompetent rock

Upper Pennsylvanian

Grant Lake Formation, Miamitown Formation, Fairview Formation Grant Lake Formation, Miamitown Formation, Fairview Formation Conemaugh Group

Middle/Lower Pennsylvanian

Allegheny and Pottsville Groups

Middle/Lower Pennsylvanian

Allegheny and Pottsville Groups

Mostly competent rock

Lower Mississippian

Black Hand Member of the Cuyahoga Formation

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 311–324

323

Admassu and Shakoor Appendix. 1. Continued. Site

Lithology

MEG-33-6 MEG-33-15 MUS-70-11 RIC-30-12.5 STA-30-27 WAS-7-18

Slope Type

Geologic Age

Red claystone/mudstone Interlayered competent/ interlayered with sandstone incompetent rock Red claystone/mudstone Interlayered competent/ interlayered with sandstone incompetent rock Sandstone interlayered with shale Interlayered competent/ incompetent rock Sandstone Mostly competent rock

Formation or Group Name

Upper Pennsylvanian

Monongahela Group

Lower Permian/Upper Dunkard Group Pennsylvanian Middle/Lower Pennsylvanian Allegheny and Pottsville Groups

Upper and Lower Logan and Cuyahoga Formations Mississippian Mostly incompetent rock Middle/Lower Pennsylvanian Allegheny and Pottsville Groups Interlayered competent/ Lower Permian/Upper Dunkard Group incompetent rock Pennsylvanian

Shale with minor siltstone Red claystone/mudstone interlayered with sandstone

APPENDIX 2 Rockfall Roll Out Distances Obtained from RocFall Simulation (Modified from Admassu and Shakoor, 2013-b) Stratigraphic Type

Slope Height (ft) 0.25H:1V Slope Angle

Catchment Ditch Slope

0.5H:1V Slope Angle

Catchment Ditch Slope

1H:1V Slope Angle

Catchment Ditch Slope

1.5H:1V Slope Angle

Catchment Ditch Slope

324

3H:1V 4H:1V 6H:1V Flat 3H:1V 4H:1V 6H:1V Flat 3H:1V 4H:1V 6H:1V Flat 3H:1V 4H:1V 6H:1V Flat

III

100 80 60 40 20 100

60 40 20 100

40 20 100 80 60 40 20

1 1 5 29 3 6 17 46 14 24 29 41 14 18 21 26

50 68 76 88 39 31 61 76 30 35 100 49 20 30 31 36

44 50 56 66 41 26 51 59 19 26 36 45 20 20 25 30

54 54 74 109 51 54 66 116 39 46 54 127 41 36 51 85

46 56 56 90 45 51 60 100 44 45 55 76 30 35 47 96

16 21 21 41 19 24 28 49 19 22 28 47 19 27 29 54

1 1 3 23 1 6 14 36 14 19 24 36 11 15 16 19

1 1 4 19 2 5 11 27 10 15 19 21 11 11 14 14

1 1 2 15 1 3 17 19 6 10 14 19 7 7 11 12

1 1 2 8 1 2 4 12 4 6 6 11 4 4 5 6

69 75 86 115 69 40 90 101 65 64 145 95 35 35 35 53

15 21 25 27 16 12 19 23 13 14 17 19 9 10 11 12

11 12 14 16 11 9 14 15 9 9 11 12 5 6 5 6

71 82 104 140 72 85 94 117 74 75 95 130 77 77 83 131

13 14 20 24 12 12 17 28 12 13 16 31 13 13 15 31

1 1 7 35 3 7 18 60 26 37 50 100 37 47 60 138

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 311â&#x20AC;&#x201C;324

1 1 7 27 2 7 17 50 21 34 42 86 30 40 50 91

1 1 4 22 2 7 16 45 15 26 35 57 25 31 39 68

1 1 3 16 2 6 11 29 12 17 22 50 17 21 26 54

1 1 2 9 1 3 7 19 7 9 12 29 10 11 15 29

Understanding Karst Leakage at the Kowsar Dam, Iran, by Hydrogeological Analysis MORTEZA MOZAFARI1 EZZATOLLAH RAEISI2 Department of Earth Sciences, College of Sciences, Shiraz University, 71454, Shiraz, Iran

Key Terms: Dam, Reservoir, Karst Aquifer, Leakage Problem, Grout Curtain

ABSTRACT The Kowsar Dam is constructed on the Kheyrabad River at the northern limb of the Duck Anticline, close to its NW plunge, SW Iran. The dam is built on the karstic Asmari Formation, and the reservoir is in direct contact with this formation from the dam body to the upstream impermeable Gachsaran Formation. After reservoir impounding, several new springs emerged from the Asmari Formation, adjacent to the old small springs at the southern limb of the Duck Anticline. The discharge of these downstream springs was not reduced despite grout curtain treatment works. The main water leakage route is not below or through the grout curtain, as shown by considering rock permeability in pilot and check holes, cement consumption in grouting boreholes, borehole water levels, spring locations, and discharge of the dam galleries. The Duck Anticline is hydrogeologically connected to the adjacent Dill and Pahn Anticlines, comprising the Asmari Karst Aquifer, which is divided into three karst sub-aquifers. The general flow direction of karst water in each sub-aquifer was determined using hydrogeological analysis and water balance calculations. Two alternative models are proposed for water flow at the northern limb of the Duck Anticline. The main leakage route toward the downstream springs is most probably through a relict karst conduit system, developed along bedding planes of the Asmari Formation at the NW plunge of the anticline, according to the first proposed model. In this case, the water leakage can be significantly reduced by extension of the grout curtain further into the upstream impermeable Gachsaran Formation. INTRODUCTION Karst is defined as a terrain with distinctive landforms and hydrology that is developed on 1 2

Email address: mmozafari@shirazu.ac.ir. Corresponding author email: e_raeisi@yahoo.com.

especially soluble rocks such as limestone, marble, dolomite, halite, and gypsum (Ford and Williams, 2007). Karst covers approximately 20 percent of Earth’s surface and is characterized by fluted and pitted rock surfaces, sinking streams, enclosed depressions, caves, sinkholes, shafts, springs, and subsurface drainage systems (White, 1988; Ford and Williams, 2007). Karst aquifers are characterized by three types of porosity: matrix (or inter-granular) porosity, fracture porosity, and conduits (White, 1969, 1977; Ford and Williams, 2007). Depending on whether the flow path is through matrix, fractures, conduits, or a combination, there are different types of flow systems in karst aquifers (Martin and Screaton, 2001). Diffuse flow systems occur predominantly within matrix and fracture porosity, while conduit flow occurs within conduits (Pitty, 1968; Paterson, 1979). Development of conduits within karst aquifers results in high permeability and allows rapid transfer of groundwater (Martin and Screaton, 2001). Due to a variety of intrinsic geological and hydrological features, karst systems are among the most vulnerable settings in the world for design and development of engineering projects, including dams (Gutierrez et al., 2014; Parise et al., 2014). Leakage from karst dam sites has been reported all over the world. The difficulties involved in constructing a dam on a karstified bedrock were first documented at the Hales Bar Dam, which was built on the Tennessee River between 1905 and 1913 (Donnelly et al., 2009). Milanovic´ (2004) reviewed in detail the leakage problems of several dams constructed on karst regions all over the world; most of them work successfully, but a few have problems with unacceptably heavy leakage from the reservoir. In many dams, huge leakages were reduced after complicated sealing programs. Leakage appears at preexisting springs or as new resurgences. Leakage routes are often karst conduits developed along bedding planes, joints, and/or their intersections. The leakage rate can vary from a few liters to several cubic meters per second. Even with extensive grouting, sometimes leakage is serious from the beginning

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 325–339

325

Mozafari and Raeisi

Figure 1. Location of the Kowsar Dam site, hydrogeological map of the study area, and the conceptual model of general flow direction in three sub-aquifer areas, SP, NDD, and SPNDD.

of impoundment and may increase greatly with time (Palmer, 1988). Most papers on the topic focus on remediation methods. An overview of the articles shows that one or several of the following techniques have been used to identify the source(s) and route(s) of leakage: response of spring discharge and borehole water level to reservoir water-level changes (Sahuquillo, 1985; Pantzartzis et al., 1993; Turkmen, 2003; Unal et al., 2007; Bonacci and Rubinic, 2009; and Bonacci and Bonacci, 2013), tracing tests (Quinlan, 1985; Turkmen et al., 2002; and Mozafari et al., 2011), isotopic studies (Hansen and Teter, 1970; Gu¨nay et al., 1995; Crilley and Torak, 2002; and Laksiri, 2007), hydrochemistry (Gu¨nay et al., 1995; Qingzhi et al., 1998; Montoroi et al., 2001; Ghobadi et al., 2005; Torak et al., 2006; and Al-Omosh et al., 2008), exploratory drilling, rock permeability, and grout curtain characterization (Zogovic´, 1993; Jarvis, 2003; Turkmen, 2003; Milanovic´, 2004; Schaefer, 2009; Milanovic et al., 2010; and Mozafari et al., 2011), and geophysical methods (Al-Saigh et al., 1993; Ginther and Charlton, 2009; and Bedrosian et al., 2012). Mohammadi et al. (2007) proposed three steps for leakage study at a karst dam site, including: (1) preparation of the hydrogeological map; (2) delineation and func-

326

tional analysis of the karst system by means of structure and functioning approaches; and (3) assessment of the leakage potential using the results of steps 1 and 2 and determination of the most probable leakage zones. The concrete gravity Kowsar Dam, with a height of 144 m and a reservoir capacity of 580 million cubic meters (MCM), was constructed by the Iranian Ministry of Power in SW Iran (Figure 1) to annually supply 200 and 300 MCM of drinking and irrigation water, respectively. The dam is built on the Kheyrabad River, on the karstified Asmari Formation at the northern limb of the Duck Anticline (Figure 1). Here, the Kheyrabad River, with an average discharge of 21.5 m3/s, flows almost perpendicular to the anticline axis near NW plunge, developing the V-shape Kheyrabad Valley in the Asmari Formation. The reservoir normal water level (RNWL) is 625 m above sea level (a.s.l.), about 125 m above the Kheyrabad River floodplain. The reservoir is in direct contact with the Asmari Formation vertically from the riverbed to the RNWL, and on the sides from the dam body to the upstream impermeable Gachsaran Formation. Soon after impounding, leakage occurred in the downstream preexisting springs and at several new springs, located on the Asmari Formation at the

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 325–339

Understanding Karst Leakage, Kowsar Dam, Iran

southern limb of the Duck Anticline. The maximum total leakage stabilized at 5.61 m3/s at the RNWL. The main objective of this paper is to identify the main leakage route of the reservoir by means of hydrogeological analysis of the Asmari Karst Aquifer (AKA), rock permeability tests in pilot and check holes, cement consumptions in grouting boreholes, borehole water levels, springs locations, and discharge of the dam galleries. The paper explains the role of a general hydrogeological study and evaluation of relict karst development mechanisms in delineation of the reservoir leakage route. The results would be useful to determine potential leakage routes in other karst dam sites in Iran and elsewhere. HYDROGEOLOGICAL SETTING The study area is located in the Simply Folded Zone of the Zagros Orogenic System in SW Iran, where the sedimentary rocks have been folded into parallel NW-SE anticlines and synclines since the Miocene (Falcon, 1961). In the Simply Folded Zone, most of the karst formations are sandwiched between two impervious formations, forming independent karst aquifers (Raeisi, 2008). The study area is composed of five anticlines parallel to the general structure of the Simply Folded Zone, namely, the Duck, Dill, Pahn, Mish, and Khami Anticlines (Figure 1). The structural characteristics and stratigraphy of the Zagros sedimentary sequence have been described in detail by Stocklin and Setudehnia (1977) and Alavi (2004). The geological formations in the study area in decreasing order of age are (Figure 1): Khami Group carbonates and marls (Jurassic to Lower Cretaceous), Bangestan Group carbonates and shale (Upper Cretaceous), Pabdeh-Gurpi shale and marl (Cretaceous–Tertiary), Asmari Formation, marly limestone and marl (Oligocene–Miocene), Gachsaran marl and evaporites (Tertiary), and Quaternary alluvium. The distribution of the geological formations in the Dill, Duck, Pahn, Mish, and Khami Anticlines is presented in Figure 2. Due to action of thrust on the Mountain Front Fault (MFF) (Figure 1), the older geological formations of the Mish and Khami Anticlines are located next to the younger formation of the Dill Anticline. The Khami Group constitutes the core of the Mish and Khami Anticlines, overlain by the exposed Bangestan Group. The Pabdeh-Gurpi Formation outcrops only at the Mish Anticline and makes an important aquiclude unit beneath the Asmari Formation in the core of the Duck, Dill, and Pahn Anticlines. The Asmari Formation outcrops at the top of the Dill, Duck, and Pahn Anticlines where the impermeable overlying Gachsaran Formation has been eroded, and it is

mostly exposed at the foot of the anticlines or buried under a thin alluvium on the adjacent plains. The Gachsaran Formation is divided into salt and nonsalt equivalents (Bahrudi and Koyi, 2004). The Salt Gachsaran Formation outcrops at the study area. Stocklin and Setudehnia (1977) divided it into seven members based on a type section from wells in the Gachsaran oil field. Member 1, which consists of about 40 m of inter-bedded anhydrite and limestone associated with shale, is known as an important sealing unit over the Asmari Formation oil reservoirs. The Kowsar Dam is built on the Asmari Formation at the northern limb of the Duck Anticline (Figure 3). Here, the Asmari Formation is classified into three members: Lower Asmari (LAs), Middle Asmari (MAs), and Upper Asmari (UAs) (Figure 3), based on detailed geological mapping and borehole logs (Fars Regional Water Authority, 1997b). The LAs unit, which extends vertically from 70 m above and to at least 100 m below the valley floor, based on borehole log data, consists of 4- to 8-m-thick beds of dense crystalline massive limestone with rare marly limestone inter-beds. The MAs unit is composed of more than 100 m of 1.5- to 3-m-thick beds of crystalline limestone inter-bedded with marl and marly limestone layers (marls represent about 7 to 8 percent of the MAs unit). A 1.5-m-thick marly layer is located near the lower boundary of the MAs unit (Figure 3). The UAs unit consists of 150 m of medium- to thinbedded, crystalline limestone with relatively more inter-bedded marly limestone and marls. Caves up to 0.8 m were observed in the exposed LAs and MAs units, filled partly with clay materials. In addition, a few open caverns were discovered in the Asmari Formation during borehole drilling and gallery excavation. At the dam axis, the lithologic beds dip very gently from the left abutment to the right abutment, but in the right abutment, they have a sleeper slope. In addition, by moving from the dam axis toward the reservoir, lithologic beds dip upstream. Before the dam construction, the groundwater level measured in the six pilot holes along the dam axis was located about 6 to 10 m below the riverbed (Figure 4). There was no spring at the northern limb of the Duck Anticline, and just a few small springs emerged on both the banks of the Kheyrabad River, from the MAs unit in the southern limb of the anticline, about 1.2 km downstream of the dam body (Figure 3). There, a near-vertical minor fault (F1 fault in Figure 3) extended from the springs toward the UAs unit. Unfortunately, there are no data for discharge of the springs before the dam construction, but a few measurements indicated that in addition to the springs, there was about 0.5 m3/s direct seepage

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 325–339

327

Mozafari and Raeisi

Figure 2. Distribution of geological formations at the: (a) Pahn, Dill, and Mish Anticlines; b) Dill and Mish Anticlines, and (c) Duck Anticline (modified after NIOC, 1959). Cross sections A-A9, B-B9, and C-C9 are illustrated in Figure 1.

from the AKA into the riverbed at the southern limb of the anticline. Grout Curtain The water-tightness system of the Kowsar Dam includes a hanging grout curtain in the LAs and lower parts of the MAs units at both abutments (Figures 3 and 5). The hanging grout curtain was designed using the results of a FILTER software model based on measured rock permeabilities obtained from the eight pilot boreholes (Gidrosproekt, 1996) (Figure 4). The model predicted seepage less than 0.4 m3/s at the RNWL, by considering a 65- to 150-m-deep grout curtain with 280 and 230 m extensions into the right and left abutments, respectively. The as-built grout curtain is 100 m longer in the right abutment and 30 m deeper than the pre-built model (Tuzhikhin and Kolichko, 2001). The grout curtain was constructed in one, two, or three rows with a distance interval of

328

1.7 m and by a three-level gallery system. The final grouting borehole spacing was in average 1.5 m, but it was reduced at the zones with high cement consumption. The diameter of the grouting boreholes was 0.076 m, and grouting was carried out in 5-m-long sections. A two-level drainage gallery system was designed and built parallel and 25 m downstream of the grout curtain. The water transport gallery was designed and built at the right abutment. METHODOLOGY The geologic map of the dam site (Figure 1) is based on 1:100,000 (National Iranian Oil Company Exploration and Production [NIOC], 1959) and 1:5,000 (Fars Regional Water Authority, 1997a) geologic maps. Reservoir and borehole water levels, hydrochemistry, and discharge of springs were measured by the Fars Regional Water Authority (2010). Reservoir and boreholes water levels were

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 325â&#x20AC;&#x201C;339

Understanding Karst Leakage, Kowsar Dam, Iran

Figure 3. Hydrogeological map of the Kowsar Dam site and the proposed main leakage route.

measured monthly during a 7 year period. The discharge of every spring was measured individually by volumetric and weir methods for about 2 years. Leakage from the dam galleries was measured using weirs monthly for 7 years. A hydrometric station was constructed at the end of Kheyrabad valley, downstream of the springs (Figure 3). Here, the total

discharge of springs and dam galleries, and the AKA direct seepage into the riverbed were measured when all of the dam outlets were closed. The water electrical conductivity (EC) was measured monthly in the reservoir and largest springs on both banks of the Kheyrabad River during the spring to summer 2009. The following equation was used to determine the

Figure 4. Longitudinal section along the dam axis and measured rock permeabilities and groundwater level in the pilot holes.

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 325â&#x20AC;&#x201C;339

329

Mozafari and Raeisi

Figure 5. Plan view of the dam body, galleries, and grout curtain, and the layout of the as-built grout curtain.

mean annual recharge of each aquifer during one hydrological year (Q) in m3/s: Q 5 API/t,

(1)

where A is the aquifer area (m2), P is the 20 year mean annual rainfall (m), I is the recharge coefficient, and t is the 1 year period in seconds. The mean rainfall over the aquifer surface was calculated based on the relationship between the elevations and 20 year mean rainfall from adjacent climatologic stations. ANALYSIS The AKA includes the Asmari Formation at the Duck, Dill and Pahn Anticlines, which is underlain by the impermeable Pabdeh-Gurpi Formation, surrounded by the impermeable Gachsaran Formation and thin alluvial deposits; it is limited to the NE by the MFF thrust fault (Figures 1 and 2). The AKA and the Bangestan Group are separated by the 800m-thick impermeable Pabdeh-Gurpi Formation (Figure 2A and B). The impermeable Member 1 of the Gachsaran Formation prevents any hydraulic communication between the AKA and the adjacent Mish and Khami Anticlines. The Gachsaran Formation is covered by alluvial deposits in some parts

330

around the AKA. With the exception of the Lishter alluvium at the southwest of the Pahn Anticline, which was exploited by more than twenty 30- to 75m-deep pumping wells (Figure 1), the other overlying alluvium deposits are thin. The water level in the AKA is most probably lower than the contact of the Gachsaran Formation and those alluvium deposits (Figure 2); therefore, karst water cannot flow into them. In addition, since alluvial aquifers are often exploited by pumping wells in Iran, the lack of pumping wells in those alluviums suggests that there is no water recharge from the AKA. The AKA recharge source is direct rainfall on the karst aquifer body. There is no recharge from the surrounding impermeable Gachsaran Formation or alluvium because those are located at lower elevations than the AKA outcrops (Figure 2). Recharge from the southern limbs of the Mish and Khami Anticlines to the AKA is unlikely because they are structurally and hydraulically disconnected by the impermeable Gachsaran Formation and fractured zone of the MFF thrust fault (Figure 2A and B). In addition, the infiltrated water of the southern limb of the Mish Anticline discharges into the downstream Emamzadeh Jafar Aquifer, located about 10 km to the SE of the study area (Sharifi, 2009). The AKA water discharge zones are the Kheyrabad River and the

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 325â&#x20AC;&#x201C;339

Understanding Karst Leakage, Kowsar Dam, Iran Table 1. Annual recharge of the AKA sub-aquifers and Lishter Aquifer, calculated based on Eq. 1. Aquifer or Sub-Aquifer SP NPSDD NDD Lishter1

Area (km2)

Recharge Coefficient

Recharge by Rainfall (MCM)

25 69 49 36

0.4 0.4 0.4 0.2

6.3 21.2 15.3 4

Lishter alluvium, controlling general flow direction in the karst aquifer. The proposed discharge zones are confirmed based on the following reasons: (1) There are no springs at the contact of the AKA and the surrounding Gachsaran Formation and alluviums. (2) Before the dam construction, there were a few springs on both banks of the Kheyrabad River at the southern limb of the Duck Anticline. In addition, a few measurements indicated that there was about 0.5 m3/s of direct seepage from the AKA into the Kheyrabad River at the southern limb of the Duck Anticline. (3) The underlying impermeable PabdehGurpi Formation prevents any downward water flow from the AKA into the Bangestan Group. (4) The surrounding impermeable Gachsaran Formation and the fractured zone of the thrust MFF prevents any water flow from the AKA to the adjacent anticlines. (5) The contact elevations of the Asmari Formation and the Gachsaran Formation are located about 150 to 800 m above the riverbed all around the AKA, except at the Kheyrabad Valley inlet and outlet. The transverse Kheyrabad River is the main base level of erosion and the main discharge zone of the AKA because it has the lowest elevation within the AKA. Based on the hydrogeological setting and the catchment areas of the discharge zones, the AKA is divided into three sub-aquifer areas (Figure 1): the southern limb of the Pahn Anticline sub-aquifer (SP sub-aquifer), which discharges into the Lishter Aquifer; the northern limbs of the Dill and Duck Anticlines sub-aquifer (NDD sub-aquifer), which discharges into the Kheyrabad River; and the northern limb of the Pahn and southern limb of the Dill and Duck Anticlines sub-aquifer (NPSDD subaquifer), which discharges into the Kheyrabad River. Since there is no spring discharge measurement and therefore no exact data on discharge of the AKA before the dam construction, the discharge of the AKA sub-aquifers during one hydrological year was estimated based on Eq. 1 and using the estimated recharge coefficient according to previous studies on the karst regions of Iran (Pezeshkpour, 1991; Karst Research Centre of Iran, 1993; Karimi, 2003; and Karimi et al., 2005) (Table 1). A conceptual model of the general flow direction is proposed for the three AKA sub-aquifers (Figure 1).

Discharge by Springs, Seepage, or Wells (MCM) â&#x20AC;&#x201D; 17 46 9.1

The elevation of the impermeable Pabdeh-Gurpi Formation under the crest of the Pahn and Dill Anticlines is higher than the contact of the Asmari Formation and the surrounding Gachsaran Formation or alluvium at the foot of anticlines (Figure 2B), disconnecting the hydraulic connectivity of both limbs. Therefore, every limb becomes an independent sub-aquifer with catchment area limited to the crest of the anticline. In each sub-aquifer, the water flows initially along bedding planes until it reaches the foot of the limb, where it flows parallel to the strike of the folds. This hypothetical general flow direction along the fold strike has been demonstrated by Ashjari and Raeisi (2006) in most of the Zagros anticlines. In the SP sub-aquifer, water initially flows along bedding planes and finally discharges into the adjacent Lishter Aquifer (Figure 1). The Lishter Aquifer annual recharge by rainfall is about 4 MCM, while its annual discharge by pumping wells is 9.1 MCM (Table 1). Therefore, the missing 5.1 MCM/yr must be supplied by inflow from the adjacent SP sub-aquifer, because: (1) the Lishter alluvium is surrounded by the impermeable Gachsaran Formation in all directions except adjacent to the Pahn Anticline; and (2) the Pahn Anticline is the nearest karstic limestone to the Lishter Aquifer, and its elevation is higher than the Lishter alluvium. Based on Eq. 1, the required karst catchment area for providing 5.1 MCM/yr water is about 16.5 km2. The southern limb of the Pahn Anticline (SP sub-aquifer) is a feasible catchment area, since its exposed area is about 25 km2. The contact elevation of the Pabdeh-Gurpi Formation with the Asmari Formation in the SE plunge of the Dill Anticline is about 650 m higher than in the NW plunge of the Duck Anticline. Therefore, at the NPSDD and NND sub-aquifers, a main conduit system has probably developed parallel to the fold strike, starting from the SE plunge of the Dill Anticline and going toward the NW plunge of the Duck Anticline. The NPSDD sub-aquifer collects groundwater coming from the northern limb of the Pahn Anticline and southern limb of the Dill Anticline. This karst water flows along the anticline strike toward the southern limb of the Duck Anticline and finally emerges into the Kheyrabad River

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 325â&#x20AC;&#x201C;339

331

Mozafari and Raeisi

Figure 6. Proposed models of the general flow direction inside the Asmari Formation at the northern limb of the Duck Anticline.

through the springs on the left bank or direct seepage into the riverbed. In the NDD sub-aquifer, groundwater flows along the foot of the northern limb of the Dill Anticline toward the northern limb of the Duck Anticline. Two alternative schematic models are proposed for the water flow route at the northern limb of the Duck Anticline (Figure 6). In the first model, water flows through a conduit system developed most probably along bedding planes at the NW plunge and finally emerges into the Kheyrabad River as the right bank springs. The NDD subaquifer water may be flowing below the Kheyrabad River and emerging at the right bank springs, since the groundwater level in the dam site was 6 to 10 m lower than riverbed before dam construction. The fact that there were no springs in the Kheyrabad Valley at the northern limb supports this model. Oberlander (1965) analyzed the origin of drainages and streams at the Zagros Range in detail and showed a model of Zagros stream superposition. The Kheyrabad River was most probably superimposed upon the newly exposed crest of the Duck Anticline, eroding it and leading to the formation of the Kheyrabad valley. Once river incision started, and bearing in mind that the riverbed at the southern limb was lower than at the northern limb, part of the river water may have seeped into the joints and bedding planes to develop a new karst conduit system in the NW plunge, ending at the right bank springs. As the river continued cutting down into

332

the bedrock and water base level continued to drop, new conduits formed, others enlarged, and the highest relict conduits were dewatered and abandoned. The impermeable Gachsaran Formation prevents emergence of karst water from the plunge nose. At the present time, the elevation of the downstream springs is at least 140 m below the contact elevation of the Asmari Formation and the Gachsaran Formation at the NW plunge nose of the Duck Anticline. Two cavities with a height of 3 m were observed in borehole TD9 (Figure 3) at elevations of 508 and 561 m a.s.l., probably confirming the relict conduit development in the proposed model route. In the alternative model, the karst water flows perpendicular to the Duck Anticline axis, parallel to the Kheyrabad River and emerges through the downstream springs (Figure 6). Hydraulic connection between the northern and southern limbs of the Duck Anticline is possible since the elevation of the downstream springs is higher than the contact of the Asmari Formation and the PabdehGurpi Formation at the anticline axis (Figure 2C). This second flow route model seems to be unlikely since water should flow perpendicular to the limestone bedding planes, and marly inter-beds up to 1.5 m thick (Figure 3) may create an impermeable barrier and prevent conduit development perpendicular to the bedding plane. In addition, some of the thick nonkarstified limestone layers of the LAs unit may create an impermeable barrier to water and prevent

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 325â&#x20AC;&#x201C;339

Understanding Karst Leakage, Kowsar Dam, Iran

Figure 7. Time series of the reservoir water level and (a) discharges of the hydrometric station, springs, galleries, and the AKA direct seepage into the riverbed through 2004; and (b) discharge at the hydrometric station through 2012.

conduit development perpendicular to bedding plane, an idea supported by the lack of spring or seepage at the LAs unit. Leakage History and Remedial Efforts Reservoir impounding started in November 2002. By the time that the reservoir water level reached 546.1 m a.s.l. in February 2003, the discharge of the downstream springs located in the southern limb of the Duck Anticline started to increase, and several new springs emerged adjacent to them. No new spring or seepage was observed in the northern limb of the anticline, and water just emerged from some of the dam galleries, especially from those located at the right abutment. As explained already, the total discharge of the downstream springs, dam galleries, and the AKA direct seepage into the riverbed was measured at the hydrometric station when all of the dam outlets were closed. In addition, the discharge of

every spring was measured individually until the end of 2004. Figure 7A presents the time series of the total discharge of the hydrometric station, right bank springs, left bank springs, dam galleries, and the AKA direct seepage into the riverbed until the end of 2004. The AKA direct seepage into the riverbed was estimated by subtracting the discharges of the springs and dam galleries from the discharge of the hydrometric station. Results show that the discharges of the springs and dam galleries correlated with the reservoir water-level changes, but the AKA direct seepage into the riverbed was most affected by the seasonal variations. At the reservoir water level of 613 m a.s.l, the discharge of the hydrometric station was 3.4 m3/s, and the apportioning of the right bank springs, left bank springs, galleries, and the AKA direct seepage was 46, 20, 15, and 19 percent, respectively. The time series of the reservoir water level and the discharge of the hydrometric station for a period of 10 years are presented in Figure 7B. The figure shows

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 325â&#x20AC;&#x201C;339

333

Mozafari and Raeisi

Figure 8. Rock permeabilities in the check holes.

a tight relationship between the reservoir water level and the changes in the discharge of the hydrometric station. The discharge of the hydrometric station reached 5.61 m3/s, more than one fourth of the mean annual discharge of the Kheyrabad River (21.5 m3/s), at the RNWL in 2005. In order to reduce reservoir water leakage, watertightness treatments were applied by constructing a new 90-m-long and 100-m-deep grout curtain along the transport gallery at the right abutment and additional grouting in some parts of the as-built grout curtain (Figure 5). The treatment works were successful in reducing nearly 60 percent of discharge from the dam galleries (Fars Regional Water Authority, 2010), but they did not affect the discharge of the downstream springs. The discharge of the hydrometric station was about 4 m3/s at a reservoir water level of 623.5 m a.s.l. (1.5 m below the RNWL), in May 2012. RESULTS AND DISCUSSION Water from the reservoir is leaking through the downstream springs, mainly through those ones located on the right bank of the Kheyrabad River. The potential leakage route could be below and/or through the grout curtain, or through a relict conduit system at the NW plunge, based on the first proposed model of the general flow direction in the northern limb of the Duck Anticline. The grouting work quality was evaluated based on the rock permeability in the pilot and check holes (Figures 4 and 8) and cement consumption in different stages of grouting. Before grouting, rock permeability was more than 5 Lu in 92 percent of the measured sections in the pilot holes, and it was even more than 100 Lu in 13 percent of sections. The permeability was reduced

334

significantly by grouting, since after grouting, it was less than 5 Lu, 5 to 10 Lu, and 10 to 18 Lu in 76, 21, and 3 percent of the measured sections of the check holes, respectively. The total cement consumption in the grouted sections is presented on Figure 9. The cement consumption decreased by stages of grouting; the average cement take was more than 1000 kg/m at the first stage, but it was decreased to less than 40 kg/m at the last stage. In spite of high permeability before grouting, the grouted sections in the middle part of the left abutment had low total cement consumptions (Figure 9) due to transfer of the grouting mix during grouting of the upper sections. The grout curtain works properly, based on the borehole water levels, discharge of the dam galleries, and location of the springs. The difference between boreholes water levels upstream and downstream of the grout curtain was at least 60 m at the RNWL (Figure 10). The linear regression coefficient between the reservoir water level and borehole water levels upstream of the grout curtain was 0.99, while at the downstream, it was less than 0.75, and mainly less than 0.74 at the left and right banks, respectively. The total discharge of the dam galleries was about 0.5 m3/s at the RNWL, which is not far from the predicted leakage by the FILTER modeling (0.4 m3/s). The Kheyrabad River is parallel to the dam axis about 300 m downstream of the grout curtain (Figure 3), but no spring or seepage was observed on the river banks at this area. The dam outlets were closed several times, and the main leakage zone was at the downstream springs area. These evidences confirm that the main leakage route is not below or through the grout curtain. The main leakage route is most probably from a relict conduit system that developed along bedding planes of the Asmari Formation at the NW plunge of the Duck Anticline (Figure 3). The reservoir is in

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 325â&#x20AC;&#x201C;339

Understanding Karst Leakage, Kowsar Dam, Iran

Figure 9. Total cement consumption in the grouted sections of the grout curtain.

direct contact with the LAs, Mas, and UAs units. However, since the hydraulic relation between the reservoir and the underlying LAs unit is disconnected by the grout curtain (Figures 3 and 5), reservoir water can leak into the MAs and UAs units and then flow through the relict conduit system and finally emerge as the right bank springs. The karst water in the UAs and MAs units could be directed toward the springs by the F1 fault. Borehole P38 is located 330 m from the reservoir, in the UAs and MAs units and inside the proposed leakage route (Figure 3). The water level of this borehole has a similar response to reservoir waterlevel variations (Figure 10). The linear regression

coefficient between the water levels of the reservoir and borehole P38 is 0.99, indicating a tight correlation and therefore a strong hydraulic connection between the reservoir and the proposed leakage route. The borehole P38 water level was 12.8 m lower than RNWL, and the reservoir water flowed toward it with a hydraulic gradient of 4 percent. The water EC of the largest spring on the right bank was about 80 to 115 ms/cm higher than the reservoir during March to July 2009, when the reservoir water level was near 595 m a.s.l. (Figure 11). The EC of the reservoir water must be increased by moving through the proposed leakage route toward the downstream springs.

Figure 10. Borehole water levels upstream and downstream of the grout curtain.

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 325â&#x20AC;&#x201C;339

335

Mozafari and Raeisi

Figure 11. Water electrical conductivity (EC) of the reservoir and largest springs located on both the banks of the Kheyrabad River.

The total discharge of the right bank springs was 46 MCM during 2004 (Table 1), which is much more than the expected annual discharge of the NDD subaquifer (15.3 MCM). Even considering that all NDD sub-aquifer water emerged at the right bank springs after reservoir impoundment, the reservoir water leakage through the right bank springs is calculated to be 30.7 MCM. The total annual discharge of the left bank springs was 16.6 MCM during 2004 (Table 1). Two assumptions can be proposed for the water source of these springs. Based on the first assumption, the spring water is completely supplied by the karst water of the NPSDD sub-aquifer area. The expected annual discharge of the NPSDD sub-aquifer is 21.2 MCM, i.e., 4.6 MCM higher than the annual discharge of the left bank springs (16.6 MCM), which can be seeping directly into the riverbed. The lower water EC of the largest spring on the left bank compared to the reservoir (Figure 11) supports this assumption. Based on the second assumption, the source of the springs could be mixing of the NPSDD sub-aquifer karst water and the reservoir water. At the springs area, the limestone layers are repeated at both the banks of the Kheyrabad River, since they are eroded almost perpendicular to the bedding planes by the Kheyrabad River. Some reservoir water could be transferred from the right bank to the left bank of the river by a probable conduit

336

system in the limestone layers below the riverbed. The crossing of a karst conduit below a riverbed without any hydrogeological relationship has been proved by a dye tracing test at the Doosti Dam site by Mozafari et al. (2011). The transferred reservoir water could be mixed with the water of the NPSDD sub-aquifer and emerge as the left bank springs. The EC of the transferred reservoir water can be reduced by mixing with the karst water of the NPSDD subaquifer. CONCLUSIONS The reservoir water leakage route toward the downstream springs is not below or through the grout curtain. The proper function of the grout curtain is evidenced by the evaluation of rock permeability in the pilot and check holes, cement consumptions of the grouting boreholes, borehole water levels, springs locations, and discharge of the dam galleries. Understanding the hydrogeology of the AKA acts as an effective tool to determine water leakage route toward the downstream springs. The hydraulic relation between the AKA and adjacent aquifers, except the Lishter Aquifer, is disconnected from the underlying impermeable Pabdeh-Gurpi Formation, surrounding impermeable Gachsaran Formation, and function and fractured zone of the MFF thrust fault. The AKA recharge

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 325â&#x20AC;&#x201C;339

Understanding Karst Leakage, Kowsar Dam, Iran

source is direct rainfall on the karst aquifer, and there is no recharge from the surrounding alluvium, impermeable Gachsaran Formation, and adjacent Mish and Khami Anticlines. The AKA consists of three sub-aquifer areas, SP, NPSDD, and NDD. While the SP sub-aquifer area is discharged into the Lishter alluvial aquifer, the Kheyrabad River is the main discharge zone of the NPSDD and NDD subaquifers, controlling the general flow direction in these sub-aquifers. The proposed conceptual model of general flow direction in the three AKA sub-aquifers is in agreement with the general flow direction in most of the anticlines of the Zagros Mountain Range in Iran. A relict conduit system has likely been developed along bedding planes in the NW plunge of the Duck Anticline by karst water flow of the NDD sub-aquifer and action of the Kheyrabad River. The reservoir water leakage route is mainly through this relict conduit system. By raising the reservoir water level, water seeps into the Asmari Formation at the northern limb, converges into the relict conduit system in the MAs and UAs units, flows along the plunge of the anticline, and finally emerges through the right bank springs at the southern limb. The as-built grout curtain and the treatment works in the right abutment were not able to disconnect the hydraulic relation between the reservoir and the envisioned relict conduit system. The hydrometric station mean annual discharge is about 95.6 MCM. Considering 37.5 MCM annual discharge of the NDD and NPSDD sub-aquifers, the annual water leakage of the reservoir is calculated to be 58.1 MCM. This volume of water is about 20 and 30 percent of the proposed annual irrigation and drinking water, respectively. The irrigation diversion canal is a few kilometers downstream of the springs; therefore, the leakage water is used for agricultural purposes. The Kowsar Dam is located in an arid and semiarid part of Iran, and it is the most reliable source of drinking water for the downstream arid areas. Because of low precipitation in the Kheyrabad River catchment area, provision of proposed irrigation and even drinking water could be associated with an uncertainty during drought periods. Therefore, prevention of the current reservoir leakage seems to be reasonable. Construction of a new grout curtain perpendicular to the bedding planes in the upper parts of the MAs and UAs units and toward the upstream impermeable Gachsaran Formation at the right abutment would be effective in reducing of the water leakage amount. Dye tracing test is recommended for detection of specific leakage route and subsequently the correct path for the probable additional grout curtain.

ACKNOWLEDGMENTS The authors gratefully appreciate the sincere cooperation of the Fars Regional Water Authority of Iran for providing useful data. The authors also would like to thank Mr. Ahmadi for information and Dr. J.P. Galve, Dr. J. Guerrero, and Prof. F. Gutie´rrez for their reviews and comments. REFERENCES ALAVI, M., 2004, Regional stratigraphy of the Zagros FoldsThrust Belt of Iran and its proforeland evolution: American Journal of Science, Vol. 304, pp. 1–20. AL-OMOSH, H.; AL FARAJAT, M.; AND ZUNIC, F., 2008, Leakage in Bayer Dam in Jordan: Its causes and consequences, Jordan: Journal of Civil Engineering, Vol. 2, No. 4, pp. 363–375. AL-SAIGH, N. H.; MOHAMMED, Z. S.; AND DAHHAM, M. S., 1993, Detection of water leakage from dams by self-potential method: Engineering Geology, Vol. 37, No. 2, pp. 115–121. ASHJARI, J. AND RAEISI, E., 2006, Influences of anticlinal structure on regional flow, Zagros, Iran: Cave and Karst Science, Vol. 68, No. 3, pp. 118–129. BAHRUDI, A. AND KOYI, H. A., 2004, Tectonic-sedimentary framework of the Gachsaran Formation in the Zagros foreland basin: Marine and Petroleum Geology, Vol. 21, pp. 1295–1310. BEDROSIAN, P. A.; BURTON, B. L.; POWERS, M. H.; MINSLEY, B. J.; PHILLIPS, J. D.; AND HUNTER, L. E., 2012, Geophysical investigations of geology and structure at the Martis Creek Dam, Truckee, California: Journal of Applied Geophysics, Vol. 77, pp. 7–20. BONACCI, O. AND RUBINIC J., 2009, Water losses from a reservoir built in karst: The example of the Boljuncica reservoir (Istria, Croatia): Environmental Geology, Vol. 58, pp. 339–345. BONACCI, T. R. AND BONACCI, O., 2013, The possible negative consequences of underground dam and reservoir construction and operation in coastal karst areas: An example of the Hydro-Electric Power Plant (HEPP) Ombla near Dubrovnik (Croatia): Natural Hazards and Earth System Sciences, Vol. 13, pp. 2041–2052. CRILLEY, D. M. AND TORAK, L. J., 2002, Physical and hydrochemical evidence of lake leakage and assessment of karst features in the vicinity of Lake Seminole, southwestern Georgia and northwestern Florida. In American Geophysical Union Fall Meeting Abstracts: American Geophysical Union, Washington, DC, abstract H72C-0864. DONNELLY, C. R.; HINCHBERGER, S.; AND MOHAMMADIAN, E., 2009, The design of foundation treatment measures for dams on karst foundations: Canadian Dam Association Bulletin, Vol. 20, No. 4, pp. 20–27. FALCON, N. L., 1961, Major earth-flexing in the Zagros Mountains of southwest Iran: Quaternary Journal of Geological Society of London, Vol. 117, pp. 367–376. FARS REGIONAL WATER AUTHORITY, 1997a, Water Supply Project of Boushehr, Geological Map of Kowsar Dam Site: Fars Regional Water Authority, Shiraz, Iran, scale: 1:5,000, 1 sheet. FARS REGIONAL WATER AUTHORITY, 1997b, Water Supply Project of Boushehr, Engineering Geology of Kowsar Dam Site: Fars Regional Water Authority, Shiraz, Iran [Report in Persian], 275 p. FARS REGIONAL WATER AUTHORITY, 2010, Kowsar Dam Site Condition Reports: Fars Regional Water Authority, Shiraz, Iran [Report in Persian], 98 p.

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 325–339

337

Mozafari and Raeisi FORD, D. C. AND WILLIAMS, P. W., 2007, Karst Geomorphology and Hydrology: 2nd ed.: John Wiley & Sons, Chichester, U.K., 562 p. GHOBADI, M. H.; KHANLARI, G. R.; AND DJALALY, H., 2005, Seepage problems in the right abutment of the Shahid Abbaspour Dam, southern Iran: Engineering Geology, Vol. 82, pp. 119–126. GIDROSPROEKT, 1996, Model Studies of Seepage and Cutoff Arrangement: Tange Duck Water Development Report, Iran, Kowsar Dam: Gidrosproekt, Tehran, Iran 1913-NSI-1-15702. GINTHER, C. H. AND CHARLTON, J. C., 2009, Comprehensive foundation rehabilitation at Bear Creek Dam. In USSD 29th Annual Conference: United States Society on Dams, Nashville, TN, 4/20/09. GU¨NAY, G.; EKMEKCI, G. M.; AND TEZCAN, L., 1995, Hydrogeochemical and isotopic evaluation of the hydrogeological system at Ceyhan-Berke Dam site: Application of tracers in arid zone hydrology. In Proceedings of the Vienna Symposium, August 1994: International Association of Hydrological Sciences Publication 232, pp. 395–408. GUTIERREZ, F.; PARISE, M.; DE WAELE, J.; AND JOURDE, H., 2014, A review on natural and human-induced geohazards and impacts in karst: Earth-Science Reviews, Vol. 138, pp. 61–88. HANSEN, R. L. AND TETER, G. A., 1970, Use of Radioisotopes in Tracing Reservoir Leakage at Anchor Dam: U.S. Department of the Interior, Bureau of Reclamation, Office of Chief Engineer, Washington, DC, 27 p. JARVIS, T., 2003, The money pit: Karst failure of Anchor Dam, Wyoming. In JOHNSON, K. S. and NEAL, J. T. (Editors), Evaporite Karst and Engineering/Environmental Problems in the United States: Oklahoma Geological Survey Circular 109, pp. 271–278. KARIMI, H., 2003, Hydrogeological Behavior of the Alvand Basin Karst Aquifers, Kermanshah: Ph.D. thesis, Shiraz, University of Shiraz, 281 p. KARIMI, H.; RAEISI, E.; AND ZARE, M., 2005, Physicochemical parameters time series of karst spring as a tool to differentiate the source of spring water: Carbonates and Evaporites, Vol. 20, No. 2, pp. 138–147. KARST RESEARCH CENTRE OF IRAN, 1993, Comprehensive Study and Research in Water Resource of the Maharlu Basin, Fars: Karst Research Centre of Iran Report, Volumes 1–4. LAKSIRI, L. B. K., 2007, Investigation of Water Leakage Mechanism in the Karstic Dam Site, Samanalawewa, Srilanka: Ph.D. thesis, Department of Engineering Systems and Technology, Graduate School of Science and Engineering, Saga University, Saga, Japan, 136 p. MARTIN, J. B. AND SCREATON, E. J., 2001, Exchange of Matrix and Conduit Water with Examples from the Floridan Aquifer: U.S. Geological Survey Karst Interest Group Proceedings, Water Resources Investigations Report 01–4011, p. 38–44. MILANOVIC´, P. T., 2004, Water Resources Engineering in Karst: CRC Press, Boca Raton, FL, p. 312. MILANOVIC, S.; STEVANOVIC, Z.; AND JEMCOV, I., 2010, Water losses risk assessment: An example from Carpathian karst: Environmental Earth Sciences, Vol. 60, pp. 817–827. MOHAMMADI, Z.; RAEISI, E.; AND BAKALOWICZ, M., 2007, Method of leakage study at the karst dam site. A case study: Khersan 3 Dam, Iran: Environmental Geology, Vol. 52, pp. 1053–1065. MONTOROI, J. P.; GRU¨NBERGER, O.; AND NASRI, S., 2001, Hydrochemical behaviour of waters in a Tunisian hill reservoir watershed and reservoir impact on alluvial aquifer. In: HYDROMED: se´minaire international les petits barrages dans le monde me´diterrane´en: recueil des re´sume´s. Paris (FRA); Tunis: IRD; INRGREF, 2001, 2 p. Multigr. Les.

338

Petits Barrages dans le Monde Me´diterrane´en: Se´minaire International, Tunis (TUN), 2001/05/28–31. MOZAFARI, M.; RAEISI, E.; AND ZARE, M., 2011, Water leakage paths in the Doosti Dam, Turkmenistan and Iran: Journal of Environmental Earth Sciences, Vol. 65, No. 1, pp. 103–117. NATIONAL IRANIAN OIL COMPANY EXPLORATION and PRODUCTION (NIOC), 1959, Geological Quadrangle Map of Iran No. E-10: Geological Map of Behbahan: (with explanatory text in Persian): NIOC, Tehran, Iran scale: 1:250,000, 1 sheet. OBERLANDER, T., 1965, The Zagros Streams: A New Interpretation of Transverse Drainage in an Orogenic Zone: Syracuse Geographical Series No. 1, Syracuse University Press, New York, 168 p. PALMER, A. N., 1988, Solutional enlargement of opening in the vicinity of hydraulic structures in karst regions. In Proceedings of the 2nd Conference on Environmental Problems in Karst Terranes and Their Solutions: Association of Groundwater Scientists and Engineers, Nashville, TN, pp. 3–15. PANTZARTZIS, P.; EMMANUELIDIS, G.; KRAPP, L.; AND MILANOVIC, P., 1993, Karst phenomena and dam construction in Greece. In Hydrogeological Processes in Karst Terranes, Proceedings of the Antalya Symposium and Field Seminar: International Association of Hydrological Sciences Publication 207, pp. 65–74. PARISE, M.; CLOSSON, D.; GUTIERREZ, F. AND STEVANOVIC Z., 2014, Facing engineering problems in the fragile karst environment. In LOLLINO, G.; MANCONI, A.; GUZZETTI, F.; CULSHAW, M.; BOBROWSKY, P.; and LUINO, F. (Editors), Engineering Geology for Society and Territory. Volume 5—Urban Geology, Sustainable Planning and Landscape Exploitation. Springer International Publishing, Switzerland, 2015, pp. 479–482. PATERSON, K., 1979, Limestone springs in the Oxfordshire Scarplands: The significance of spatial and temporal variations in their chemistry: Zeitschrift fu¨r Geomorphologie N.F., Vol. 32, pp. 46–66. PEZESHKPOUR, P., 1991, Hydrogeological and Hydrochemical Evaluation of Kuhe-Gar-Barm-Firooz Springs: M.Sc. thesis, University of Shiraz, Shiraz, Iran, 282 p. PITTY, A. F., 1968, Calcium carbonate content of water in relation to flow-through time: Nature, Vol. 217, pp. 939–940. QINGZHI, H.; KEFENG, Z.; XING, L.; JIANHONG, Y.; GUOQING, Z.; AND DINGGUO, S., 1998, Study on hydrochemical characteristics and seepage of regional karst water: Geological Science and Technology Information, Vol. 17, No. 2, pp. 9–14. QUINLAN, J. F., 1985, Application of dye-tracing to dam-site evaluation in a Kentucky karst area, U.S.A. In Proceedings of the International Symposium on Karst Water Resources: Ankara, Turkey, July 1985: International Association of Hydrological Sciences Publication 161, pp. 533–544. RAEISI, E., 2008, Groundwater storage calculation in karst aquifers with alluvium or no-flow boundaries: Journal of Cave and Karst Studies, Vol. 63, pp. 62–70. SAHUQUILLO, A., 1985, Spanish experience in karst water resources. In Proceedings of the International Symposium on Karst Water Resources, Ankara, Turkey, July 1985: International Association of Hydrological Sciences Publication 161, pp. 133–147. SCHAEFER, J. A., 2009, Risk evaluation of dams on karst foundations. In Proceedings of the 29th Annual USSD Conference, April 20–24, 2009: United States Society on Dams, Nashville, TN, pp. 541–579. SHARIFI, M., 2009, Evaluation of Groundwater Contamination in the Emamzadehjafar Aquifer, Gachsaran City: M.Sc. thesis, University of Shiraz, Shiraz, Iran, 298 p. STOCKLIN, J. AND SETUDEHNIA, A., 1977, Stratigraphic Lexicon of Iran: Geological Survey of Iran Report 18–1971, 370 p.

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 325–339

Understanding Karst Leakage, Kowsar Dam, Iran TORAK, L. J.; CRILLEY, D. M.; AND PAINTER, J. A., 2006, Physical and Hydrochemical Evidence of Lake Leakage near Jim Woodruff Lock and Dam and Ground-Water Inflow to Lake Seminole and an Assessment of Karst Features in and near the Lake, Southwestern Georgia and Northwestern Florida: U.S. Geological Survey Scientific Investigations Report 2005-5084, 89 p. TURKMEN, S., 2003, Treatment of the seepage problems at the Kalecik Dam (Turkey): Engineering Geology, Vol. 68, pp. 159–169. TURKMEN, S.; OZGULAR, E.; TAGA, H.; AND KARAOGULLARINDAN, T., 2002, Seepage problems in the karstic limestone foundation of the Kalecik Dam (south Turkey): Engineering Geology, Vol. 63, pp. 147–157. TUZHIKHIN, G. G. AND KOLICHKO, A. V., 2001, Cementation operations in the building of the Kousar Dam, Iran: Hydrotechnical Construction, Vol. 35, No. 8, pp. 402–405. UNAL, B.; EREN, M.; AND YALCIN, M. G., 2007, Investigation of leakage at Ataturk Dam and hydroelectric power plant by

means of hydrometric measurements: Engineering Geology, Vol. 93, pp. 45–63. WHITE, W. B., 1969, Conceptual models for limestone aquifer: Groundwater, Vol. 7, pp. 15–21. WHITE, W. B., 1977, Conceptual models for carbonate aquifers: Revisited. In DILAMARTER, R. R. and CSALLANY, S. C. (Editors), Hydrologic Problems in Karst Terrains: Western Kentucky University, Bowling Green, KY, pp. 176–187. WHITE, W. B., 1988, Geomorphology and Hydrology of Karst Terrains: Oxford University Press, New York, NY, 464 p. ZOGOVIC´, D., 1993, Storage water tightness problems and technical solutions in the highly developed karst area of the Niksˇic´ Polje. In Proceedings of the Hydrogeological Processes in Karst Terranes, Antalya Symposium and Field Seminar, October 1990: International Association of Hydrological Sciences Publication 207, pp. 85–93.

Environmental & Engineering Geoscience, Vol. XXI, No. 4, November 2015, pp. 325–339

339