Environmental & Engineering Geoscience

Environmental &Engineering GeoscienceNOVEMBER 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 pub-lished quarterly by the Association of Environmental and EngineeringGeologists (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 mail-ing offices.

EDITORIAL OFFICE: Environmental & EngineeringGeoscience journal, Department of Geology, Kent StateUniversity, Kent, OH 44242, U.S.A. phone: 330-672-2968, fax:330-672-7949, [email protected].

CLAIMS: Claims for damaged or not received issues will behonored for 6 months from date of publication. AEG membersshould contact AEG, 1100 Brandywine Blvd, Suite H, Zanesville, OH43701-7301. Phone: 844-331-7867. GSA members who are notmembers of AEG should contact the GSA Member Service cen-ter. All claims must be submitted in writing.

POSTMASTER: Send address changes to AEG, 1100Brandywine Blvd, Suite H, Zanesville, OH 43701-7301. Phone: 844-331-7867. Include both old and new addresses, with ZIP code.Canada agreement number PM40063731. Return undeliverableCanadian addresses to Station A P.O. Box 54, Windsor, ON N9A6J5 Email: [email protected].

DISCLAIMER NOTICE: Authors alone are responsible forviews expressed in articles. Advertisers and their agencies aresolely responsible for the content of all advertisements printed andalso assume responsibility for any claims arising therefromagainst the publisher. AEG and Environmental & EngineeringGeoscience reserve the right to reject any advertising copy.

SUBSCRIPTIONS:

Member subscriptions: AEG members automatically receive the journal as part of their AEG membership dues. Additionalsubscriptions may be ordered at $40 per year. GSA memberswho are not members of AEG may order for $40 per year ontheir annual GSA dues statement or by contacting GSA.

Nonmember subscriptions are $175 and may be ordered fromthe subscription department of either organization. A postagedifferential of $10 may apply to nonmember subscribers outsidethe United States, Canada, and Pan America. Contact AEG at303-757-2926; contact GSA Subscription Services, c/oAmerican Institute of Physics, 2 Huntington Quadrangle, Suite1N01, Melville, NY 11747-4502.

Single copies are $50.00 each. Requests for single copies shouldbe sent to AEG, 1100 Brandywine Blvd, Suite H, Zanesville, OH43701-7301. Phone: 844-331-7867.

© 2015 by the Association of Environmental and EngineeringGeologists

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, orby any information storage and retrieval system, without permission in writing from AEG.

THIS PUBLICATION IS PRINTED ON ACID-FREE PAPER

ABDUL SHAKOORDepartment of GeologyKent State University

Kent, OH 44242330-672-2968

[email protected]

BRIAN G. KATZFlorida Department of Environmental

Protection2600 Blair Stone Rd.Tallahassee, FL 32399

[email protected]

EDITORSCover photo

Air photo of the Lower San Diego River Valley showing the Qualcomm Stadiumarea north of downtown San Diego, California. The yellow outline is that of aplume of tertiary butyl alcohol (TBA), a biodegradation product of the gasolineadditive MTBE, which was unintentionally released in the late 1980s from thefuel terminal shown at the upper right corner of the photo (see article on pages000–000).

SUBMISSION OF MANUSCRIPTS

Environmental & Engineering Geoscience (E&EG), is a quar-terly journal devoted to the publication of original papers thatare of potential interest to hydrogeologists, environmental andengineering geologists, and geological engineers working in siteselection, feasibility studies, investigations, design or construc-tion of civil engineering projects or in waste management,groundwater, and related environmental fields. All papers arepeer reviewed.

The editors invite contributions concerning all aspects of envi-ronmental and engineering geology and related disciplines.Recent abstracts can be viewed under “Archive” at the website, “http://eeg.geoscienceworld.org”. Articles that report onresearch, case histories and new methods, and book reviewsare welcome. Discussion papers, which are critiques of print-ed articles and are technical in nature, may be published withreplies from the original author(s). Discussion papers andreplies should be concise.

To submit a manuscript go to http://eeg.allentrack.net. If youhave not used the system before, follow the link at the bottom ofthe page that says New users should register for an account.Choose your own login and password. Further instructions willbe available upon logging into the system. Please carefully readthe “Instructions for Authors”.

Authors do not pay any charge for color figures that are essen-tial to the manuscript. Manuscripts of fewer than 10 pages maybe published as Technical Notes.

For further information, you may contact Dr. Abdul Shakoor atthe editorial office.

JOHN W. BELL

Nevada Bureau of Mines andGeologyRICHARD E. JACKSON

(Book Reviews Editor)Geofirma Engineering, Ltd.JEFFREY R. KEATON

AMEC AmericasPAUL G. MARINOS

National Technical Universityof Athens, GreeceJUNE E. MIRECKI

U.S. Army Corps of EngineersPETER PEHME

Waterloo Geophysics, IncNICHOLAS PINTER

Southern Illinois University

PAUL M. SANTI

Colorado School of MinesROBERT L. SCHUSTER

U.S. Geological SurveyROY J. SHLEMON

R. J. Shlemon& Associates, Inc.GREG M. STOCK

National Park ServiceRESAT ULUSAY

Hacettepe University, TurkeyCHESTER F. “SKIP” WATTS

Radford UniversityTERRY R. WEST

Purdue University

EDITORIAL BOARD

ASSOCIATE EDITORS

JEROME V. DEGRAFF

USDA Forest ServiceTHOMAS J. BURBEY

Virginia Polytechnic InstituteSYED E. HASAN

University of Missouri, Kansas City

ROBERT H. SYDNOR

ConsulantCHESTER F. WATTS (SKIP)Radford University

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, Ground-water 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 high-

permeability 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 ,80percent of its water supply from distant sources thatcould be interrupted by seismic events, severing theaqueducts from the Colorado River (San DiegoProject) and northern California (State Water Pro-ject). In addition, surface-water sources such as theCalifornia State Water Project and the ColoradoRiver are threatened by drought and thus are nolonger as reliable as in the past.

Prior to World War II (WWII), the City of SanDiego utilized groundwater from a high-permeabilityalluvial aquifer in the lower San Diego River Valley(“the valley”). This aquifer, which is the subject ofthis article, was developed in 1914 but abandonedbefore WWII. It is the city’s goal to re-develop thisgroundwater resource once remediation of the Mis-sion Valley Terminal (MVT) fuel release has beencompleted and thus provide further diversification ofthe city’s water supply. It is the intent of thisremediation to restore background groundwaterquality conditions (San Diego RWQCB, 2005). Ourmotivation in preparing this article is to identify thebackground groundwater quality (GWQ) in the valleyto bring closure to the MVT remediation that beganin 1992; a discussion of background GWQ concludesthis article.

The study area (Figure 1) consists of the lower SanDiego River Valley, from its outlet near Mission Bayeastward and upstream along the San Diego River tothe vicinity of Qualcomm Stadium, a distance ofapproximately 8 km (5 mi). The main axis of thevalley contains the San Diego River and its relatedfloodplain. Murphy Canyon Creek is now confined toa concrete channel that directs flow into the river andthence to the Pacific Ocean. The lower San Diego

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

River Valley is frequently identified as “MissionValley”; however, we use this term only to refer tothe high-permeability alluvial aquifer, i.e., the Mis-sion Valley Aquifer (MVA), which was pumped as thepre-WWII groundwater supply.

Our primary objective is to present a hypothesisdescribing the Quaternary history of the valley basedupon similar histories elsewhere along the southernCalifornian coastline (e.g., Edwards et al., 2009) andevidence from borehole logs and cores collected in thevalley. We then use that hypothesized history toexplain the occurrence of brackish GWQ in theQuaternary-age alluvial deposits and the deeper,Eocene-age sedimentary rocks beneath the Valley.

GEOLOGIC HISTORY OF THE LOWER SANDIEGO RIVER VALLEY

Sources of Information

The valley contains Cretaceous through Holocenerocks and sediments deposited over the past 145 m.y.under a multitude of paleoenvironmental conditions.Understanding the geologic history and stratigraphicrelationships of the rock layers is important tocharacterize the aquifer now intended for sustainabledevelopment. Accordingly, we summarize the stratig-raphy and geologic history of the valley based mainly

on the investigations of Geofirma and INTERA from2010 through 2014, on reconstructions of the broadgeologic history of the San Diego area (Abbott, 1999;Kennedy and Peterson, 2001), and on interpretation ofnew, site-specific subsurface lithologic and GWQ data.

Most of our detailed stratigraphic interpretationsof the Quaternary alluvial deposits are based ongeologic logs of groundwater monitoring wells pre-pared for Kinder Morgan Energy Partners (KMEP).KMEP is the owner and operator of the MissionValley Terminal (MVT) bulk fuel storage facilitylocated at the mouth of Murphy Canyon. In addition,the City of San Diego has installed since 2011 severalmonitoring wells in the buried channel aquifer at theintersection of Interstates 8 and 805, as well as onenorth of the MVT within Murphy Canyon and twomore on the northern edge of the Qualcomm Stadiumparking lot (see Figure 1). These new wells nowprovide site-specific information about the alluviumand Friars Formation bedrock through lithologic andgeophysical logs, petrographic analysis, and labora-tory grain-size analysis. In 2004, the U.S. GeologicalSurvey (USGS, 2014) installed a monitoring wellcluster known as the Aquaculture Well (SDAQ,San Diego), a multi-depth monitoring well consistingof five nested piezometers, which also providesgeologic context. This well is located on the southside of the river across from Qualcomm Stadium (see

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

Sengebush, Heagle, and Jackson

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

Figure 1); it provides a vertical stratigraphic columnto a depth of approximately 940 ft (286 m) belowground surface (bgs). The lithologic and geophysicallogs of this well and periodic water sampling andanalysis provide a deep vertical profile of Quaternarysediments, Eocene bedrock, and GWQ within thevalley.

Stratigraphic History

Figure 2 is a stratigraphic column illustrating thegeologic 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

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

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


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 non-

conformable 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 sedimentarysequences. Alluvial fans were built seaward, pushingthe shoreline to the west (Abbott, 1999). Sedimentsdeposited during this time consisted of both the LaJolla Group, west of the study area, and the PowayGroup, which constitutes the rocks in the San DiegoRiver Valley in the vicinity of Qualcomm Stadium.The La Jolla Group is only represented by the FriarsFormation in the study area. The Poway Groupincludes the Stadium, Mission, and Pomerado For-mations and consists of sediments laid down by theEocene Ballena River, an ancestral west-flowingchannel originating east of San Diego but nowdisplaced by lateral and vertical movements of theSan Andreas and related fault systems (Abbott andSmith, 1989).

The late middle Eocene–age Friars Formation (Tf)of the La Jolla Group overlies the Santiago PeakVolcanics throughout the San Diego River Valleyarea. The Friars Formation crops out extensively onthe walls of Murphy Canyon and San Diego Rivercanyon and is named for exposures along the northside of the valley near Friars Road on the La Jollaand La Mesa geologic quadrangles (Kennedy andPeterson, 2001). The Friars Formation consists ofsand- and clay-stone and contains both non-marineand lagoonal facies, up to 492 ft (150 m) thick in thestudy area. Table 1 presents the mineralogical com-position of the Friars Formation.

The Stadium Formation (Stadium conglomerate)lies directly over the Friars Formation in the studyarea and contains clasts rounded from fluvial trans-port 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 Canyonand the sides of Serra Mesa, bordering QualcommStadium. Abbott and Smith (1989) postulated thatthe Stadium formation originated from volcanicsources near Sonora, Mexico, an area that has sincebeen separated from San Diego by lateral slip along

the intervening San Andreas and related plate-boundary faults.

The Stadium conglomerate is overlain by theMission Valley Formation (Tmv) and the PomeradoConglomerate (Tp), completing the Poway Group ofEocene sand and gravel deposits. The Mission ValleyFormation is about 200 ft (60 m) thick, while thePomerado is about 180 ft (55 m) thick. These unitsinter-tongue from east to west.

The Pliocene units directly overlie the Eocenestrata. The Pliocene rocks consist of the San DiegoFormation (Tsd) marine sandstone. During Pliocenetime (5.3–1.8 m.y. ago), continental glaciers grew inthe Northern Hemisphere, resulting in a sea-level fallof about 625 ft (190 m), according to Abbott (1999).Invertebrate, marine mammal, fish, and bird fossilsare abundant in the 100-ft-thick (30-m-thick) SanDiego Formation, readily seen in road cuts on thesouth side of the valley near the mesa top.

Quaternary Geology and Hydrogeology

The Pleistocene rocks and sediments in the studyarea consist of the Linda Vista Formation (Ql) andterrace alluvium (Qt). The Linda Vista Formation,which consists of redeposited sand and conglomeratederived from nearby older sediments, is mapped onthe top of the mesa bordering the valley. ThePleistocene stream-terrace deposits within the valley,as mapped by Kennedy and Peterson (2001), arelocated at the foot of the mesas on the north side ofthe valley and extend from the mesa slopes to thebanks of the river, a distance of up to ,3000 ft(1 km). Holocene deposits are San Diego Riveralluvium (Qal) and slopewash (Qsw) (Kennedy andPeterson, 2001). These sediments have been largelycovered by urban development, including the pavedparking lot of Qualcomm Stadium, and are describedas “poorly consolidated, conglomeratic sand depos-its” (Kennedy and Peterson, 2001, p. 50).

Petrographic analysis of the terrace alluvium inMW-03 at 59 ft (18 m) bgs reveals these sands arecomposed of almost entirely polycrystalline porphy-ritic dacite volcanic and pyroclastic lithic grains,equigranular granite to granodiorite intrusives, andsubordinate metamorphic lithic grains and mono-crystalline quartz. Table 1 presents the mineralogicalcomposition by DePangher (2014); locations of wellsare shown in Figure 1.

The geomorphic feature extending from the mouthof Murphy Canyon to the modern San Diego Riverhas been interpreted as the Murphy Canyon alluvialfan (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



Table 2. These dimensions suggest that the MurphyCanyon alluvial fan fits the type II fans identified byBlair and McPherson (1994) with sediment transportdominated by sheetflow.

We believe the Murphy Canyon fan was derivedprimarily from the Poway Group and the Friarssandstone. The evidence for this interpretation stemsmainly from lithofacies in the boring logs ofmonitoring wells throughout the area. In particular,the rounded volcanic cobbles described in the KMEPboring logs and observed in core from the city’s recent(INTERA, 2014) monitoring well borings—oftendescribed as up to 4 in. (100 mm) in diameter—areevidence of the correlation between the cobbles in the

Stadium and Pomerado conglomerates and thecobbles in the basal and surficial gravel deposits inthe Murphy Canyon alluvial fan.

Textural analysis of the basal gravel within thealluvium (MW-3, 59 ft or 18 m bgs) describes thematerial as poorly graded gravel with sand: mediangrain size d50 5 12 mm and uniformity coefficientCu 5 d60/d10 5 69. By comparison, the FriarsFormation (MW-3, 67 ft or 20 m bgs) is finergrained: d50 5 0.17 mm and Cu 5 171 (Daniel B.Stephens and Associates, Inc., 2014). Figure 3 pre-sents grain-size distribution curves for these samples.

Lithologic logs from monitoring wells installed inthese sediments for the MVT remediation projecthave been used to map the subsurface lithofacies ofthe Qt stream-terrace deposits (Geofirma EngineeringLtd. and INTERA, 2011). Schematic geologic crosssection A-A9 (Figure 4) illustrates the vertical andlateral distribution of various sediment types withinthe Qt deposits. The cross section identifies threeprincipal layers within the Qt beneath the MVT andoff-terminal remediation area: the basal gravel

Table 2. Dimensions of the Murphy Canyon alluvial fan.

Parameter Value

Width of toe of fan, m (mi) 1,584 (1)Length from mouth to toe, m (mi) 800 (0.5)Slope from mouth to toe, m (ft) 6 (20)Angle of slope, deg 0.43

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 26 Friars sandstonePlagioclase 20 Scale: ,27 mm acrossK-feldsparClayOrganic matterCarbonateBiotiteIlliteChlorite

202053221

The sample is an unconsolidated sandderived almost entirely froma carbonaceous clayey arkosesandstone protolith

From borehole at MW-3, sampled at67 ft (20 m) bgs (DePangher, 2014).

Plagioclase 74 Basal gravel unit of the MVAQuartz 10 Scale: ,27 mm acrossK-feldsparHornblendeActinoliteFerric oxideSericiteBiotite and chlorite

82221

,1

The sample is an unconsolidated lithicsand composed almost of entirelypolycrystalline lithic grains:polycrystalline porphyritic dacitevolcanic and pyroclastic lithic grains,equigranular granite to granodioriteintrusives, and subordinatemetamorphic lithic grains andmonocrystalline quartz

From borehole at MW-3, sampled at59 ft (18 m) bgs (DePangher, 2014).



(eroded into the Friars Formation); the middle sandlayer (including zones of silt and clay within thesand); and the upper gravel layer.

These three units are generally identifiable in thewells across the MVT site (although the upper gravel isabsent in some of the lithologic logs). The basal gravelcontains a channel structure (“paleochannel”) thattrends northeast to southwest beneath the QualcommStadium parking lot. This basal gravel deposit, in-cluding the paleochannel, together with the overlyingmiddle sand unit comprise the MVA in the context ofboth the MVT remediation program and its future useas a city water-supply aquifer (Geofirma EngineeringLtd. and INTERA, 2011, 2013). This paleochannel,now obscured beneath the stadium parking lot, was thelocation of several of the City of San Diego water-supply wells completed and operated prior to WWII(Fay, 1914; Ellis and Lee, 1919). Additional city wellswere located along the axis of the river downstreamfrom the alluvial fan. This combination of unconsol-idated gravel, sand, silt, and clay beneath QualcommStadium represents a Pleistocene fluvial depositionalenvironment with its major source in Murphy Canyon.A schematic block diagram (Figure 5) illustrates thevarious layers and their spatial relationships.

We describe three Qt lithofacies—the basal gravel,middle sand, and upper gravel—using Miall’s (1985)

classification. We interpret the basal gravel deposit,with its principal lithofacies of massive, matrix-supported gravel (Gms) with massive or crudelybedded gravel (Gm) and coarse to very coarse sand(St), to be a channel (CH) element. The coarse sandmatrix would result in high hydraulic conductivityfor this basal gravel layer, which is consistent withthe grain-size analysis and hydraulic conductivityestimates from sediment samples of the basal gravel inthe Murphy Canyon and stadium parking lot wellsinstalled by the City of San Diego (INTERA, 2014).In addition, the concave-up erosional base (as definedby the elongate and trough-like contact between thebasal gravel and the Friars Formation) fits with thethird-order contact within the hierarchy of beddingcontacts as described by Allen (1983). Our classifica-tion is consistent with the CH architectural elementinterpretation of Miall (1985, 1992).

Because Quaternary terrace deposits do not crop outin the immediate study area, we illustrate the de-positional model from well-known examples nearSocorro, New Mexico (Figure 6). The photographshows a typical matrix-supported gravel overlyinga sand unit similar to the Qt channel contact with theunderlying Friars sandstone. Our lithofacies classifica-tion is provided in Table 3, and the stratigraphicarchitectural elements are given in Table 4 (Miall, 1985).

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-3borehole, front entrance to the Qualcomm Stadium, Friars Road, San Diego.



Pleistocene Paleochannels

The Murphy Canyon alluvial fan contains a Pleisto-cene gravel channel beneath the Qualcomm Stadiumparking lot (the “paleochannel”), which extends fromthe mouth of Murphy Canyon southwest beneath theQualcomm Stadium parking lot to the river. Thispaleochannel (within the area mapped as Quaternaryterrace deposits, Qt; Kennedy and Peterson, 2001)incises the Friars Formation and is mapped by lines ofequal gravel thickness (Figure 7) based on the gravellithofacies thickness in monitoring well logs (GeofirmaEngineering Ltd. and INTERA, 2011). Additionally,the elevation of the contact surface at the base of thegravel and the top of the Friars Formation was mapped(Figure 8). This shows a thalweg that spatially coincideswith the axis of the gravel thickness isopach map. Thissubsurface structure ranges from about 300 ft (91 m) to600 ft (183 m) wide and is 3,600 ft (1,097 m) long from

the mouth of Murphy Canyon to the toe of the alluvialfan. The slope of the channel parallel to the channel axisis 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 southwestcorner of the stadium parking lot is 3.8 degrees.

Bull (1991, p. 172) observed that “unusually warmsea-surface temperatures at about 125 ka should havefavored stronger and more frequent tropical storms inthe San Gabriel Mountains” in Los Angeles County;this is no doubt true for San Diego County at the sametime. Such storms would have accelerated the erosionof Cenozoic sediments in the coastal areas of southernCalifornia, potentially leading to the erosion of the SanDiego River Valley and its tributaries. The paleochan-nel down cut through existing Eocene sediments in thevalley, in particular, the Friars Formation, and wasfilled with an upward-fining sequence of gravels, sands,

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.



and silts similar to modern gravel channels foundthroughout the Basin and Range Province. Thisburied-channel aquifer beneath the Qualcomm Stadi-um parking lot (within the Murphy Canyon alluvial

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

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

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



gallons per day of high-quality groundwater (,400 mgtotal dissolved solids [TDS]/L; Ellis and Lee, 1919).

The buried paleochannel within the Murphy Can-yon alluvial fan is interpreted as analogous to the latePleistocene buried channel of the lower Santa Marga-rita River, as described by Shlemon (1979), and toother buried Pleistocene channels along the southernCalifornia coast that grade to marine isotope stage 2,indicating the eustatically lowered sea levels of the LastGlacial Maximum (LGM) (Shlemon, 1979; Edwards etal., 2009; and Lee and Normark, 2009). The SantaMargarita River and its estuary are located in northernSan Diego County, 40 mi (64 km) north of San DiegoRiver Valley. The buried channel lies beneath themodern Santa Margarita River and is identified byburied gravels 75 ft (23 m) thick, extending to a depthof 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 thechannel was steep compared to that of the modernSanta Margarita River. Subsequent sea-level risecovered the channel with finer-grained sediments,resulting in a fining-upward sedimentary sequence.

We believe that the paleochannel mapped beneaththe Qualcomm Stadium parking lot intersects and istributary to a larger and deeper paleochannel beneaththe San Diego River (Figure 9), similar to the SantaMargarita River buried paleochannel. This paleochan-nel beneath the river was the location of several of thecity’s pre-WWII MVA water-supply wells (Figure 10).

Last Interglacial Sea-Level Highstand

In the San Diego area, and in the lower San DiegoRiver Valley in particular, the Pleistocene was a timeof repeated sea-level rise and fall as worldwideglaciers advanced and retreated (Abbott, 1999).Erosional features related to these sea-level changesmay be seen today as terraces in and around La Jolla,San Diego, and on the mesas above San Diego RiverValley. The number and spacing of terraces weredetermined by the rate of tectonic uplift and nature ofcoastal processes. The oldest terrace is generallycorrelated with the early Pleistocene (1.18 Ma to120 ka) according to Muhs et al. (2002).

Kern and Rockwell (1992) documented 16 separatemarine terraces, ranging in age from 1.29 Ma to 80 ka.These erosional wave-cut platforms mark the highestsea-level elevations maintained during glacial/inter-glacial time. Muhs et al. (2002) also described marineterraces with a focus on those near Point Loma, thelower Bird Rock terrace at about 26 ft (8–9 m) abovepresent sea level, and the higher, Nestor terrace, about75 ft (23–24 m) above present sea level. The Nestorterrace dates to 120 ka, while the Bird Rock terrace ismore recent, dated at 80 ka. The Bird Rock terraceformed at about 26 ft (22 m) relative to present sealevel, while the higher Nestor terrace formed about 20ft (6 m) above present sea level (Table 3; Kern andRockwell, 1992). Based on the estimated tectonicland-surface uplift that has taken place over the past

Table 3. Lithofacies classification for Qt deposits (from Miall, 1985).

Facies Code Lithofacies Sedimentary Structures Interpretation

Gms Massive, matrix-supported gravel Grading Debris-flow depositsGm Massive or crudely bedded gravel Horizontal bedding,

imbricationLongitudinal bars, lag deposits, sieve

depositsGt Gravel, stratified Trough cross-beds Minor channel fillsGp Gravel, stratified Planar cross-beds Linguoid bars or deltaic growths from

older bar remnantsSt Sand, medium to v. coarse, may

be pebblySolitary (theta) or grouped

(pi) trough cross-bedsDunes (lower flow regime)

Sp Sand, medium to v. coarse, maybe pebbly

Solitary (alpha) or grouped(omikron) planar cross-beds

Linguoid, transverse bars, sand waves(lower flow regime)

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

Element Symbol Principal Lithofacies Assemblage Geometry and Relationships

Channels CH Any combination Finger, lens or sheet; concave-up erosional base;scale and shape highly variable; internal concave-upsecondary erosion surfaces common

Gravel bars andbed forms

GB Gm, Gp, Gt Lens, blanket; usually tabular bodies; commonlyinterbedded with SB

Sandy bed forms SB St, Sp, Sh, Sl, Sr, Se, Ss Lens, sheet, blanket, wedge; occurs as channel fills,crevasse splays, minor bars



120,000 years, the inland extent of the 120 ka (Nestor)sea-level invasion may be estimated.

Uplift resulting from offset on the Rose Canyonfault dominates the tectonic history of the San Diegocoast. Uplift has occurred at a rate of 0.13–0.14 m/k.y., with both higher and lower rates near the RoseCanyon 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 Nestorterrace was 19 ft (6 m) above the present sea levelprior to uplift. Therefore, the elevation of the marineinundation of the valley would have been 74 ft (23 m)above the present sea level, which is shown inFigure 11. This illustration shows the marine in-undation of the valley to a maximum position in thevicinity of the San Diego Mission, with someinundation of Murphy Canyon. Abbott (1999, pages202–203) has calculated and shown a similar marinehighstand during this interglacial.

GROUNDWATER QUALITY IN THE LOWERSAN DIEGO RIVER VALLEY

The late-middle Eocene–age sediments, such as theFriars Formation, were deposited ca. ,50 to ca. 34Ma and were inundated repeatedly by rising sea levelsduring the Pleistocene. These Eocene sedimentaryrocks contain brackish groundwater with TDS,2,000 mg/L, which is either (1) connate watertrapped during sedimentary deposition or (2) seawa-ter that inundated the valley during Pleistocene time.In both cases, the residual salinity would have becomediluted by freshwater recharge flowing throughthe valley flow system. In this section, we evaluatethe role of two Pleistocene events: (1) the marine in-undations during the last interglacial (marine oxygenisotope substage 5e; Shackleton, 1969) and earlierPleistocene interglacials and (2) the subsequent MVAdeposition during the LGM. Then, we attempt todetermine how they might have influenced the GWQin the valley since 1915, when measurements began.

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



We initially present USGS data for the San DiegoAquaculture monitoring-well cluster that provide a ref-erence set of GWQ data for the sedimentary sequencebeneath the valley. Other USGS, DWR, and city GWQdata are then presented, followed by stable isotopevalues of oxygen and deuterium to show the differencesevident in groundwaters from monitoring wells through-out the valley. The GWQ data are subsequentlyinterpreted to demonstrate “freshening” of the Eocenegroundwaters reported in the 1965 California Depart-ment of Water Resources (DWR) GWQ data; thisfreshening is accompanied by increased salinity in theMVA itself. We then identify background GWQconditions prior to the urban development of the valleyduring the 1960s based on the evidence of geologicalhistory and hydrogeochemical analysis.

Sources of Information

The GWQ in the valley prior to the development ofboth 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 samplecollected was a composite sample from the 13 drilledwells of the city’s Mission Valley well field, which haddepths ranging from 15 to 30 m (45 to 90 ft) bgs.Figure 10 shows the approximate position of the first12 of the 13 wells based on records of the San DiegoWater Department. Presumably, because just onesample from June 1915 was analyzed (see Table 5),the sample was collected from a manifold at thepumping station that mixed the groundwater from the13 wells drilled the year before by the city (Fay, 1914).

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



Ellis and Lee’s table 46 referred to this sample as K 47and identified its origin as 13 drilled wells belongingto the City of San Diego located in the “Pueblo landsand Ex Mission of San Diego.” These data arereproduced as Table 5.

DWR evaluated the GWQ in the valley prior to theconstruction of the MVT tank farm and the QualcommStadium in the 1960s but following the abandonment ofthe 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 thevalley shown in Figure 1, including data (see Table 6)from a number of residential and farm wells that arelocated on Figure 10. Well construction and screendepths for these wells were not reported. A final report(DWR, 1967) identified the variability in GWQ in bothinland and coastal regions. These DWR studies wereconducted over the same area “in support of theactivities of the San Diego Regional Water QualityControl Board” (DWR, 1967, p. xiii), which exists tothis day as the regional regulator.

The seven wells for which the data are presented inTable 6 (DWR, 1965) are those along or adjacent tothe axis of the Murphy Canyon paleochannel (i.e., itsthalweg) and thus constitute a set of data that isuseful in reconstructing the ambient or baseline GWQof the MVA prior to the establishment of the MVT in1963. Three of the wells, 5D1, 5M1, and 5N1, are inMurphy Canyon itself (see Figure 1), while twoothers are at the mouth of Murphy Canyon (17D1and 17D2; see Figure 10). The sixth well, 18Q3, waslikely the city’s former well no. 6 in the pre-WWIIwell field (see Figure 10). A seventh well, 18N1, wasoutside the MVA and located in an area now builtover near Friars Road and is shown in Figure 10.

The USGS San Diego Hydrogeology Project beganan extensive study of the San Diego region in 2001(http://ca.water.usgs.gov/sandiego/) that has yieldedmuch useful information on background GWQ in theregion (Wright et al., 2005; Wright and Belitz, 2011;and Anders et al., in review). In particular, Anderset al. (2014) has studied GWQ in the Pliocene-age SanDiego Formation. The data presented in Table 7 are

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

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

SiO2 Fe Ca Mg Na + K CO3 HCO3 SO4 Cl NO3 TDS

24 Trace 57 17 54 0.0 151 81 85 1.0 394



from the data compilation of Anders et al. (in review).These analyses are considered “complete” in thehydrogeochemical sense in that a full suite of majorinorganic ions and some stable and radiogenicisotopes were analyzed. Table 7 presents the resultsof three samples from the valley.

The USGS Aquaculture (SDAQ) well cluster wasinstalled in 2004 on the south side of the river oppositethe Qualcomm Stadium (see Figure 1). This monitoringwell cluster contains five 2 in. (5 cm) nested piezometersinstalled within a 17.5 in. (44 cm) borehole. These wellsare 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 baseof the MVA at 20 ft (6 m) above mean sea level (amsl)and penetrating the Friars Formation. Thus, SDAQ-J7provides a mixed sample of MVA and Friars FormationGroundwater, including tritium from the Quaternarysediments and high TDS (1,900 mg/L) from the FriarsFormation, as shown in Figure 12.

In addition to sampling the SDAQ multi-level well,the USGS analyzed samples from several other wells inMission Valley, including that of the River Walk GolfCourse number 2 well (“RWGC2”), which is furtherdown Mission Valley (see Figure 1). Both RWGC2 andSDAQ J7 have relatively high salinity, with RWGC2(TDS ,3,579 mg/L) being higher than J7 (TDS ,1,840mg/L), compared with those shown in Table 8. Theproximity of the RWGC2 well to the San Diego RiverFloodway (0.5 mile, 800 m) and the typical extractionrates of irrigation wells suggest that the high salinity inthis well is due to modern seawater intrusion.

The Groundwater Quality Monitoring Act(California, 2001) initiated the Groundwater Ambient

Monitoring and Assessment (GAMA) Priority BasinProject “to assess and monitor the quality of ground-water in California. (Wright and Belitz, 2011, p. 2)”Wright and Belitz (2011, p. 1) indicated that the“GAMA San Diego study was designed to providea statistically robust assessment of untreated-ground-water quality within the primary aquifer systems.” Oneof the four primary aquifer systems tested is identifiedas “Alluvial Basins,” which would include aquifersystems such as the San Diego River alluvial basin thatcontains the MVA. The USGS sampled a total of 17alluvial basin wells in 2004, including two public water-supply wells in the San Diego River Valley fartherupstream from Qualcomm Stadium. The range ofmeasured GWQ parameters in these 17 wells ispresented in Table 8. The rationale for including inthis article groundwater samples from alluvial wellscollected by the USGS outside the valley, i.e.,Table 8, but within the San Diego DrainagesHydrogeological Province (Wright and Belitz,2011), is that they are derived from sediments ofsimilar geochemical nature to the alluvial sedimentswithin the valley and thus are representative ofGWQ within the valley.

Table 9 presents data from two City of San Diegomonitoring wells that were installed and sampled in2011. These monitoring wells are situated downgradient of the SDAQ multi-level well but in theMVA and were, at the time of sampling, at the leadingedge of a plume of contaminated groundwater contain-ing the gasoline additive methyl tertiary-butyl ether(MTBE) and its biodegradation product tertiary-butyl alcohol (TBA). Since these samples werecollected, the TBA and TDS concentrations have

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) 5D1 5M1 5N1 17D1 17D2 18N1 18Q3

Sample date Feb 1959 May 1960 April 1959 May 1960 Feb 1963 April 1959 April 1955Temperature (uC) n.m. 28 n.m. 26 n.m. n.m. 22pH (pH units) 7.3 7.5 7.6 7.2 7.1 7.8 7.2SEC (mS/cm) 1,400 1,406 1,432 3,160 n.m. 2,931 1,786TDS{ (mg/L) 1,039 975 994 2,155 1,776 1,944 1,105Sodium (mg/L) 132 214 230 363 n.m. 340 215Potassium (mg/L) 2 8 4 5 n.m. 5 4Calcium (mg/L) 149 54 51 226 189 122 103Magnesium (mg/L) 35 28 20 68 93 115 46Chloride (mg/L) 216 206 241 615 593 703 368Sulfate (mg/L) 288 111 118 383 360 154 146Bicarbonate (mg/L) 250 346 315 432 n.m. 416 303Nitrate-NO3 (mg/L) 7.6 0 0 9 2.0 4 2.5Boron (mg/L) 0.13 0.4 0.08 0.6 n.m. 0.44 0.14

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



risen in this part of the MVA; thus, no more recentdata from these wells are included in this assessment.

These five data sets (Tables 5 to 9) containincreasingly large analyte lists from a limited numberof water-supply and monitoring wells in the LowerSan Diego River Valley. We now apply severalmethods of hydrogeochemical analysis in order toidentify the origin and evolution of the groundwaterin the valley. However, no single well has beencontinually sampled in the 100 year period since the1915 USGS analysis shown in Table 5. Samples fromthe MVA prior to its contamination by the MVTgasoline releases of 1987–1991 are limited to just the1919 USGS (Table 5) and 1965 DWR (Table 6) datasets. Thus, we will compare data from this valleywith that from elsewhere in the San Diego hydro-geologic region as collected, analyzed, and compiledby the USGS (Wright et al., 2005; Wright and

Belitz, 2011; Anders et al., 2014; and Anders et al.,in review).

Stable Water Isotopes

Measurements of the stable water isotopes 18O anddeuterium (2H) provide information on the origin thegroundwaters in the valley. Stable water isotope datain this paper are reported in per mil (%) compared toVienna Standard Mean Ocean Water (VSMOW), i.e.,d18O and d2H, and shown in Figure 13 for July andOctober 2014. USGS San Diego River samplinglocations in the lower valley are shown on Figure 1,while groundwater sampling was conducted with thecompletion of the city’s monitoring well networkshown in Figure 14.

The San Diego River water samples (open trian-gles) represent storm runoff during five events in

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



2004–2010, which was obtained from the USGSNational Water Information Database for the threelocations shown in Figure 1. These samples fall on orabout the global meteoric water line (GMWL),although two plot to the right of the GMWL. Sucha displacement is often considered evidence ofevaporation prior to sampling (Clark and Fritz,1997); however, it appears that here it reflects theannual or seasonal variability in the isotopic charac-ter of the winter rains (Williams and Rodoni, 1997).

The groundwater samples are mainly from citymonitoring 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 “MVAMW-2 MW-3” or “MVA MW-1” for that well inMurphy Canyon. These are within the MVA paleo-channel, except for the deepest sampling port in eachcase, which was installed in the Friars Formation justbeneath the paleochannel gravels; these are identifiedas “Friars MW-1-2-3.” A few samples are includedfrom 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 (dia-mond-shaped data) fall on or around a “GWcorrelation” line, in which the most depleted samples(i.e., most negative) are from furthest up MurphyCanyon at multi-level well MW-1 (see Figure 14). Theshallow SDAQ J7 sample also plots on this GWcorrelation line, as do MW-2 and MW-3 samples fromthe MVA. These samples appear to indicate that theMVA, i.e., both the paleochannel beneath QualcommStadium and that beneath the main channel of theriver (DB data), is recharged by discrete storm runoffevents producing the unique spatial pattern in theMVA shown in Figure 13. It is also possible that thispattern in some way reflects the infiltration ofirrigation at various golf courses above the stadiumand the DB site and perhaps infrastructure leaks.

More data are required to resolve this uncertainty,because the runoff data and the groundwater data arefrom different times; consequently, it is difficult to

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) SDAQ-J7 SDAQ-J7 RWGC-2

Screen elevation (ft a.m.s.l.) 20 20 250 to 280Sample date Aug 2010 May 2005 Jan 2004Temperature (uC) 24.5 22 21pH (pH units) 7.0 (field) 7.1 7.1 (field)Dissolved oxygen (mg/L) 2.1 0.5 0.7Specific electrical conductance (mS/cm) 2,950 (field) 3,000 4,600 (field)Total dissolved solids, residue (mg/L) 1,930 1,840 2,813Alkalinity (mg/L CaCO3) 294 340 755Sodium (mg/L) 300 301 644Potassium (mg/L) 2.6 3.9 11.2Calcium (mg/L) 221 219 303Magnesium (mg/L) 71.1 75 137Iron (mg/L) 0.524 0.92 2.43Manganese (mg/L) 2.79 3.05 3.25Chloride (mg/L) 742 631 1061Sulfate (mg/L) 224 237 477Bicarbonate (mg/L) 338 414.4 852Nitrate-N (mg/L) ,0.04 0.022 ,0.06Boron (mg/L) 0.22 0.23 0.359Arsenic (mg/L) 0.0026 0.00713 0.0118Dissolved organic carbon (mg/L) 4.0 No sample No sampleOxygen-18, d18O (%) 25.2 25.58 25.5Deuterium, d2H (%) 240 241.1 237Tritium, 3H (tritium units) 5.8 5.6 2.9SI CaCO3, calcite 0.2 0.33 0.7SI Fe(OH)3, iron hydroxide ,0.0 2.0 ,0.0SI MnO2, pyrolusite ,0.0 210.19 210.47Na/Cl ratio (mmol/L) 0.62 0.74 0.94Cl/Br ratio (mmol/L) 814 661 644

Note: SI indicates the saturation index of the sample, where 0.0 indicates equilibrium with the mineral, negative values indicate mineraldissolution, 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.



define reliable end points for mixing calculations.Nevertheless, a comparison of hydraulic heads inSDAQ J7 with the stage height of the river showsevidence 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 discretetimes and localities that cannot be identified with ourpresent data and monitoring well locations.

Figure 11. The 74 ft (23 m) a.m.s.l. topographic contour showing the marine inundation of the San Diego River Valley, approximately120,000 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.

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



Those groundwater samples from sampling portslocated in the Friars Formation (yellow circles) appearto be a mixture of MVA and deep bedrock ground-waters but have low TDS values (,1,200 mg/L), ratherthan Friars Formation groundwaters, which exhibitedhigher salinity from the inundation of the valley. Itappears that low TDS groundwater present beneaththe Quaternary/Friars contact may discharge upwardsinto the MVA from the deeper Eocene sedimentarybedrock (see TDS data in Figure 12).

Thus, the paleochannels appear to be acting asfocused linear discharge areas through which theregional groundwater flow system discharges into theMVA under the artesian conditions noted above. Thishas resulted in the deepest sampling ports from MW-1,MW-2, and MW-3 plotting midway between the deepSDAQ samples and the DB samples on Figure 13.

Impacts of the MVT Gasoline Release onGroundwater Quality

An additional complication has been posed by thepresence of high TDS concentrations induced by thebiodegradation of the very large gasoline leak(,200,000 gallons or ,800 m3) that occurred in1987–1991 at the MVT. The MVT is situated (seeFigure 1) at the neck of Murphy Canyon, whichallowed the gasoline to directly penetrate the MVAgravels and for the dissolved phase contamination—principally MTBE, which biodegraded to TBA—to betransported throughout the MVA to the DB monitor-ing wells. The TBA plume shown in Figure 10 exactlytraces the MVA paleochannel due to its very highpermeability relative to the Friars Formation.

Biodegradation of the gasoline within this plumeresulted in an increase in TDS due to the production ofprotons caused by the hydrolysis of the dissolvedcarbon dioxide, i.e., CO2 + H2O 5 H2CO3 5 H+ +HCO3

2. The acid produced attacks mineral surfaces(see Bennett et al., 1993; Borden et al., 1995; andMcMahon et al., 1995), thus causing their dissolutionand an increase in TDS. Such biodegradation-induced TDS may be identified by the simultaneousoccurrence of fuel hydrocarbons in the groundwatersample, e.g., MTBE and TBA. For this reason, wehave restricted our analysis of groundwater fresheningto samples collected by DWR prior to the constructionof the MVT in 1963 and gasoline releases.

Freshening Process and Its Effect on the MissionValley Aquifer

We have defined the MVA as the Pleistocene(LGM) paleochannel deposit underlying both theQualcomm Stadium and the Lower San Diego River

Valley (see section on “Pleistocene Paleochannels”).Because of the cutting and filling of the paleochannelduring the Pleistocene lowstand of the sea level, theEocene sediments, which were inundated by seawaterduring the last interglacial, now surround the MVApaleochannel throughout its length as shown inFigure 9.

The hydrogeochemical consequence of this Pleisto-cene aquifer (i.e., MVA) being embedded in anEocene aquitard is that the aquifer acts as a naturalhydraulic drain throughout the valley. When the citybegan to extract groundwater from the MVA in 1914(Fay, 1914), the natural process of freshening theEocene sedimentary rock was enhanced throughinduced seepage to the MVA, causing the MVA tobecome somewhat brackish while the Eocene sedi-ments underwent freshening. This process of drainagevia the MVA and freshening of the Eocene sedimentswas accelerated by the heavy pumping that occurredduring the remediation of the MTBE/TBA plume thatmigrated from the MVT to the DB monitoring wellsshown in Figure 1. Not only was brackish waterinduced to flow into the MVA by this pumping, butalso the groundwater became more brackish (i.e.,higher TDS) due to the effects of the bioremediationof the gasoline released from the MVT (see above).

Head measurements in SDAQ indicate a strongupward hydraulic gradient across the Friars Forma-tion into the Quaternary sediments. This gradientproduces an artesian head, exhibited by the piezo-meters beneath the Quaternary/Friars contact being,10–15 ft (3–5 m) above ground surface. Thus,SDAQ monitors hydraulic head and GWQ in a re-gional groundwater discharge area being recharged inthe Peninsular Range to the east. The FriarsFormation, a poorly indurated sandstone with 20percent clay-sized particles, is regarded as an aquitard(K , 1E205 m/s) in the valley. By contrast, thePleistocene sands and gravels of the MVA have muchhigher hydraulic conductivities (K . 1E204 m/s).The MVA acts hydraulically as a line sink throughthe center of the Friars Formation such that Friars’groundwater has slowly drained into the MVAnaturally or has been induced to seep more rapidlyby groundwater extraction in the MVA.

Depth profiles of TDS, specific electrical conduc-tance, chloride/boron ratios, 14C, tritium and sulfurisotopic values (d34SO4), and specific electricalconductance (SEC) for SDAQ are presented inFigure 12. The non-detect tritium results in the lowerfour ports (J3 through J6) of the SDAQ piezometersindicate the groundwater was recharged before 1953.The uppermost port, J7, had small amounts of tritium(near 6 TUs), which suggests some groundwaterrecharge occurred after 1953. The 14C results in the



lower four ports (, 10 pmc) indicates the groundwa-ter is relatively old and has not been recentlyrecharged.

We note that SEC, TDS, and the chloride/boronratio indicate a clear trend in the upper 300 ft (100 m)towards a more saline shallow groundwater at theFriars/Quaternary contact (i.e., J7, 12 m or 40 ft bgs),which is confirmed by a similar trend in the sulfurisotope data towards the seawater d34SO4 value of21% (Clark and Fritz, 1997, p. 140). Sulfate-sulfurisotope results are reported relative to the ViennaCanyon Diablo Troilite.

We associate these features with the marine in-undation of 120 ka, which, we propose, produced thebrackish groundwaters of the Friars Formation. Inthis hypothesis, the high TDS at the Friars/Quater-nary contact reflects inundation of the FriarsFormation by seawater during the last interglacial(see Figure 11) of 120 ka, when there was a sufficientlyhigh head of seawater (specific gravity 5 1.02) to sinkthrough the Quaternary sediments and be trapped atthe Friars/Quaternary contact. This trapping isclearly evident in the resistivity and gamma logs fromthe SDAQ well construction diagram (Figure 9H;Aqua Culture Monitoring Well, http://ca.water.usgs.gov/projects/sandiego/wells/summary.html).

In this assessment, we use ionic ratios andconcentrations, as well as the stable water isotopesdiscussed above (Figure 13), to elucidate processesaffecting the observed patterns of GWQ. Therelationships between bromide and chloride (Daviset al., 1998), as well as sodium and chloride (seeAppelo, 1994; Ravenscroft and McArthur, 2004;Andersen et al., 2005; and chapter 6 in Appelo andPostma, 2005), are common tools used to identifyfreshening of brackish aquifers.

Bromide and chloride both have high aqueoussolubilities, and their movement in brackish or freshgroundwater is considered to be conservative (Daviset al., 1998). Their relationship (in mg/L) is shown inFigure 15, although the DWR (1965) data cannot beshown because bromide was not analyzed by DWR.The seawater ratio of Cl:Br is 284 (by mass) based onthe seawater concentrations shown in Hem (1985).The ratio was developed for the range of Cl shown inthe figure. The data from the MVA and the USGSGAMA wells plot on or slightly below the seawaterratio line and suggest there is a seawater Cl:Brcomponent in the groundwater.

Table 8. Ranges of results for the parameters determined for thealluvial wells in the San Diego Hydrogeologic Province as part of theGAMA Project (Wright et al., 2005).

Groundwater QualityParameter (Unit)

Range in Alluvial BasinWells

Dissolved oxygen (mg/L) 0.1–5.5pH (standard units, field measured) 6.8–7.5Specific conductance (mS/cm @

25uC, field) 805–2,787Total hardness (mg/L as CaCO3) 201–922Alkalinity (mg/L as CaCO3) 133–300Nitrate + nitrite (mg/L as N) 0.04–9.14Dissolved organic carbon (mg/L) 0.4–2.1Major ions in ppm

TDS (residue on evaporation, mg/L) 685–1,800Calcium (mg/L) 43–234Magnesium (mg/L) 22.5–81.6Potassium (mg/L) 2.51–9.1Sodium (mg/L) 68–295Bromide (mg/L) 0.17–1.74Chloride (mg/L) 113–540Sulfate (mg/L) 61.7–421

Trace elements in ppbArsenic (mg/L) 0.5–2.0Barium (mg/L) 21–144Boron (mg/L) 51–228Iron (mg/L) 4–2,120Manganese (mg/L) 0.1–492Strontium (mg/L) 409–1,130Uranium (mg/L) 0.46–7.91

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

Parameter (Unit)

Well

DB-1 DB-2

Screen elevation (ft amsl) +6 to 214 4.5 to 220.5Sample date Apr-11 Jun-11Temperature (uC) — —pH (pH units) 7.7 (lab) 7.1 (lab)Dissolved oxygen (mg/L) — —Specific electrical conductance

(mS/cm) 2,710 (lab) 2,650 (lab)Total dissolved solids (mg/L) 1,540 1,640Alkalinity (mg/L CaCO3) 416 325Sodium (mg/L) 233 263Potassium (mg/L) — —Calcium (mg/L) 160 172Magnesium (mg/L) 67.6 62.1Iron (mg/L) 4.11 7.83Manganese (mg/L) 1.66 2.61Chloride (mg/L) 545 555Sulfate (mg/L) 203 212Bicarbonate (mg/L) 472 374Nitrate-N (mg/L) ,0.05 0.14Arsenic (mg/L) ,0.002 ,0.002SI CaCO3, calcite 0.9 0.3SI FeCO3, siderite 1.4 1.1SI MnCO3, rhodochrosite 1.2 0.8SI Fe(OH)3, iron hydroxide ,0.0 ,0.0SI MnO2, pyrolusite ,0.0 ,0.0Na/Cl 0.66 0.73

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



Figures 16 through 18 present evidence of ground-water within the Lower San Diego River Valley beingfreshened over the past 20,000 years. The exception iswell 18Q3, which was installed in the MVA (seeFigure 10); the paleochannel aquifer acts as the drainfor the Eocene sediments and thus becomes in-creasingly saline over time. Ellis and Lee’s (1919)composite sample is used to represent the freshwaterend member, i.e., background conditions.

Figure 16 shows the sodium and chloride concentra-tions for the DWR, SDAQ, RWGC2, and DB samples.A two-component mixing line is shown betweenseawater from Hem (1985; Na 5 10,500 mg/L, Cl 5

19,000 mg/L) and the Ellis and Lee (1919) sample (Na5 54 mg/L and Cl 5 85 mg/L). Ravenscroft andMcArthur (2004) and Anders et al. (2014) have useda similar graphical technique to identify aquifers thatare being either freshened or salinized.

Figure 16 shows that 5D1, 18Q3, and 17D1 plot onthe mixing line and have a seawater signature. Thedata 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 concentra-tion relative to Cl. The data points below the linerepresent groundwater that is being influenced by the

addition of more saline water. RWGC2 shows rela-tively less evidence of freshening, which is consistentwith the likelihood that it is undergoing modernseawater intrusion that comprises about 5 percent ofthe sample. Those samples close to the freshwater endmember, the Ellis and Lee (1919) sample, indicate thatfreshening is well advanced. 18Q3 reflects salineseepage into the MVA and is thus more saline thanwould otherwise be expected.

Freshening of seawater-inundated sediments is alsoapparent in Figure 17, which presents evidence of thecation exchange processes that are to be expected, suchas the replacement of seawater sodium in the Eocenesedimentary rock by freshwater calcium (Appelo, 1994):

1=2Ca2zz Na{X?1=2Ca{X2z Naz

where X represents the ion exchanger, such as clayminerals or other oxide surfaces in the sediments, i.e.,desorption of Na from marine-inundated sedimentsby freshwater Ca.

The mixing line in Figure 17 is drawn with seawaterdata from Hem (1985; Ca 5 410 mg/L) and from theEllis and Lee (1919) results (Ca 5 57 mg/L). Datapoints that lie below the mixing line are indicative of

Figure 13. Stable water isotopes in the Lower San Diego River Valley. The groundwater data from city monitoring wells (DB-2A, MVA atDB, Friars GW, MVA Qualcomm, and MW-1) are from July and October 2014. The USGS SDAQ data are from 2010, while the USGSdata from the River Walk Golf Course (RWGC2) and that from the Mission wells are from 2005. River water samples are USGS data from2004–2010, with one collected in March 2015 at the DB site (identified as DB Mar-15); sampling locations are shown in Figure 1.



freshwater flushing, and data points that lie abovethe mixing line are indicative of marine inundation(Ravenscroft and McArthur, 2004) and brackish waterflushing. Sample RWGC2, which contains ,5 percentmodern seawater, represents a saline end member ofthe freshening process, while the Ellis and Lee (1919)sample represents the freshwater end member. Verygenerally, the freshening process of the Eocenesediments progresses from RWGC2 towards 18Q3,which represents saline drainage within the MVA.

Similarly, Figure 18 illustrates the reaction involv-ing the desorption of the borate anion in Eocenesediments by freshwater bicarbonate:

HCO{3 zB(OH)4{X~B(OH){4 zHCO3{X

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

ionic strength, equilibrium re-adjustment, and, possi-bly, competitive exchange with HCO3/CO3.” Theyconcluded that “enrichment of both Na and B resultsfrom desorption from mineral surfaces in response toflushing by fresh groundwater of previously salineaquifers.” A comparison of Figures 16 and 18 showssome similarity in Na and B desorption in therelationship of samples 17D1 and 18N1 to 18Q3.

DISCUSSION: BACKGROUND, BASELINE,AND AMBIENT GWQ

“Background” and “baseline” are adjectives usedto describe the GWQ prior to anthropogenic de-velopment that might affect the chemical compositionof groundwater. We believe that the following termsare consistent with North American usage:

N Background describes the pristine GWQ derivedfrom natural geological, biological, or atmospheric

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



sources in the absence of identifiable anthro-pogenic influences (see Langmuir, 1997, p. 304),whereas

N baseline describes the GWQ at the beginning ofmonitoring 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 ofhuman activities, e.g., coliform bacteria from septictanks or nitrate from fertilizer applications, althoughthese analytes may or may not be reported. It isnoteworthy that European usage of “background”and “baseline” is exactly the opposite of NorthAmerican usage (see Edmunds and Shand, 2008).

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

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



In addition, some institutions such as the USGSuse a third term:

N Ambient GWQ is that GWQ measured at sometime and place without any assumption beingmade as to anthropogenic influences.

The USGS (2013, p. 1) states that the CaliforniaGroundwater 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 17. Calcium and chloride concentrations in the Lower San Diego River Valley prior to biodegradation-induced elevation of totaldissolved solids. The solid line is a mixing line between the background sample of Ellis and Lee (1919) and seawater.

Figure 18. Evidence of freshwater bicarbonate exchange for seawater boron in the DWR (1965) and SDAQ J7 (2010) data caused byaquifer “freshening,” i.e., displacement of brackish groundwater by freshwater recharge. The simple linear regression indicates that, ifRWGC2 is excluded, 81 percent of the variance is explained by this ion-exchange reaction. DB boron values were not available for 2011.



ents in groundwater.” Thus, “baseline” and “ambient”GWQ can indicate the same GWQ; the first termmerely indicates that it is the GWQ measured prior tosome event that may affect it. Our purpose is identifythe particular meaning of each term with respect to thegroundwater samples considered in Tables 5 through 9.

The 1915 USGS analysis shown in Table 5 is thatof a fresh groundwater with relatively low TDS (,400mg/L) and trace quantities of iron and 1.0 mg/L ofnitrate, indicating an oxygenated groundwater. Thecommon occurrence of dissolved oxygen (DO) andnitrate is discussed by Appelo and Postma (2005, pp.458–464) and by Langmuir (1997, p. 418); it is alsowell documented in the technical literature (e.g.,Jackson et al., 1990). The data in Table 5 representwhat is referred to for regulatory purposes as“background water quality conditions” (San DiegoRWQCB, 2005).

Tables 6 and 9 represent subsequent sampling ofthe MVA over approximately 100 years following theUSGS sampling event of 1915. Table 6 data werecollected during 1955–1965 by DWR, by which time(1959) the city well field shown in Figure 10 had beenabandoned and the wells destroyed, although 18Q3was sampled before abandonment (see DWR, 1959,p. B-20). Table 9 represents data collected during2011 from the city’s pilot well field (DB-1 and DB-2),which at the time was beginning to show evidence ofincreased contamination (TDS, MTBE, and TBA)that we associate with the MVT gasoline release of1987–1991. Therefore, these three sets of dataas represented by Tables 5, 6, and 9 provide(1) background GWQ data (i.e., the 1915 USGSsampling in Table 5) and (2) two sets of supplementalanalyses representing the evolution of GWQ in theMVA approximately 45 and 96 years later.

The background GWQ of the MVA that emergesfrom these studies suggests that the MVA groundwa-ter in 1915 was rather typical of alluvial aquifersfound throughout the U.S. Southwest, i.e., aquifersediments derived from plutonic rocks producinga sediment rich in feldspars and silica (“felsic”). In hisstudy of the Southwestern alluvial basins, F.N.Robertson (1991, p. C-16) of the USGS commentedthat “The basin-fill sediments were transported intothe basin and deposited under oxidizing conditions.”Robertson’s model identifies groundwater in therecharge area as a calcium-bicarbonate water withpH 5 7.2, DO ranging from 3 to 7 mg/L, and a meanand standard deviation of TDS 5 495 6 68 mg/L.The major geochemical reactions in the rechargeareas according to Robertson (1991, p. C86) are: (1)generation of carbonic acid in the recharge area fromthe dissolution of soil-zone carbon dioxide, includingplant respiration, in the recharging groundwater

(CO2 + H2O 5 H2CO3), producing an acidicgroundwater (pH , 6); (2) weathering of feldsparsand ferromagnesian minerals; (3) dissolution ofcarbonate minerals; and (4) formation of montmoril-lonite clay and iron oxides.

Table 5, showing the USGS 1915 analysis, is anaccurate representation of groundwater acquiredthrough Robertson’s three processes and indicatesa TDS , 400 mg/L and the presence of DO. Thesetwo parameters—TDS and DO—concisely define themajor ion and redox state in the background GWQ ofthe 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 continentalPleistocene sediments has a TDS concentration fallingwithin 200 to 600 ppm.” The background GWQ ofthe Friars Formation is unknown, but 17D1 and17D2 had TDS , 2,000 mg/L in the 1960s (Table 4).

The baseline GWQ in the MVA prior to urbandevelopment of the valley in the 1960s is representedby sample 18Q3 in Table 6, which is similar to theUSGS GAMA data shown in Table 8. This high-quality groundwater existed in the MVA untilcontamination from the MVT caused its deteriora-tion, although some deterioration may have beencaused by agricultural development in the area nowoccupied by the Qualcomm Stadium, e.g., the lownitrate concentrations in Table 6. Groundwaterextraction by the city prior to WWII would havecaused brackish water inflow into the MVA from theadjacent Friars Formation and is likely the reason forthe increase in TDS from ,400 mg/L in 1915 to,1,100 mg/L in 1955. In addition, the pavementsurrounding Qualcomm Stadium would have de-creased the infiltration of DO and low-TDS rechargeto the MVA after the mid-1960s. Consequently,urban development of Mission Valley in the 1960sled to further deterioration of the MVA prior to therelease of fuels from the MVT.

Because knowledge of the MVA used by the cityfaded from memory after WWII, the perception ofrelatively high TDS concentrations in the MissionValley groundwaters developed. This was likely theresult of a review of the DWR reports that failed todiscriminate between the Pleistocene MVA andEocene formations, where TDS concentrations rangedup to 3,485 mg/L. The mean and standard deviationof 75 TDS samples reported by DWR (1965) were1,694 6 723 mg/L, i.e., about the same as in thepresent city monitoring well DB-2 (Table 9).

Of particular interest is that all three samplescollected and analyzed by DWR (1959) from thepresent Qualcomm Stadium area shown in Table 6(17D1, 17D2, and 18Q3) contained measurabledissolved nitrate. As noted above (Appelo and



Postma, 2005, pp. 458–464; Langmuir, 1997, p. 418;Jackson et al., 1990), the presence of nitrate indicatesthe probable presence of DO in these groundwaters atthat time prior to urban development.

The USGS GAMA data (Table 8) indicate theambient GWQ in coastal southern California alluvialaquifers that were not affected by the marine invasionduring the last interglacial. All six USGS GAMAwells contained some DO, although their mean wasonly 1.6 mg/L. The 1955 DWR sample from 18Q3also appears to have contained DO because of thepresence of nitrate in the sample. The six AlluvialBasin wells sampled by the USGS during its GAMAsurvey shown in Table 8 had a mean TDS concen-tration of only 1,021 mg/L, which is similar to thatreported for 18Q3 (1,105 mg/L) in 1955 (DWR, 1965)and displayed in Table 6.

The ranges in Table 8 provide guidance as to whatmight be considered reasonable present-day concen-trations within MVA groundwaters. That is, valuesbeyond the ranges reported are likely due to either (1)poor sampling technique, such as incorporation ofsub-micron fines, resulting in erroneous concentra-tions, or (2) contamination, e.g., the migration of theuncaptured plume of MTBE, TBA, and TDS fromthe MVT gasoline releases.

In 2005, the San Diego Regional Water QualityControl Board (San Diego RWQCB, 2005, p. 3)ordered that the gasoline contamination from thegasoline tank farm (MVT; see Figure 1) that wascontaminating the MVA should be remediated “toattain background water quality conditions” by theend of 2013. The board defined these backgroundconditions as “the concentrations or measures ofconstituents or indicator parameters in water or soilthat have not been affected by waste constituents/pollutants from the Site”; i.e., from the MVT.However, it has been long established (Bennettet al., 1993; Borden et al., 1995; and McMahonet al., 1995) that all inorganic constituents are affectedby the intrinsic bioremediation such as that which hasoccurred within that part of the MVA contaminatedby the gasoline components. This is because theoxidation of hydrocarbons produces carbon dioxide,which causes a chain of hydrogeochemical processes:It dissolves in water and thus lowers the pH byforming carbonic acid, causing dissolution of aquiferminerals and raising TDS concentrations. Similarly,the dissolved hydrocarbons change the redox state byreducing DO, nitrate, and sulfate and causing thereductive dissolution of iron and manganese oxideson aquifer minerals. Thus, this definition of back-ground GWQ by the regulator is based on a falseassumption; i.e., there is no effect of intrinsicbiodegradation of organic hydrocarbons on the

concentrations of inorganic groundwater constitu-ents, and the Board’s definition of backgroundconditions is therefore incorrect.

The attainment of “background water qualityconditions” could theoretically be associated with theattainment of the 1915 GWQ measured by the USGS(Ellis and Lee, 1919), i.e., TDS , 400 mg/L andpresence of DO, which is the real background GWQ.However, this would likely require the recharge ofdistilled water to the aquifer but recharge of anywater—treated or not—has been continually rejectedby the owner of the MVT and by the board in recentyears. The attainment of the baseline GWQ, repre-sented by 18Q3 (TDS , 1,100 mg/L, DO present)measured in 1955, prior to the development of theMVT, is however practical given today’s desalinizationtechnology. It should be noted that the DWR (1965,pp. 42–43) advised the San Diego RWQCB that theEocene sediments of the valley were brackish and that“the most practical way to alleviate the problem ofpost-nate [i.e., brackish] water seepage is to increasethe relative head of ground waters by means of groundwater recharge of Mission San Diego Basin” (i.e., theLower San Diego River Valley).

SUMMARY AND CONCLUSIONS

In order to explain the current pattern of GWQ inthe Lower San Diego Valley, we have reconstructedthe likely Quaternary evolution beginning with thelast interglacial about 120,000 years ago. Pleistocenesea-level fluctuations have caused periodic incision ofLower San Diego River channels and inundation byrising seas. From water-well data, we have identifiedchannels that developed during the LGM that cutinto the Eocene and early Quaternary sediments. Thischannel probably extended far offshore relative to thepresent coastline as it graded to a lowered base (sea)level about 17,000 years ago, similar to other majorcoastal rivers in southern California. These gravelsbecame the principal hydraulic unit in the city’s pre-WWII groundwater supply—the MVA.

We hypothesize that the inundation of the Eocenesediments, such as the Friars Formation, by theseshallow seas during the last interglacial is recordedin the brackish GWQ of water wells completed inthe valley sediments, with the exception of thosewells completed in the gravel paleochannel(s) thatdate to the LGM, during which the MVA wasdeposited. However, groundwater extraction fromMVA wells has induced brackish groundwater flowinto the MVA from the adjacent Friars Formationsediments, and biodegradation-induced naturalattenuation of hydrocarbons has further raisedTDS concentrations.



Several lines of evidence support this hypothesis:

N Abbott’s (1999, pp. 202–203) popular historyof regional geology illustrates the marine in-vasion of 120 ka, confirms the geographic extentof the invasion, and outlines the same causes ofsea-level rise since the LGM and tectonic upliftover ,120,000 years.

N Similar gravel aquifers to the MVA were laiddown in all the major southern Californianvalleys following the LGM; recent drillingevidence from the MVA has confirmed this.

N The USGS 1915 sampling of the GWQ in theMVA showed clearly that the aquifer had lowTDS and was oxygenated prior to the beginningof large-scale municipal extraction.

N Numerous water-supply wells sampled by DWRin the 1950s, i.e., prior to urbanization of thevalley, indicate TDS concentrations varyingfrom 700 to 3,500 mg/L; the lower TDS valueswere associated with the alluvial gravel aquifer.

N DWR also concluded in the 1960s that marinewaters explained the brackish GWQ of theLower San Diego River Valley and that artificialrecharge of water was necessary to keep theTDS concentrations low.

N The USGS SDAQ (Aquaculture) multi-depthmonitoring well in the valley records this brackishwater (TDS , 2,000 mg/L) at the contact of theQuaternary and Eocene sediments.

We conclude that the confusion concerning theGWQ in the valley is due to a loss of institutionalmemory of documentation and a failure to re-visitthe excellent work of the DWR in the 1950s and1960s conducted for the present San Diego RegionalWater Quality Control Board. While new informa-tion is always helpful, it does not negate thebountiful data readily available in the publishedliterature. Restoration of the MVA as a municipalwater supply is being planned. Evidence of stormrunoff recharge events is apparent in stable waterisotope data from new monitoring wells; however,the nature of the recharge process is uncertainbecause of limited data at present. High residualTDS concentrations will require a significant water-treatment initiative to make MVA groundwaterssuitable for public consumption in the area of thepre-WWII city well field.

ACKNOWLEDGMENTS

We are grateful to Anna Fyodorova, RobAnders, and Roy Shlemon for their most helpfulreviews of the original and revised drafts of thismanuscript.

REFERENCES

ABBOTT, P. L., 1999, The Rise and Fall of San Diego: 150 MillionYears of History Recorded in Sedimentary Rocks: SunbeltPublications: San Diego, CA, 231 p.

ABBOTT, P. L. AND SMITH, T. E., 1989, Sonora, Mexico, source forthe Eocene Poway Conglomerate of southern California:Geology, Vol. 17, No. 4, pp. 329–332.

ALLEN, J. R. L., 1983, Studies in fluviatile sedimentation: Bars, barcomplexes and sandstone sheets (low sinuosity braidedstreams) in the Brownstones (L. Devonian), Welsh Borders:Sedimentary Geology, Vol. 33, No. 4, pp. 237–293.

AMERICAN PETROLEUM INSTITUTE (API), 2009, Hydraulic FracturingOperations—Well Construction and Integrity Guidelines: APIGuidance Document HF-1, First Edition, October 2009,American Petroleum Institute, Washington, D.C., 36 p.

ANDERS, R.; DANSKIN, W. R.; AND MENDEZ, G. O., inreview Geologic, Hydrologic, and Water Quality Data fromSurface-Water and Groundwater Sites in San Diego County,California, 2002–2009. To be published in the USGSData Series by the US Geological Survey, California WaterScience Center, 4165 Spruance Rd., Suite 200, San Diego, CA92101.

ANDERS, R.; MENDEZ, G. O.; FUTA, K.; AND DANSKIN, W. R., 2014,A geochemical approach to determine sources and movementof 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 theseawater/freshwater interface of a sandy aquifer: Geochimicaet Cosmochimica Acta, Vol. 69, No. 16, pp. 3979–3994.

APPELO, C. A. J., 1994, Cation and proton exchange, pHvariations, 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, Groundwaterand Pollution, 2nd ed.: A. A. Balkema Publishers, NewYork, 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 Geo-chemistry, Vol. 8, No. 6, pp. 529–549.

BLAIR, T. C. AND MCPHERSON, J. G., 1994, Alluvial fans and theirdistinction from rivers based on morphology, hydraulicprocesses, 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, Geo-chemical indicators of intrinsic bioremediation: GroundWater, 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: WaterCode sections 10780-10782.3, State of California, Sacra-mento, CA.

CLARK, I. D. AND FRITZ, P., 1997, Environmental Isotopes inHydrogeology: Lewis Publishers, Boca Raton, FL, 328 p.

DANIEL B. STEPHENS AND ASSOCIATES, INC., 2014, LaboratoryReport, San Diego MVT: Unpublished consultant report, forINTERA, 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: SpectrumPetrographics Inc. Vancouver, Washington, June 22, 5 p.

DEPARTMENT OF WATER RESOURCES (DWR), 1959, Ground WaterGeology, San Diego River Valley: A Progress Report on



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), Depart-

ment 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 ofWater 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 coastalaquifer 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 AmericaSpecial Paper 454, pp. 319–344.

EINARSON, M. AND CHERRY, J. A., 2002, A new multilevelground 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 DepartmentArchives.

GEOFIRMA ENGINEERING LTD. AND INTERA INC., 2011, Hydro-

geology of the Mission Valley Aquifer: Revision 0, preparedfor the Office of the City Attorney and the Public UtilitiesDepartment, The City of San Diego, San Diego, CA, January25, 2011, 143 p.

GEOFIRMA ENGINEERING LTD. AND INTERA INC., 2013, Remedia-

tion of MTBE and TBA, Mission Valley Aquifer: Revision 0,prepared for the Office of the City Attorney and the PublicUtilities 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. GeologicalSurvey Water Supply Paper 2254, 263 p.

INTERA INC., 2014, Murphy Canyon Monitoring Well Installation

and Sampling Report: Revision 1, prepared for the WaterDepartment, City of San Diego, Water Resources andPlanning Division, October, Albuquerque, NM, 9 p., 1figure, 1 table, and 3 appendices.

JACKSON, R. E.; MUTCH, J. P.; AND PRIDDLE, M. W., 1990,Persistence of aldicarb residues in the sandstone aquifer ofPrince 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 andGeology Bulletin 200, p. 56.

KENNETT, J. P., 1995, Latest Quaternary benthic oxygen andcarbon isotope stratigraphy: Hole 893A, Santa BarbaraBasin, 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.

KERN, J. P. AND ROCKWELL, T. K., 1992, Chronology anddeformation of Quaternary marine shorelines, San DiegoCounty, California. In Fletcher, C. H. and Wehmiller, J. F.(Editors), Quaternary Coasts of the United States: Marine andLacustrine 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 andthe Urban Ocean, The Southern California ContinentalBorderland: Geological Society of America Special Paper454, 481 p.

MCMAHON, P. B.; VROBLESKY, D. A.; BRADLEY, P. M.; CHAPELLE,F. H.; AND GULLETT, C. D., 1995, Evidence for enhancedmineral dissolution in organic acid-rich shallow groundwater: Ground Water, Vol. 33, No. 2, pp. 207–216.

MIALL, A. D., 1985, Architectural-element analysis: A new methodof facies analysis applied to fluvial deposits: Earth-ScienceReviews, Vol. 22, No. 4, pp. 261–308.

MIALL, A. D., 1992, Alluvial deposits. In Walker, R. G. andJames, N. P. (Editors), Facies Models: Response to Sea LevelChange: 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 ofNorth America: Timing and paleoclimate: Geological Societyof America Bulletin, Vol. 114, No. 5, pp. 569–592.

PARKHURST, D. L. AND APPELO, C. A. J., 1999, User’s Guide toPHREEQC (Version 2)—A Computer Program for Specia-tion, Batch-Reaction, One-Dimensional Transport, and InverseGeochemical Calculations: U.S. Geological Survey Water-Resources Investigations Report 99-4259, 312 p.

RAVENSCROFT, P. AND MCARTHUR, J. M., 2004, Mechanism ofregional enrichment of groundwater by boron: The examplesof Bangladesh and Michigan, USA: Applied Geochemistry,Vol. 19, No. 9, pp. 1413–1430.

ROBERTSON, F. N., 1991, Geochemistry of Ground Water in AlluvialBasins of Arizona and Adjacent Parts of Nevada, New Mexico,and California: U.S. Geological Survey Professional Paper1406-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, SanDiego Region, April 13, Addendum No. 5, 8 p.

SHACKLETON, N. J., 1969, The last interglacial in the marine andterrestrial record: Proceedings of the Royal Society ofLondon B, Vol. 174, pp. 135–154.

SHLEMON, R. J., 1979, Late Pleistocene channel of the lower SantaMargarita River, San Diego County, California. In Fife,D. L. (Editor), Geological Guide of the San Onofre NuclearGenerating Station and Adjacent Regions of SouthernCalifornia: Pacific Sections, American Association of Petro-leum Geologists, Society of Economic Mineralogists andPaleontologists, and Society of Exploration Geophysicists,Guidebook Number 46, pp. A63–A70.

SLOTO, R. A., 2013, Baseline Groundwater Quality from 20Domestic Wells in Sullivan County, Pennsylvania, 2012: U.S.Geological Survey Scientific Investigations Report 2013-5085, 27 p.

TANAKA, H.; SMITH, T. E.; AND HUANG, C. H., 1984, The SantiagoPeak volcanic rocks of the Peninsular Ranges Batholith,southern California: Volcanic rocks associated withcoeval gabbros: Bulletin of Volcanology, Vol. 47, No. 1,pp. 153–171.

U.S. GEOLOGICAL SURVEY (USGS), 2013, California GroundwaterAmbient Monitoring and Assessment (GAMA) Program



Priority Basin Project: Shallow Aquifer Assessment: Fact Sheet2012-3136, 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 effectsand application to hydrologic investigations in southwesternCalifornia: Water Resources Research, Vol. 33, No. 7,pp. 1721–1729.

WRIGHT, M. T. AND BELITZ, K. 2011, Status and Understanding ofGroundwater Quality in the San Diego Drainages Hydro-geologic Province, 2004: California GAMA Priority BasinProject: U.S. Geological Survey Scientific InvestigationsReport 2011-5154, 100 p.

WRIGHT, M. T.; BELITZ, K.; AND BURTON, C. A., 2005, CaliforniaGAMA Program—Ground-Water Quality in the San DiegoDrainages Hydrogeologic Province, California, 2004: U.S.Geological Survey Data Series 129, 91 p.



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, ForestCover, Recovery

ABSTRACT

Watersheds recently burned by wildfires can besusceptible to debris flow, although little is knownabout how long this susceptibility persists and how itchanges over time. We use a compilation of 75 debris-flow response and fire-ignition dates, vegetation andbedrock class, rainfall regime, and initiation processfrom throughout the western United States to addressthese issues. The great majority (85 percent) of debrisflows occurred within the first 12 months followingwildfire, with 71 percent occurring within the first 6months. Seven percent of the debris flows occurredbetween 1 and 1.5 years after a fire, or during thesecond rainy season to impact an area. Within thefirst 1.5 years following fires, all but one of the debrisflows initiated through runoff-dominated processes,and debris flows occurred in similar proportions inforested and non-forested landscapes. Underlyinggeologic materials affected how long debris-flowactivity persisted, and the timing of debris flowsvaried within different rainfall regimes. A second,later period of increased debris flow susceptibilitybetween 2.2 and 10 years after fires is indicated by theremaining 8 percent of events, which occurredprimarily in forested terrains and initiated largelythrough landslide processes. The short time periodbetween fire and debris-flow response within the first1.5 years after ignition and the longer-term responsebetween 2.2 and 10 years after fire demonstrate thenecessity of both rapid and long-term reactions byland managers and emergency-response agencies tomitigate hazards from debris flows from recentlyburned areas in the western United States.

INTRODUCTION

There are few vegetative communities throughoutthe world in which wildfire does not occur periodi-cally. Global fire images taken by the NationalAeronautic and Space Administration (NASA) Terrasatellite from March 2000 to April 2014 (EarthObservatory website, NASA, 2014) and forest ecol-ogy literature (e.g., Dwyer et al., 2000; Lloret andMari, 2001; Weisberg and Swanson, 2003; and Floydet al., 2004) illustrate this point. In the United States,vegetative communities affected by large wildfires inrecent years include, for example, sagebrush-grass-lands (Murphy Complex Fires, Idaho), spruce forestand woodlands (Taylor Complex Fires, Alaska),chaparral (Station Fire, California), Ponderosapine–Douglas fir (Hayman fire, Colorado), and pineplantations and swamp lands (Big TurnaroundComplex, Georgia) (NIFC, 2011).

A broad spectrum of potential hydrologic re-sponses can be triggered where wildfires burn onsteep slopes that are subsequently affected by rain-storms. At the most destructive end of the post-firerunoff and erosion spectrum, debris flows posea serious threat because they can move rapidly anddeliver a significant destructive force along their flowpaths (DeGraff et al., 2007; Cannon et al., 2011).Debris flows following a wildfire pose a particularhazard when burned watersheds are adjacent todensely populated areas (Cannon and DeGraff,2009; Cannon et al., 2009), but they can also affectless populated, rural settings (DeGraff et al., 2011).Since the 1970s, debris flows have been widelyrecognized as a post-wildfire phenomenon withinmuch of the western United States (Scott, 1971;Wells, 1987; Florsheim et al., 1991; Wohl andPearthree, 1991; Cannon and Reneau, 2000; Cannon,2001; Cannon et al., 2001, 2008, 2010; Meyer et al.,


2001; Wondzell and King, 2003; Gartner et al., 2005;Giraud and McDonald, 2007; and Wagner et al.,2013). Debris flows have also been documented fromburned watersheds in Canada, Australia, Spain,France, the Swiss Alps, and parts of the Mediterra-nean Basin (Conedera et al., 2003; Jordan andCovert, 2009; Fox, 2011; Nyman et al., 2011; Pariseand Cannon, 2012; Santi and Morandi, 2012; andGarcia-Ruiz et al., 2013).

Debris flows initiated from burned watersheds havebeen reported to have been triggered by a variety ofprecipitation conditions including high-intensity rain-fall during summer convective storms (DeGraff et al.,2011), long-duration winter and summer frontalstorms (Cannon et al., 2008), cells of high-intensityrainfall within frontal storms (Cannon et al., 2011;Kean et al., 2011), rapid snowmelt (Schulz et al.,2006), and intense rainfall on melting snow (Meyeret al., 2001; Shaub, 2001).

Moody and Martin (2009) used a synthesis ofmeasured sediment yields following wildland fire(including some debris flows) to demonstrate thatthe rainfall that results in sediment movement willvary spatially with prevailing rainfall characteristics;measured post-fire sediment yields varied with theseasonality of storm rainfall as well as with the short–recurrence interval rainfall intensities typical of anarea. Moody and Martin (2009) developed a mapshowing areas of similar rainfall conditions, or whatthey termed rainfall regimes, for the western UnitedStates.

Debris flows following wildfires have been found toinitiate by either progressive bulking of surface runoffwith sediment eroded from hillslopes and channels ormobilization of a discrete landslide mass triggered byinfiltration processes (Wells, 1987; Cannon et al.,2001; Meyer et al., 2001; Wondzell and King, 2003;and Parise and Cannon, 2012). A lack of discretelandslides bounded by shear surfaces at the heads ofthe great majority of post-wildfire debris flowsindicates the prevalence of surface runoff-erosionprocesses in their initiation, rather than infiltration-dominated processes. Debris flows initiated throughsurface runoff processes can be particularly erosive,with material entrained from both hillslopes andchannels (Santi et al., 2008; Parise and Cannon, 2012).

Although there exists considerable study related tohow vegetation recovers following wildfires (e.g.,Cerda and Doerr, 2005; Lentile et al., 2007), littlework has been done to evaluate how a fire-relatedincreased susceptibility to debris flow changes overtime. On a broad scale, Meyer et al. (2001) suggestedthat the timing of post-wildfire debris flows will bedriven by the initiation process; runoff-initiated debrisflows would be expected in the first few years after

a wildfire, while landslide-initiated failures wouldoccur several years after the event, as roots decayand soil shear strengths decrease. With slightly moreprecision, Cannon et al. (2009) postulated that theiremergency assessment of debris-flow hazards fromrecently burned areas is appropriate for the first 2years following fires, and Santi and Morandi (2012)found that most post-fire sedimentation in southernCalifornia occurs in the first year following fires.However, a more precise understanding of when anincreased likelihood of debris flows following wildfiresexists, and of how this likelihood changes over time, isimportant to land managers and emergency respon-ders tasked with minimizing risks to public safety andproperty (Santi et al., 2011; Moody et al., 2013).

In this article we use a compilation of data on thetiming of wildfires and post-fire debris-flow eventsthroughout the western United States to examine thetemporal distributions of post-fire debris flows interms of vegetative cover, underlying bedrock,initiation process, and rainfall regime. This workserves to better identify the timeframes during whichland managers and emergency-response personnelneed to be aware of potential hazards in differentsettings throughout the western United States.

DATA

Starting with an existing database of post-fire debrisflow events (Gartner et al., 2005), we reviewedpublished papers and our own field records, accessedthe database of historically significant wildfires(NIFC, 2011), and interviewed other researchers toidentify events for which both the date of the fireignition and the date (to the day) of a debris flow–specific response were documented. We were able tobuild a database of 75 fire-ignition date–debris-flowresponse pairs from 52 fires in Arizona, California,Colorado, Idaho, Montana, New Mexico, and Utah(Figure 1 and Table 1). The database includes eventpairs that we could confidently consider to haveoccurred as a result of the effects of wildfire and thatwere not the result of extreme rainfall events thatwould have triggered debris flows even without fireeffects; we incorporated data only in cases in whichabundant debris flows were not triggered fromadjacent unburned terrain or in which the authorsdocumented a higher density of debris flows within theburned area than within the unburned area. Althoughmany fires had only one triggering event that resultedin debris flows and are thus represented by a singleentry, fires in which multiple storms resulted inmultiple debris flows are represented by an entry foreach storm. For each entry in the database, wedocumented the dates of fire ignition and debris-flow

DeGraff, Cannon, and Gartner


response and, where available, the vegetation class(forested, non-forested, mixed), bedrock class (granit-ic, metamorphic, volcanic [including meta-volcanic],sedimentary, mixed), initiation process (infiltration orrunoff), and primary rainfall regimes (Moody andMartin, 2009). Table 2 includes the rainfall seasonal-ities and the range of 2-year recurrence, 30 minute–duration rainfall intensities associated with each of thefour primary rainfall regimes identified by Moody and

Martin (2009). Note that although more detailedinformation on vegetation configuration and under-lying materials may be available, it was necessary touse the rather broad classes identified above todevelop a consistent record.

The data compiled here include primarily oppor-tunistic documentation of post-fire debris-flow eventsfrom the literature and observations from ourmulti-year efforts maintaining rain gauge networks,

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

Post-Fire Debris Flow Timing


Ta

ble

1.

Da

tau

sed

toev

alu

ate

tim

ing

of

po

st-w

ild

fire

deb

ris

flo

wo

ccu

rren

ce,

incl

ud

ing

iden

tifi

cati

on

nu

mb

er,

fire

na

me,

sta

te,

da

teo

ffi

reig

nit

ion

,d

ate

of

deb

ris-

flo

wre

spo

nse

,n

um

ber

of

da

ys

an

dm

on

ths

bet

wee

nig

nit

ion

an

dre

spo

nse

,d

ebri

s-fl

ow

init

iati

on

pro

cess

,ve

get

ati

on

cla

ss,

bed

rock

cla

ss,

sou

rce

of

info

rma

tio

n,

an

dse

aso

na

lra

infa

llre

gim

e(

fro

mM

oo

dy

an

dM

art

in,

20

09

).

Tim

eB

etw

een

Ign

itio

na

nd

Deb

ris

Flo

w

Fir

eID

Fir

eN

am

eS

tate

Fir

eIg

nit

ion

Da

te1

Deb

ris

Flo

wD

ate

Da

ys

Mo

nth

s

Deb

ris

Flo

wIn

itia

tio

nP

roce

ssV

eget

ati

on

Cla

ss2

Bed

rock

Cla

ss3

So

urc

e

Sea

son

al

Ra

infa

llR

egim

e(M

oo

dy

an

dM

art

in,

20

09

)

1N

ort

hH

ills

Mo

nta

na

8/2

7/1

98

48

/31

/19

84

40

Ru

no

ffF

ore

sted

Met

am

orp

hic

Pa

rret

t,1

98

7P

lain

s2

Mo

un

tain

Ca

lifo

rnia

7/1

5/2

01

37

/21

/20

13

60

Ru

no

ffN

on

-fo

rest

edG

ran

itic

Th

isst

ud

yP

aci

fic

3S

eele

yU

tah

6/2

6/2

01

27

/7/2

01

21

10

Ru

no

ffF

ore

sted

Sed

imen

tary

R.

Gir

au

d,

ora

lco

mm

.,2

01

3S

ub

-Pa

cifi

c

4H

arv

ard

Ca

lifo

rnia

10

/4/2

00

51

0/1

7/2

00

51

30

Ru

no

ffN

on

-fo

rest

edM

eta

mo

rph

icG

ard

ner

eta

l.,

20

09

Pa

cifi

c5

Piu

teC

ali

forn

ia6

/29

/20

08

7/1

2/2

00

81

30

Ru

no

ffF

ore

sted

Mix

edD

eGra

ffet

al.

,2

01

1P

aci

fic

6M

ou

nta

inT

rail

Ca

lifo

rnia

10

/23

/19

78

11

/11

/19

78

19

1R

un

off

Fo

rest

edM

ixed

Wel

ls,

19

87

Pa

cifi

c7

Mo

nu

men

tA

rizo

na

6/1

2/2

01

17

/10

/20

11

28

1R

un

off

Fo

rest

edM

ixed

Yo

ub

erg

,2

01

4A

rizo

na

8W

illo

wA

rizo

na

6/2

4/2

00

47

/23

/20

04

29

1R

un

off

Mix

edV

olc

an

icP

eart

hre

e,2

00

4A

rizo

na

9S

chu

ltz

Ari

zon

a6

/20

/20

10

7/2

0/2

01

03

01

Ru

no

ffF

ore

sted

Vo

lca

nic

Yo

ub

erg

,2

01

4A

rizo

na

10

Pea

kA

rizo

na

6/1

0/1

98

87

/11

/19

88

31

1R

un

off

Fo

rest

edM

ixed

Wo

hl

an

dP

eart

hre

e,1

99

1A

rizo

na

3S

eele

yU

tah

6/2

6/2

01

27

/31

/20

12

35

1R

un

off

No

n-f

ore

sted

Sed

imen

tary

R.

Gir

au

d,

ora

lco

mm

.,2

01

3S

ub

-Pa

cifi

c

11

Asp

enA

rizo

na

6/1

7/2

00

37

/24

/20

03

37

1R

un

off

Fo

rest

edG

ran

itic

A.

Yo

ub

erg

,o

ral

com

m.,

20

14

Ari

zon

a

12

Pic

ken

sC

ali

forn

ia1

1/2

1/1

93

31

2/3

1/1

93

34

01

Un

kn

ow

nN

on

-fo

rest

edM

ixed

Ea

ton

,1

93

6P

aci

fic

13

Gra

nd

Pri

xa

nd

Old

Ca

lifo

rnia

11

/14

/20

03

12

/25

/20

03

41

1R

un

off

No

n-f

ore

sted

Mix

edB

rock

eta

l.,

20

07

Pa

cifi

c

14

Wa

llo

wA

rizo

na

5/2

9/2

01

17

/11

/20

11

43

1R

un

off

Fo

rest

edV

olc

an

icY

ou

ber

g,

20

14

Ari

zon

a1

5B

ak

erC

ali

forn

ia1

0/1

2/1

99

71

2/1

/19

97

50

2R

un

off

No

n-f

ore

sted

Sed

imen

tary

Ca

nn

on

,2

00

1P

aci

fic

16

So

uth

Ca

ny

on

Co

lora

do

7/1

1/1

99

49

/1/1

99

45

22

Ru

no

ffN

on

-fo

rest

edS

edim

enta

ryC

an

no

net

al.

,1

99

8P

lain

s2

Mo

un

tain

Ca

lifo

rnia

7/1

5/2

01

39

/9/2

01

35

62

Ru

no

ffN

on

-fo

rest

edG

ran

itic

Th

isst

ud

yP

aci

fic

9S

chu

ltz

Ari

zon

a6

/20

/20

10

8/1

6/2

01

05

72

Ru

no

ffF

ore

sted

Vo

lca

nic

Yo

ub

erg

,2

01

4A

rizo

na

18

Co

al

Sea

mC

olo

rad

o6

/8/2

00

28

/5/2

00

25

82

Ru

no

ffN

on

-fo

rest

edS

edim

enta

ryC

an

no

net

al.

,2

00

3P

lain

s1

9M

issi

on

ary

Rid

ge

Co

lora

do

6/9

/20

02

8/8

/20

02

60

2R

un

off

Fo

rest

edV

olc

an

icC

an

no

net

al.

,2

00

3A

rizo

na

7M

on

um

ent

Ari

zon

a6

/12

/20

11

8/1

1/2

01

16

02

Ru

no

ffM

ixed

Mix

edY

ou

ber

g,

20

14

Ari

zon

a2

0D

om

eN

ew Mex

ico

4/2

5/1

99

66

/26

/19

96

62

2R

un

off

Fo

rest

edV

olc

an

icC

an

no

net

al.

,1

99

7A

rizo

na

7M

on

um

ent

Ari

zon

a6

/12

/20

11

8/1

3/2

01

16

22

Ru

no

ffM

ixed

Mix

edY

ou

ber

g,

20

14

Ari

zon

a2

1G

lad

iato

rA

rizo

na

5/1

3/2

01

27

/14

/20

12

62

2R

un

off

Mix

edG

ran

itic

Yo

ub

erg

,2

01

4A

rizo

na

22

Ho

rses

ho

e2

Ari

zon

a5

/8/2

01

17

/11

/20

11

64

2R

un

off

Mix

edV

olc

an

icY

ou

ber

g,

20

14

Ari

zon

a1

1A

spen

Ari

zon

a6

/17

/20

03

8/2

3/2

00

36

72

Ru

no

ffM

ixed

Mix

edA

.Y

ou

ber

g,

ora

lco

mm

.,2

01

4A

rizo

na

3S

eele

yU

tah

6/2

6/2

01

29

/1/2

01

26

72

Ru

no

ffN

on

-fo

rest

edS

edim

enta

ryR

.G

ira

ud

,o

ral

com

m.,

20

13

Su

b-P

aci

fic

23

Po

pp

etC

ali

forn

ia9

/23

/19

97

12

/1/1

99

76

92

Ru

no

ffN

on

-fo

rest

edM

eta

mo

rph

icC

an

no

n,

20

01

Pa

cifi

c2

4U

nn

am

edC

ali

forn

ia1

0/2

9/1

93

01

/7/1

93

17

02

Un

kn

ow

nN

on

-fo

rest

edM

eta

mo

rph

icE

ato

n,

19

36

Pa

cifi

c2

5M

arg

ari

taC

ali

forn

ia9

/4/1

99

71

1/1

5/1

99

77

22

Ru

no

ffF

ore

sted

Vo

lca

nic

Ca

nn

on

,2

00

1P

aci

fic



Tim

eB

etw

een

Ign

itio

na

nd

Deb

ris

Flo

w

Fir

eID

Fir

eN

am

eS

tate

Fir

eIg

nit

ion

Da

te1

Deb

ris

Flo

wD

ate

Da

ys

Mo

nth

s

Deb

ris

Flo

wIn

itia

tio

nP

roce

ssV

eget

ati

on

Cla

ss2

Bed

rock

Cla

ss3

So

urc

e

Sea

son

al

Ra

infa

llR

egim

e(M

oo

dy

an

dM

art

in,

20

09

)

26

Cer

roG

ran

de

New M

exic

o5

/4/2

00

07

/16

/20

00

73

2R

un

off

Fo

rest

edV

olc

an

icC

an

no

net

al.

,2

00

1A

rizo

na

28

Bu

llo

ckA

rizo

na

5/2

0/2

00

28

/5/2

00

27

73

Ru

no

ffF

ore

sted

Gra

nit

icA

.Y

ou

ber

g,

ora

lco

mm

.,2

01

4A

rizo

na

27

Sa

yre

Ca

lifo

rnia

11

/20

/20

08

2/5

/20

09

77

3R

un

off

No

n-f

ore

sted

Sed

imen

tary

Th

isst

ud

yP

aci

fic

19

Mis

sio

na

ryR

idg

eC

olo

rad

o6

/9/2

00

28

/29

/20

02

81

3R

un

off

Fo

rest

edM

eta

mo

rph

icC

an

no

net

al.

,2

00

3A

rizo

na

27

Sa

yre

Ca

lifo

rnia

11

/20

/20

08

2/1

4/2

00

98

63

Ru

no

ffN

on

-fo

rest

edS

edim

enta

ryT

his

stu

dy

Pa

cifi

c1

9M

issi

on

ary

Rid

ge

Co

lora

do

6/9

/20

02

9/5

/20

02

88

3R

un

off

Fo

rest

edS

edim

enta

ryC

an

no

net

al.

,2

00

3A

rizo

na

47

Hem

lock

Ca

lifo

rnia

7/5

/19

97

10

/2/1

99

78

93

Ru

no

ffN

on

-fo

rest

edS

edim

enta

ryC

an

no

n,

20

01

Pa

cifi

c1

9M

issi

on

ary

Rid

ge

Co

lora

do

6/9

/20

02

9/7

/20

02

90

3R

un

off

Fo

rest

edG

ran

itic

Ca

nn

on

eta

l.,

20

03

Ari

zon

a

29

Pa

ub

aC

ali

forn

ia8

/31

/19

97

12

/1/1

99

79

23

Ru

no

ffN

on

-fo

rest

edM

eta

mo

rph

icC

an

no

n,

20

01

Pa

cifi

c2

2H

ors

esh

oe

2A

rizo

na

5/8

/20

11

8/1

3/2

01

19

73

Ru

no

ffM

ixed

Vo

lca

nic

Yo

ub

erg

,2

01

4A

rizo

na

30

Mo

lera

Ca

lifo

rnia

8/1

/19

72

11

/12

/19

72

10

33

Ru

no

ffN

on

-fo

rest

edM

eta

mo

rph

icC

lev

ela

nd

,1

97

3P

aci

fic

31

Sta

tio

nC

ali

forn

ia8

/26

/20

09

12

/10

/20

09

10

64

Ru

no

ffN

on

-fo

rest

edM

ixed

Ca

nn

on

eta

l.,

20

11

Pa

cifi

c3

1S

tati

on

Ca

lifo

rnia

8/2

6/2

00

91

2/1

1/2

00

91

07

4R

un

off

No

n-f

ore

sted

Mix

edC

an

no

net

al.

,2

01

1P

aci

fic

17

Joh

nst

on

Pea

kC

ali

forn

ia7

/20

/19

60

11

/5/1

96

01

08

4R

un

off

Fo

rest

edV

olc

an

icD

oeh

rin

g,

19

68

Pa

cifi

c2

0D

om

eN

ew Mex

ico

4/2

5/1

99

68

/19

/19

96

11

64

Ru

no

ffF

ore

sted

Met

am

orp

hic

Ca

nn

on

eta

l.,

19

97

Ari

zon

a

31

Sta

tio

nC

ali

forn

ia8

/26

/20

09

1/1

7/2

01

01

44

5R

un

off

No

n-f

ore

sted

Mix

edC

an

no

net

al.

,2

01

1P

aci

fic

31

Sta

tio

nC

ali

forn

ia8

/26

/20

09

1/1

9/2

01

01

46

5R

un

off

No

n-f

ore

sted

Mix

edC

an

no

net

al.

,2

01

1P

aci

fic

31

Sta

tio

nC

ali

forn

ia8

/26

/20

09

1/2

0/2

01

01

47

5R

un

off

No

n-f

ore

sted

Mix

edC

an

no

net

al.

,2

01

1P

aci

fic

44

Mo

tor

Ca

lifo

rnia

8/2

5/2

01

11

/21

/20

12

14

95

Ru

no

ffN

on

-fo

rest

edM

eta

mo

rph

icT

his

stu

dy

Pa

cifi

c3

1S

tati

on

Ca

lifo

rnia

8/2

6/2

00

92

/5/2

01

01

63

5R

un

off

No

n-f

ore

sted

Mix

edC

an

no

net

al.

,2

01

1P

aci

fic

49

Rim

Ca

lifo

rnia

8/1

7/2

01

32

/26

/20

14

19

36

Ru

no

ffF

ore

sted

Gra

nit

icT

his

stu

dy

Pa

cifi

c2

1H

op

per

Ca

lifo

rnia

8/5

/19

97

2/1

5/1

99

81

94

6R

un

off

No

n-f

ore

sted

Sed

imen

tary

Ca

nn

on

,2

00

1P

aci

fic

33

Lo

ga

nC

ali

forn

ia8

/4/1

99

72

/15

/19

98

19

57

Ru

no

ffN

on

-fo

rest

edS

edim

enta

ryC

an

no

n,

20

01

Pa

cifi

c3

4U

nn

am

edC

ali

forn

ia6

/16

/19

68

1/1

/19

69

19

97

La

nd

slid

eN

on

-fo

rest

edS

edim

enta

ryM

ort

on

,1

98

9P

aci

fic

37

Mid

dle

Ca

lifo

rnia

7/2

4/1

97

72

/8/1

97

81

99

7R

un

off

No

n-f

ore

sted

Sed

imen

tary

Wel

ls,

19

87

Pa

cifi

c3

6A

rch

Ro

ckC

ali

forn

ia8

/7/1

99

03

/3/1

99

12

08

7R

un

off

No

n-f

ore

sted

Mix

edD

eGra

ff,

19

94

Su

b-P

aci

fic

38

Ov

erla

nd

Co

lora

do

10

/29

/20

03

6/2

4/2

00

42

39

8R

un

off

Fo

rest

edG

ran

itic

Ga

rdn

eret

al.

,2

00

8P

lain

s3

9F

arm

ing

ton

Uta

h7

/15

/20

03

4/6

/20

04

26

69

Ru

no

ffF

ore

sted

Met

am

orp

hic

Gir

au

da

nd

McD

on

ald

,2

00

4S

ub

-Pa

cifi

c

40

Bea

rM

on

tan

a9

/1/2

00

07

/15

/20

01

31

71

1R

un

off

No

n-f

ore

sted

Met

am

orp

hic

Pa

rret

tet

al.

,2

00

3P

lain

s4

8Y

ello

wst

on

eM

on

tan

a8

/20

/19

88

7/9

/19

89

32

31

1R

un

off

Fo

rest

edV

olc

an

icM

eyer

an

dW

ells

,1

99

7P

lain

s4

1U

nn

am

edC

ali

forn

ia1

2/3

/19

27

11

/13

/19

28

34

61

2U

nk

no

wn

No

n-f

ore

sted

Mix

edE

ato

n,

19

36

Pa

cifi

c4

8Y

ello

wst

on

eM

on

tan

a8

/20

/19

88

8/1

0/1

98

93

55

12

Ru

no

ffF

ore

sted

Vo

lca

nic

Mey

era

nd

Wel

ls,

19

97

Pla

ins

42

Iny

oC

om

ple

xC

ali

forn

ia7

/15

/20

07

7/1

2/2

00

83

63

12

Ru

no

ffN

on

-fo

rest

edM

eta

mo

rph

icD

eGra

ffet

al.

,2

01

1S

ub

-Pa

cifi

c4

3M

oll

ieU

tah

9/1

/20

01

9/1

2/2

00

23

76

13

Ru

no

ffF

ore

sted

Met

am

orp

hic

McD

on

ald

an

dG

ira

ud

,2

00

2S

ub

-Pa

cifi

c

Table

1.Continued.



Tim

eB

etw

een

Ign

itio

na

nd

Deb

ris

Flo

w

Fir

eID

Fir

eN

am

eS

tate

Fir

eIg

nit

ion

Da

te1

Deb

ris

Flo

wD

ate

Da

ys

Mo

nth

s

Deb

ris

Flo

wIn

itia

tio

nP

roce

ssV

eget

ati

on

Cla

ss2

Bed

rock

Cla

ss3

So

urc

e

Sea

son

al

Ra

infa

llR

egim

e(M

oo

dy

an

dM

art

in,

20

09

)

3S

eele

yU

tah

6/2

6/2

01

27

/16

/20

13

38

51

3R

un

off

No

n-f

ore

sted

Sed

imen

tary

R.

Gir

au

d,

ora

lco

mm

.,2

01

3S

ub

-Pa

cifi

c

44

Mo

tor

Ca

lifo

rnia

8/2

5/2

01

11

1/1

7/2

01

24

50

15

Ru

no

ffN

on

-fo

rest

edM

eta

mo

rph

icT

his

stu

dy

Pa

cifi

c4

5G

av

iota

Ca

lifo

rnia

6/1

2/2

00

41

1/9

/20

05

51

51

7R

un

off

No

n-f

ore

sted

Gra

nit

icS

an

taB

arb

ara

Co

un

tyP

ub

lic

Wo

rks

Dep

t,2

00

5

Pa

cifi

c

32

Wil

lia

ms

Ca

lifo

rnia

9/2

/20

12

2/2

8/2

01

45

44

18

Ru

no

ffN

on

-fo

rest

edM

eta

mo

rph

icD

.S

tale

y,

ora

lco

mm

.,2

01

5P

aci

fic

46

Clo

ver

Ca

lifo

rnia

5/2

8/2

00

88

/26

/20

10

82

02

7R

un

off

No

n-f

ore

sted

Gra

nit

icW

ag

ner

eta

l.,

20

13

Su

b-P

aci

fic

35

Mil

lC

reek

Ca

lifo

rnia

11

/2/1

97

52

/10

/19

78

83

12

8U

nk

no

wn

No

n-f

ore

sted

Met

am

orp

hic

Bru

ing

ton

,1

98

2P

aci

fic

19

Mis

sio

na

ryR

idg

eC

olo

rad

o6

/9/2

00

24

/8/2

00

51

,03

43

4L

an

dsl

ide

Fo

rest

edS

edim

enta

ryS

chu

lzet

al.

,2

00

6A

rizo

na

39

Lo

wm

an

Ida

ho

7/1

/19

89

12

/20

/19

96

2,7

29

91

La

nd

slid

eF

ore

sted

Vo

lca

nic

Mey

eret

al.

,2

00

1S

ub

-Pa

cifi

c5

0S

tan

isla

us

Co

mp

lex

Ca

lifo

rnia

8/1

/19

87

1/2

/19

97

3,4

42

11

5L

an

dsl

ide

Fo

rest

edG

ran

itic

Th

isst

ud

yP

aci

fic

38

Ov

erla

nd

Co

lora

do

10

/12

/20

03

9/1

1/2

01

33

,62

21

21

La

nd

slid

eF

ore

sted

Gra

nit

icC

oe

eta

l.,

20

14

Pla

ins

1D

ate

of

fire

ign

itio

nw

as

use

da

sb

ase

lin

e,ra

ther

tha

nd

ate

of

fire

con

tain

men

t,b

eca

use

con

tain

men

td

ate

isn

ot

alw

ay

sco

nsi

sten

tly

iden

tifi

ed.

2F

ore

sted

veg

eta

tio

ncl

ass

esin

clu

de

tho

sere

po

rted

as

con

ifer

,sp

ruce

,fi

r,p

ine,

oa

k,

or

com

bin

ati

on

sth

ereo

f;n

on

-fo

rest

edin

clu

des

veg

eta

tio

nre

po

rted

as

cha

pa

rra

l,sc

rub

oa

k,

oa

kb

rush

,a

nd

oa

ksa

ge;

an

dm

ixed

incl

ud

esb

oth

fore

sted

an

dn

on

-fo

rest

edte

rra

ins.

3B

edro

ckcl

ass

esin

clu

de

vo

lca

nic

an

dm

eta

-vo

lca

nic

,m

eta

mo

rph

ic,

sed

imen

tary

,g

ran

itic

,a

nd

mix

ed.

Table

1.Continued.



operating a warning system for post-fire debris flowsin southern California with the National WeatherService, and documenting the effects of debris flowswithin recently burned areas throughout the westernUnited States. We cannot be certain that the databaseincludes all events from a given burned area,particularly those events reported in the literature.However, we are more confident that most debrisflows were documented during our own efforts.

APPROACH AND METHODS

To first identify the time period following firesduring which an increased likelihood of debris flowexists, we examine frequency distributions of our 75-entry database, providing information on the timingof debris flows relative to initiation mechanism,vegetative cover, and underlying bedrock. We alsoexamine the timing of debris flows within each of thefour primary rainfall regimes defined by Moody andMartin (2009) to account for the spatial variability ofprevailing rainfall conditions in different settings inthe western United States. We also consider therelative timing of fire ignitions, storm rainfall, anddebris-flow occurrence to identify the seasons andmonths during which most debris flows can beexpected in each rainfall regime.

To further characterize how the increased suscep-tibility to debris flows following wildfires may changeover time within areas with similar rainfall conditions,we calculate cumulative probabilities of a day withdebris flows for each of the primary rainfall regimesdefined by Moody and Martin (2009). We calculatedprobabilities as the ratio of the number of eventsdivided by the number of possible results (Helsel andHirsch, 2002) or, in this case, the number of daysduring which debris flows occurred within eachrainfall regime area divided by the number of days

leading up to, and including, the debris flows. Forexample, if debris flows are first triggered on day 4following fire ignition, this day is characterized bya 1/4, or 25 percent, probability of a day with debrisflows. On a cumulative basis, then, if after 100 daysthere have been a total of 10 days with debris flowswithin a given rainfall regime area, there will be a10/100, or 10 percent, probability of a day with debrisflows for that area. Cumulative probabilities arecalculated for each rainfall regime area for each dayin which debris flows were documented, starting withthe 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 thanone entry exists for a given day, the duplicate valuesare not included in the cumulative calculation becausewe are evaluating the number of days with debrisflows, and not the total number of debris flows.A best-fit analysis identifies the functions that bestcharacterize the cumulative probability functions foreach rainfall regime.

Note that the probabilities calculated here simplycharacterize the relative chances of a given dayexperiencing debris flows based on the numbers ofdays with debris flows within each rainfall regimearea. The probabilities do not describe the immediatepost-fire debris-flow susceptibility of individual drain-age basins based on a given set of physical and rainfallconditions, as addressed by Cannon et al. (2009).

RESULTS

Debris-Flow Timing

The first debris flows from a given fire aredocumented as occurring as soon as 4 days after fireignition (Table 1). The great majority (85 percent) ofdebris flows occurred within the first 12 monthsfollowing wildfire, with 71 percent occurring within

Table 2. Rainfall characteristics of rainfall regimes, as defined by Moody and Martin (2009).

Rainfall Regime Rainfall Seasonality Rainfall Intensity Classes

2-Year Recurrence, 30-MinuteDuration Rainfall Intensity (mm/hr)

Lower Upper

Pacific Winter maximum High .36 52Summer minimum Medium .20 36

Low .15 20Arizona Winter and summer wet Extreme .52 100

Spring dry High .36 52Fall moist Medium .20 36

Sub-Pacific Winter wet Low .10 20Spring moistSummer and fall dry

Plains Winter minimum Extreme .52 100Summer maximum High .36 52

Medium .19 36



the first 6 months (Figure 2a). The importance of thefirst year following fire is in keeping with the findingsof Santi and Morandi (2012) in southern California.Seven percent of the debris flows occurred between 1and 1.5 years after a fire. Following 18 months, thereis an 8-month period over which we did not identifyany fire ignition date–debris-flow response pairs. Thefinal 8 percent of events occurred between 2.2 and 10years (26 to 120 months on Figure 2a) after the fire.

Initiation Processes and Debris-Flow Timing

All but 1 of the 69 fire-debris flow pairs documen-ted in the first 18 months following a fire werereported to have been triggered by runoff-dominatedprocesses (Figure 2a and Table 1), indicating theprevalence of such processes within this timeframe.

Of the six events that occurred between 2.2 to 10 yearsfollowing wildfire, four were reported to haveinitiated through landslide processes, one throughrunoff-dominated processes, and the initiation pro-cess of one was unknown. Although the samplenumber is small, these proportions may suggest theimportance of infiltration-dominated processes in thetriggering of debris flows over these longer time-frames.

Vegetation Class, Initiation Process, and Debris-Flow Timing

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

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



classes exerts little influence on post-wildfire debris-flow susceptibility within this timeframe. Over longertimeframes (2.2 to 10 years), four landslide-initiatedevents occurred in forested terrains, one runoff-triggered event occurred in non-forested terrain, andthe initiation mechanism of one event is unknown.Again, although the sample size is small, theproportions indicate the possibility that over longtimeframes in forested terrains, post-fire debris-flowoccurrences may be dominated by landslide processes.

Bedrock Class and Debris-Flow Timing

Our data show that the type of bedrock underlyingburned areas may influence how long debris flows willcontinue to occur after fires (Figure 2b). Debris flowswere triggered in areas underlain by granites andmetamorphic rocks through the first 18 monthsfollowing ignition, while those underlain by sedimen-tary materials produced debris flows through 14

months and those underlain by volcanic materialsthrough 12 months, with most of these occurring inthe first 4 months. Each of the four bedrock classes isrepresented in the longer timeframes (2.2 to 10 years),but no clear effect of bedrock on debris-flowoccurrence is apparent.

Rainfall Regime, Fire Ignition, and Debris-Flow Timing

The timing of debris flows varied within the fourrainfall regimes during the first 18 months followingfire ignition (Figure 2c), seasonally (Figure 3) andintra-seasonally (Figure 4). Note that the probabilityof a day with debris flows (as shown in Figure 4)cannot be calculated until debris flows actually occur;probabilities before the first occurrence are notknown.

Within the area of the Pacific rainfall regime, firesstarted during a 7-month period from June through

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



December, and most debris flows were triggered inNovember through February (Figure 3). Debris flowswere documented as soon as 6 days after fire ignition(Table 1). Although most debris flows occurredwithin the first 6 months following ignition, theycontinued to be triggered through an 18-monthperiod, with some occurring during the second winterafter a fire (Figures 2c and 4). The debris flowsproduced during the second winter are the only eventsdocumented from the areas burned by the Motor (44),Gaviota (45), and Unnamed (41) fires. Debris flowswere also reported in July and August, presumably inresponse to summer thunderstorms, and the remain-der occurred during the winter months, when rain isreported to fall at a range of low to high intensities(Figure 3 and Table 2). The cumulative probability ofa day with debris flow (Figure 4) shows a gradualexponential decrease over the 18-month period ofrecord, but with periods of increasing probability thatreflect approximately week-long periods of increaseddebris-flow activity superimposed on the general

declining trend as sequential storms move throughthe area.

Within the area of the Arizona rainfall regime, firesignited in April, May, and June, and most debrisflows followed in July and August (Figure 3), theoccurrence of which corresponds with the monsoonseason, which is characterized by medium- toextreme-intensity rainfall (Table 2). Debris flows werenot reported until 28 days following ignition(Table 1), but all of the documented events occurredwithin 4 months of the fire, with the great majorityoccurring in the first 2 months (Figure 2c).

The cumulative probability of a day with debrisflow calculation for the Arizona rainfall regime areais distinct from the three other rainfall regimes(Figure 4). No debris-flow activity was documenteduntil 28 days after the onset of ignitions, resulting ina low initial probability of a day with debris flowsrelative to the other rainfall regimes. Once stormscapable of initiating debris flows affected susceptibleareas, however, the probability of a day with debris

Figure 4. Probability of a day with debris flow over time for the Pacific, Sub-Pacific, Arizona, and Plains rainfall regimes through the first18 months following ignition or less, depending on length of record. Day 0 is the first fire ignition in each rainfall regime area, and data startat the day of the first documented debris flow. Best-fit lines to the probability functions and their equations are shown as colored lines, withx as time and y as the probability of a day with debris flows.



flow increased rapidly over a little more thana month’s time. The two nearly vertical segments onFigure 4 for the Arizona rainfall regime are the resultof 1- to 2-week periods over which debris flows occuron a near-daily basis, and the two more horizontalsegments reflect slight decreases in the frequency ofdebris-flow events. The probability of a day withdebris flow starts to decline after more than 2 monthsof debris-flow activity, presumably as the monsoonseason wanes.

Within the area of the Sub-Pacific rainfall regime,fires were ignited over a 4-month period from Maythrough September (Figure 3), and the first debrisflow occurred 11 days following ignition (Table 1).Even though the area is typified by low-intensityrainfall (Table 2), which is not generally consideredsufficient to generate debris flows, debris flows weredocumented in this area. Debris flows were triggeredup to 13 months after ignition (Figure 2c), duringMarch and April, in July and September within daysof fire ignition as well as a year later, and in Augustand December (Figure 3). The three debris flows thatoccurred more than a year after fire ignition wereproduced from the Seeley (3), Inyo (42), and Mollie(43) fires. This was the fourth debris flow producedfrom the area burned by the Seeley fire and the firstfrom the Inyo and the Mollie fire areas. Within theSub-Pacific rainfall regime area, the probability ofa day with debris flow decreases as a power law, witha higher rate at the beginning that slows over timebut that extends into the year following ignition(Figure 4).

Within the area of the Plains rainfall regime, firesstarted during a 5-month period from June throughOctober, and debris flows were triggered from Junethrough September (Figure 3), a period typified bysummer rainstorms ranging in intensity from mediumto extreme (Table 2). Debris flows first occurred only4 days following fire ignition (Table 1), and activitycontinued for an additional 12 months (Figure 2c).The debris flow produced nearly a year after ignitionwas generated from the area burned by the Yellow-stone fire (48) and was the second debris flowreported from this area. The probability of a daywith debris flow decreases as a power law similar tothat of the Sub-Pacific rainfall regime area, thedifference being that debris flows start as soon as 4days after fire ignition, rather than after the 11 daysreported in the Sub-Pacific area.

DISCUSSION

The data examined here indicate two periods ofincreased debris flow susceptibility in burned drain-age basins—the first immediately following the fire

and lasting up through 18 months and the later,second period beginning after 2.2 years and extendingup to 10 years. Debris flows are considerably morefrequent during the initial period of susceptibility,occur with similar proportions in non-forested andforested terrains, and initiate primarily through theprocess of progressive entrainment of material erodedfrom hillslopes and channels by surface runoff. Thosefew debris flows that occurred during the secondperiod of susceptibility are most common in forestedterrains and for the most part occur throughmobilization of discrete landslide masses.

The time interval of 18 months to 2.2 yearsseparating the two periods of increased debris-flowsusceptibility in burned drainage basins is seen asrepresenting recovery from fire-induced conditions,favoring erosion by the process of progressiveentrainment of material. Wildfire can change thehydrologic response of a drainage basin by (1)removing soil-mantling vegetation and litter, (2)depositing ash, (3) altering of the physical propertiesof soil and rock, and (4) enhancing, generating, ordestroying water-repellent soils (Parise and Cannon,2012). In general, these changes reduce infiltration ofwater and make the burned surface soils more proneto entrainment by overland flow. In addition, Moodyand Ebel (2012) found that hyper-dry soil conditionsaffecting infiltration are an important factor re-sponsible for the extreme runoff events that frequent-ly occur during the first rainstorm after a wildfire.

Natural restoration of soil-mantling vegetation isa key factor in reducing the amount of erosion byoverland flow. Vegetation regrowth can occur quick-ly, regardless of burn severity and ecosystem, basedon a study of post-fire vegetation burn severity ateight large western U.S. wildfires (Lentile et al., 2007).Recovery plant cover was found to be predominantlygrasses in chaparral ecosystems, forbs in mixedconifer forests, and shrubs in boreal forests. Thecomposition of post-fire vegetation was influenced bythe presence of burned vegetation types capable of re-sprouting and the condition of ungerminated seeds inthe soil (Lentile et al., 2007). Re-vegetation by herbsand shrubs in burned areas under the influence ofMediterranean climatic conditions reduced erosion tonegligible levels within 2 years after the wildfire(Cerda and Doerr, 2005). This finding is consistentwith the debris flow susceptibility due to progressiveentrainment of material effectively ceasing by 18months.

Similarly, the second period of debris-flow suscep-tibility due to infiltration-triggered landslides startingat 2.2 years and persisting for up to 10 years inforested drainage basins is consistent with research onthe influence of tree root decay on shallow soil



movement. Two years coincides with the earliest timeperiod at which tree root decay would reach a levelthat significantly weakens the strength of the soilmass (Sidle, 2005; Sidle and Ochai, 2006), and Schulzet al. (2006) described both a loss of root strength anda decline in evapotranspiration rates as contributingto landslides generated from an area that had burned3 years prior. In addition, the senior author hasexamined a number of debris flows initiated byinfiltration-type landslides in both burned and un-burned forested watersheds in the southern SierraNevada (DeGraff, 1994). In failures from unburnedforested slopes, there are commonly numerous intact,flexible roots dangling within the exposed slide planeand its margins. These are remnants of what were liveroots placed in tension during initial landslidemovement and broken off by the moving mass. Incontrast, roots exposed in landslide-initiated debris-flow scars in burned areas were generally shearedclose to or at the slide plane and were noticeably dryand brittle, indicating a decline in strength attributableto fire-induced mortality and decay (DeGraff, 1997).

Understanding the timing of debris flows followingwildfire has implications for successful protection oflife and property. The current attention and researchdevoted to debris flows that occur in the few monthsto first year following a wildfire is well placed. Thethreat is widespread and largely independent of thetype of vegetative community burned. The immediacyof this threat calls for prompt implementation ofmitigation measures. The design and placement ofthese measures needs to be effective for debris flowsresulting from progressive entrainment of materialeroded from hillslopes and channels by surfacerunoff. The likelihood of short–recurrence intervalstorms with periods of high-intensity rainfall will needto be determined as part of this effort.

When wildfire burns a forested area, the mitigationof any imminent debris-flow threat must also becoupled with similarly prompt action to limit theimpact of later debris flows. The later debris-flowthreat in burned forest area provides time formitigation within the debris-flow source area. Coun-tering the decreasing root strength provided by fire-killed trees with the increasing strength of re-plantedtrees requires action within 1 to 2 years of the wildfire.Because the later threat of debris flows in foresteddrainage basins is associated with infiltration-typelandslides, there is sufficient time to prioritize re-forestation to basins in which the soil burn severity ishigh, tree mortality is great, and infiltration-typelandslide activity has resulted in debris-flow activityin the past. Some mitigation measures implemented toaddress immediate life and property concerns fordebris flows due to progressive bulking might be

maintained for later mitigation of infiltration-typedebris flows in forested basins.

In this study we found that the cumulativeprobability of a day with debris flows varies betweenthe Pacific and Arizona rainfall regimes, but theresponses of the Sub-Pacific and Plains rainfallregime areas were remarkably similar to each other,following similar power laws. The similarities anddifferences may reflect the timing of fire ignitions andrainfall and rainfall conditions in each area. In theSub-Pacific area, most rain is reported to fall in thespring and winter months and typically at relativelylow intensities, although most debris flows weredocumented as occurring in the summer and fall. Incontrast, within the Plains rainfall regime, mostrain is reported to fall in summer at intensities thatcan range from medium to extreme, most fires occurin the summer, and most debris flows occur in thesummer. The similarities in the timing of the fires anddebris flows but contrasts in the described rainfallconditions could indicate that the rainfall character-ization for the Sub-Pacific rainfall regime is missingthe important effect of summer and fall convectivestorms, which could produce rainfall at high in-tensities. The trend of the calculated probabilities ofa day with debris flows within the Arizona rainfallregime differs from that seen in other rainfall regimes,with an overall increase in probabilities that startswith the onset of the monsoon season and starts todecrease only after the season ends. This trend may bedue to extreme rainfall intensities throughout themonsoon season that trigger debris flows on a near-daily basis, but which may also hinder vegetative andsoil property recovery at the same time. Within thePacific rainfall regime, the prolonged and nearlylinear decline in probabilities of a day with debrisflows superimposed with approximately week-longperiods of increased probabilities can perhaps beattributed to the fact that most fires are nearlyimmediately followed by repeated bands of debris-flow triggering winter storms, but any vegetativerecovery that occurs in the wet winter months willslow during the sequential dry springs and summers.

Within the Pacific, Plains, and Sub-Pacific rainfallregimes, the asymptotic decline in the probability ofa day with debris flows after about 8 months may beconsidered to indicate a background susceptibility topost-fire debris flows in these areas, given noadditional fires in the area. Note, however, that oncea new fire burns in debris-flow–susceptible terrain ineither of these two rainfall regime areas, one wouldexpect to re-set and return to the high end of each ofthe curves. That these curves do not return tozero over time indicates that a slight susceptibilityto post-fire debris flows remains after the initial



18-month period for which post-fire debris flows weredocumented in this study.

The rainfall regimes used here to characterize post-fire debris-flow timing and susceptibility are broadspatial categories. Moody and Martin (2009) alsodefine secondary categories, but the spatial distribu-tion of our data was not sufficiently dense to warrantevaluation for each of these secondary classes. It isalso possible that some of the variability in stormconditions specific to the initiation of post-fire debrisflows is not completely captured by the divisions usedhere. As discussed above, characterization of the Sub-Pacific rainfall regime lacks the high-intensity sum-mer thunderstorm activity that produces debris flowsin this area. In addition, both the Pacific Northwestof the western United States and southern Californiaare included in the Pacific rainfall regime, while thepost-fire debris-flow response can be quite differentbetween these two areas. Although runoff-initiateddebris flows are the norm in southern California, wehave not observed evidence of this process in thePacific Northwest, and this difference could beattributed to differing prevailing rainfall conditions.And finally, although summers are considered to bedry within the Pacific rainfall regime, we have recordsof debris flows being triggered during this season.Definition of additional rainfall regimes may benecessary to capture this variability and bettercharacterize conditions that lead, specifically, topost-fire debris flows.

Localized weather will be a primary source ofvariability in the timing of debris flows and of someof the uncertainty associated with the findings of thisstudy. For example, fewer storms capable of trigger-ing debris flows during drought periods, or stormslacking periods of intense rainfall, will prolong thetime between burning and when debris flows occur.We see that within the Pacific and Sub-Pacific rainfallregime areas, some of the first known debris flowsfrom given burned areas were produced during thesecond rainy season to affect the area, indicating thatthe initial rainy season was perhaps lacking insufficient rain to trigger debris flows. Additional dataand more complex analyses will be necessary tocharacterize this variability.

Additionally, although we were not able to locaterecords of post-fire debris flows beyond 18 monthsand before 2.3 years following fire ignition, this doesnot mean that they would not occur during this timeperiod; the probability curves developed in this studydo not return to zero over time, indicating a persis-tence of post-fire debris-flow susceptibility beyond theinitial 18-month period. However, because 2 yearscoincides with the earliest time at which tree rootdecay would reach a level that significantly weakens

the strength of the soil mass, we would expect debrisflows that might occur during this period to initiatethrough the process of progressive entrainment ofmaterial from hillslopes and channels by runoff,rather than through landslide initiation.

Finally, the data used in this study do not reflectthe results of a spatially extensive, long-term, in-strumental monitoring of the hydrologic response ofburned areas, which could conceivably producea more complete representation of all debris-flowactivity in a given burned area. In addition, thecalculated changes in the probability of a day withdebris flows do not reflect changes in the expectedvolumes of potential debris flows, an importantelement in assessing hazards and risk. We thussuggest that this analysis presents a first look atchanges in post-fire debris-flow susceptibility that canbe refined and improved with additional data.

SUMMARY AND CONCLUSIONS

In summary, both forested and non-forested land-scapes are likely to experience an immediate increasedsusceptibility to debris-flow occurrence in drainagebasins recently burned by wildfires. This period ofgreater debris flow susceptibility can last up to 18months but will vary depending on underlyingbedrock materials and prevailing climatic conditionsor rainfall regimes. The great majority of debris flowsthat occur during this period initiate through pro-gressive entrainment of material eroded from hill-slopes and channels by surface runoff. The data inthis study indicate that debris flows will occur inareas underlain by granites and metamorphic rocksthrough the first 18 months following wildfires, whilethose underlain by sedimentary materials can producedebris flows for up to 15 months, and those underlainby volcanic materials for up to 12 months.

Within the area included in the Pacific rainfallregime, debris flows can be expected just 6 days afterignition, and although the great majority of eventswill occur in the first 6 months, they can continue tobe triggered through an 18-month period, with someoccurring during the second winter after a fire.Although most debris flows will be triggered duringthe months of November through February inresponse to winter rainfall, some can occur in Julyand August, presumably in response to summerthunderstorms. The probability of a day with debrisflows in the area of the Pacific rainfall regime declinesas a gradual exponential function with superimposedapproximately week-long periods of increasing prob-abilities as sequential winter storms move through thearea.



Within the Arizona rainfall regime, debris flowswere not reported until 28 days following ignition, butall events can be expected within 4 months of the fire,and the great majority can be expected in the first 3months, all in response to monsoonal storms. Julyand August will host the majority of the debris flows.The probability of a day with debris flows increasesrapidly once storms capable of triggering debris flowsaffect susceptible areas, with debris flows possible ona near-daily basis during this time.

Within the Sub-Pacific rainfall regime, the firstdebris flow can ensue as soon as 11 days followingignition, and most debris flows can be expected withinthe first 13 months. Debris flows can be triggered inMarch, April, July, August, and September inresponse to spring, summer, and fall rainstorms.The probability of a day with debris flow decreases asa power law, with a higher rate at the beginning thatslows over time but that extends into the yearfollowing ignition.

In the area of the Plains rainfall regime, debrisflows can first occur 4 days following fire ignition,and activity can continue for an additional 12months, into the following summer and fall rainyseasons. Debris flows can be expected from Junethrough September. The probability of a day withdebris flow decreases as a power law very similar tothat of the Sub-Pacific rainfall regime area, thedifference being that debris flows start as soon as 4days following ignition, rather than after the 11 daysnoted in the Sub-Pacific area.

The decreasing likelihood of debris flows over timedescribed here is thought to reflect the gradualrestoration of hydrologic function as vegetative coverand soil infiltration rates return to pre-fire conditions,the temporal sequence of fires and storms, andprevailing storm-rainfall conditions in a given rainfallregime.

Forested landscapes are recognized as havinga second, later period of increased debris flowsusceptibility in burned drainage basins. Althoughprevious workers have reported that this period ofsusceptibility typically does not begin until 4 to 5 yearsfollowing the wildfire (Meyer et al., 2001; Wondzelland King, 2003), our database included events thatoccurred 2.3 to 10 years following fire ignition, with 2years coinciding with the earliest time at which treeroot decay would reach a level that significantlyweakens the strength of the soil mass. Debris flowsthat occur during these longer time frames mostfrequently are attributable to infiltration-triggeredlandslides, which mobilize into debris flow.

The short time period between fire and debris-flowresponse within the first 1.5 years after ignitiondemonstrates the necessity of a rapid response by

land managers and emergency response agencies tomitigate potential hazards from potential runoff-triggered debris flows from recently burned areas inthe western United States. The presence of a secondperiod of susceptibility to landslide-triggered debrisflows between 2.3 and 10 years after a fire in forestedterrains indicates the need to consider additionallong-term mitigation solutions specific to this initia-tion process.

ACKNOWLEDGMENTS

The authors wish to express their appreciation toRichard Giraud and other researchers who providedeither unpublished data or their time for interviewsduring compilation of our database. The insightfuland detailed comments provided by an unnamedreviewer, Jeffrey Coe, Paul Santi, Dennis Staley, andJason 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 stormwith an extraordinary response: Mapping the debris-flowresponse to the December 25, 2003 storm on the 2003 Oldand Grand Prix fire areas in southern California: GeologicalSociety 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 Manage-ment of Mediterranean-Type Ecosystems: U.S. Department ofAgriculture Forest Service, Pacific Southwest Forest andRange Experimental Station, General Technical ReportPSW-58, pp. 420–422.

CANNON, S. H., 1997, Evaluation of the Potential for Debris andHyperconcentrated Flows in Capulin Canyon as a Result ofthe 1996 Dome Fire, Bandelier National Monument, NewMexico: U.S. Geological Survey Open-file Report 97-136,20 p.

CANNON, S. H., 2001, Debris-flow generation from recently burnedwatersheds: Environmental Engineering Geoscience Vol. 7,pp. 321–341.

CANNON, S. H.; BIGIO, E. R.; AND MINE, E., 2001, A process forfire-related debris-flow initiation, Cerro Grande Fire, NewMexico: 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 post-fire debris flow emergency-response planning: NaturalHazards: 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 postwildfire debris-flow hazards. In Sassa, K. and Canuti, P.(Editors), Landslides: Disaster Risk Reduction: Springer-Verlag, 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 ofbasins burned by the 2002 Coal Seam and Missionary Ridgefires, Colorado. In Boyer, D. D.; Santi, P. M.; and Rogers,W. P. (Editors), Engineering Geology in Colorado–Contribu-tions, Trends, and Case Histories: AEG Special Publication15, Colorado Geological Survey Special Publication 55, 31 p.



CANNON, S. H.; GARTNER, J. E.; RUPERT, M. G.; MICHAEL, J. A.;REA, A. H.; AND PARRETT, C., 2010, Predicting the probabilityand volume of post-wildfire debris flows in the intermountainwest, 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, EmergencyAssessment of Postfire Debris-Flow Hazards for the 2009Station 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 flowsfrom recently burned areas in southwestern Colorado andsouthern California: Geomorphology Vol. 96, pp. 250–269.

CANNON, S. H.; POWERS, P. S.; AND SAVAGE, W. Z., 1998, Fire-related debris flows on Storm King Mountain, GlenwoodSprings, Colorado, USA: Environmental Geology Vol. 35,pp. 210–218.

CANNON, S. H. AND RENEAU, S. L., 2000, Conditions for generationof 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 vegetationrecovery on soil hydrology and erodibility following fire: An11-year investigation: International Journal Wildland Fire,Vol. 14, No. 4, pp. 423–437.

CLEVELAND, G. B., 1973, Fire + rain 5 mudflows: CaliforniaGeology, 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 intodebris-flow hazards from an extraordinary event in theColorado Front Range: GSA Today, Geological SocietyAmerica Vol. 24, pp. 4–10.

CONEDERA, M.; PETER, L.; MARXER, P.; FORSTER, F.; RICKENMANN,D.; AND RE, L., 2003, Consequences of forest fires on thehydrogeological response of mountain catchments: A casestudy of the Riale Buffaga, Ticino, Switzerland: EarthSurface Processes Landforms, Vol. 28, pp. 117–129.

DEGRAFF, J. V., 1994, The geomorphology of some debris flows inthe southern Sierra Nevada, California: Geomorphology,Vol. 10, pp. 231–252.

DEGRAFF, J. V., 1997, Geologic Investigation of the Pilot RidgeDebris Flow, Groveland Ranger District, Stanislaus NationalForest: 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 BurnedArea Emergency Response (BAER) process. In Schaefer,V. R.; Schuster, R. L.; and Turner, A. (Editors), ConferencePresentations, 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 remarkableoccurrence of large rainfall-induced debris flows at twodifferent 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 processesin the San Gabriel Mountains, California: ContributionsGeology Vol. 7, pp. 43–65.

DWYER, E.; PINNOCK, S.; GREGOIRE, J. M.; AND PEREIRA, J. M. C.,2000, Global spatial and temporal distribution of vegetationfire as determined from satellite observations: InternationalJournal Remote Sensing Vol. 21, Nos. 6 and 7,pp. 1289–1302.

EATON, E. C., 1936, Flood and erosion control problems and theirsolution.: Proceedings American Society Civil EngineersVol. 62, pp. 1302–1362.

FLORSHEIM, J. L.; KELLER, E. A.; AND BEST, D. W., 1991, Fluvial

sediment transport in response to moderate storm flowsfollowing 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 Pinon-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.

GARCIA-RUIZ, J. M.; ARNAEZ, J.; GOMEZ-VILLAR, A.; ORTIGOSA, L.;

AND LANA-RENAULT, N., 2013, Fire-related debris flowsin 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 Open-

File 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 ofdebris 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 situmeasurements of post-fire debris flows in southern Califor-

nia: Comparisons of the timing and magnitude of 24 debris-flow 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 largewildfires across the western United States: Fire Ecology Vol. 3,

No. 1, pp. 91–108.

LLORET, F. AND MARI, G., 2001, A comparison of the medievaland 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.



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 sedimentationevents 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 conditionsprovide new insights into the cause of extreme floods afterwildfire: Catena Vol. 93, pp. 58–63.

MOODY, J. A. AND MARTIN, D. A., 2009, Synthesis of sedimentyields after wildland fire in different rainfall regimes in thewestern United States: International Journal Wildland FireVol. 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 topost-wildfire runoff and erosion processes: Earth-ScienceReviews Vol. 122, pp. 10–37.

MORTON, D. M., 1989, Distribution and frequency of storm-generated soil slips on burned and unburned slopes, SanTimoteo Badlands, Southern California. In Sadler, P. M. andMorton, D. M. (Editors), Publications Inland GeologicalSociety Vol. 2, pp. 279–284.

NASA, 2014, Global Maps-Fire. Earth Observatory, NationalAeronautic and Space Administrations: Electronic document,available at http://earthobservatory.nasa.gov/GlobalMaps/view.php?d15MOD14A1_M_FIRE

NIFC, 2011, Historically Significant Wildland Fires. NationalInteragency Fire Center, Boise, Idaho: Electronic document,available at http://www.nifc.gov/fireInfo/fireInfo_stats_hist-SigFires.html

NYMAN, P.; SHERIDAN, G. J.; SMITH, H. G.; AND LANE, P. N. J.,2011, Evidence of debris flow occurrence after wildfire inupland catchments of south-east Australia: GeomorphologyVol. 125, pp. 383–401.

PARISE, M. AND CANNON, S. H., 2012, Wildfire impacts on theprocesses that generate debris flows in burned watersheds:Natural Hazards, Vol. 61, pp. 217–227.

PARRETT, C.,1987, Fire-Related Debris Flows in the Beaver CreekDrainage, Lewis and Clark County, Montana: U.S. GeologicalSurvey Water-Supply Paper 2330, pp. 57–67.

PARRETT, C.; CANNON, S. H.; AND PIERCE, K. L., 2003, Wildfire-Related Floods and Debris Flows in Montana in 2001: U.S.Geological Survey Water-Resources Investigations Report03-4319.

PEARTHREE, P. A., 2004, Fire and sediment deposition: ArizonaGeology, Vol. 34, pp. 1–2.

SANTA BARBARA COUNTY PUBLIC WORKS DEPARTMENT, SantaBarbara County Flood Control and Water ConservationDistrict, 2005, Santa Barbara County Public Works Department2004–2005 Winter Report: Electronic document, available athttp://www.worldcat.org/title/santa-barbara-county-public-works-department-2004-05-winter-report/oclc/213938602

SANTI, P. M.; DEWOLFE, V. G.; HIGGINS, J. D.; CANNON, S. H.; AND

GARTNER, J. E., 2008, Sources of debris flow material inburned 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-flowvolumes 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 2006-1341, 29 p.

SCOTT, K. M., 1971, Erosion and Sediment Yields in the Transverse

Ranges, Southern California: U.S. Geological Survey Pro-fessional 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 ondebris 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 Nevadaregion, California. In Putirka K. (Editor), Geologic Excur-

sions from Fresno, California, and the Central Valley: A Tour

of California’s Iconic Geology: Geological Society of AmericaField Guide 32, pp. 99–128. doi 10.1130/2013.0032(06)

WEISBERG, P. J. AND SWANSON, F. J., 2003, Regional synchroneityin 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 debrisflows in southern California. In Costa, J. E. and Wieczorek,G. F. (Editors), Debris Flows/Avalanches, Process, Recogni-

tion, and Mitigation: Geological Society America Reviews

Engineering Geology, No. 7, pp. 105–114.

WOHL, E. E. AND PEARTHREE, P. P., 1991, Debris flows asgeomorphic agents in the Huachuca Mountains of south-eastern Arizona: Geomorphology, Vol. 4, pp. 273–292.

WONDZELL, S. M. AND KING, J. G., 2003, Postfire erosionalprocesses in the Pacific Northwest and rocky Mountainregions: 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 ofArizona, Tucson, p. 235.



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, MO65409-0660

ADNAN M. AQEEL

Taiba University, P.O. Box 30002 Madinah, Saudi Arabia 41477email: [email protected]

NEIL ANDERSON

Missouri University of Science and Technology, 1006 Kingshighway, Rolla, MO,65409-0660, email: [email protected]

Key Terms: Engineering Geology, Terrestrial Lidar,Discontinuities, Orientation, Ground Penetrating Ra-dar

ABSTRACT

Vertical or sub-vertical discontinuities striking par-allel to rock cuts are dangerous because toppling andspontaneous raveling failures can initiate from thesesurfaces, creating hazards below. At the same time, thesurfaces of these discontinuities are often concealedbecause they do not “daylight” into the rock, and anytrace of the discontinuity that might be seen at the topof the rock cut is obscured by overburden. These hiddendiscontinuities can often be detected by groundpenetrating radar (GPR). Our new method uses GPRin conjunction with terrestrial LIDAR (light detectionand ranging) to accurately measure the orientation ofthese hidden discontinuities. The method presented inthis article establishes three control points on thesurface of the rock cut. At each control point the globalcoordinates are remotely measured using LIDAR. GPRsoundings at each control point are used to measure“the perpendicular horizontal distance” (depth) fromeach control point on the rock cut face to anydiscontinuities hidden behind the rock cut face. Thetrue perpendicular distance is added to the GPRcoordinates at each control point to form three newcontrol points on the surface of each of the hidden

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

INTRODUCTION

Rock cut stability in most rock hard massesdepends on the nature and orientation of thediscontinuities (joints) within the rock mass. Whenthey happen, failures typically start and propagatefrom and along the discontinuities. Failures can resultin property damage, lane or highway closure, andeven serious injury or death.

If the orientation of all discontinuities within therock mass behind the cut can be measured, a de-terministic analysis can be undertaken to assess thelikelihood of failure. This can be as simple asa kinematic analysis on a stereonet, as a limitingequilibrium analysis, or as complex as a numericalmodel that supports analysis of discontinuous rock.

Discontinuities can be measured on rock cuts usingthe traditional manual scanline method (Otoo et al.,2011), which is very common, inexpensive, and easy touse but is time consuming and risky when measure-ments are carried out at the base of potential failureslopes. Alternatively, advanced in situ geometrical datacollection methods are used, such as photogrammetricmethods, total station surveying methods, or, morerecently, light detection and ranging (LIDAR) methods(Post, 2001; Slob and Hack, 2004; Donovan et al., 2005;Haneberg, 2008; Sturzeinegger and Stead, 2009; andOtoo et al., 2011). Ground-based LIDAR scanners,which are sometimes called Terrestrial Laser Scanners(TLS), are geodetic instruments that have become very1Corresponding author email: [email protected].


popular for engineering and geology surveys in recentyears (Gonzalez-Jorge et al., 2011; Otoo et al., 2011).These can provide accurate point cloud data of thescanned slope within minutes, and the geometry of rockdiscontinuities can be extracted in an automated andobjective way (Pernito, 2008).

Only discontinuities that can be seen in the rock cutcan possibly be measured in this manner. Vertical orsub-vertical discontinuities that strike parallel to therock cut often are not exposed in the rock cut becausethey 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 overburdenand vegetation. In addition, if vertical borings aredone from the top of the slope the vertical or sub-vertical discontinuities will not be encountered.

Horizontal drilling can always be used to identifyhidden discontinuities, but this drilling is a time-consuming, expensive process that in some cases willrequire one or more lanes of the road to be closed. Inaddition to obtaining orientation measurements,costly oriented core drilling would also need to bedone. Ground penetrating radar (GPR) has beenshown to have the ability to both detect the hiddendiscontinuities and measure their angular trend(Maerz and Kim, 2000; Soel et al., 2001; Porsani etal., 2006; Pernito, 2008; and Torres, 2008).

GPR is an active geophysical method for non-destructive subsurface imaging based on the propaga-tion of electromagnetic (EM) waves in the subsurface(Reynolds, 1997; Conyers, 2004; Daniels, 2004; Ottoand Sass, 2006; and Sass, 2007). Once the EM wavescontact an interface plane whose electrical propertiesdiffer from those electrical properties of the surround-ing subsurface materials (such as a discontinuityplane), a portion of this energy is reflected back toand recorded in the GPR system at the ground surfaceor the rock face.

The strength of the GPR pulses reflecting froma rock discontinuity depends mainly on the apertureof the discontinuity and the infilling materials, bothof which control the reflection coefficient (Gregoire,2001). This will give a distinctive linear interface,reflector, or event, with a high amplitude compared tobackground reflection, in the radiogram image. Thisdistinctive reflection signature can be used as criteriawith which to objectively delineate the discontinuitiesin rock masses (Pernito, 2008).

METHODOLOGY

Overview

The key focus in this research is to appropriatelycombine both LIDAR and GPR data for the purpose

of measuring the geometry of concealed sub-verticaldiscontinuities. Terrestrial LIDAR can accuratelyand easily measure the position of any point visibleto the LIDAR scan, while GPR can measure thedistance (perpendicular horizontal depth) betweenany point on an exposed surface and the projection ofthat point onto the plane of a detectable hidden sub-vertical discontinuity. Consequently, if only threepoints that are co-planar, but not co-linear, on theexposed surface are projected onto the plane of thehidden discontinuity, the true orientation of thishidden discontinuity can be determined.

Measurement Protocols

Briefly, the procedure to measure hidden disconti-nuities starts with identification of a nearly flat rockface. This does not have to be a discontinuity face,but a near-vertical surface is optimum for identifyingand measuring hidden sub-vertical discontinuities.Then, three none colinear control points are markedon the rock face such that their positions in space interms of coordinate triplets (x, y, z) can be measuredusing LIDAR (Figure 1). These control points couldalso be measured using other survey techniques.

At a minimum, one GPR sounding must be done ateach control point; however, obtaining multiple GPRtraverses will be more effective in isolating thegeometry of the concealed structures behind the rockslope face (Figure 1). Once GPR data are acquired,processing these data is required to enhance the qualityof the data and to improve the clarity of the resultingradiogram images. After that process is complete, theapparent horizontal distance (depth) to the hiddendiscontinuities can be measured on the radiograms.

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

Maerz, Aqeel, and Anderson


As a final step in GPR processing data, a GPR datamigration method is used to calculate the truehorizontal distance to the hidden structure in a di-rection perpendicular to the rock face. By addingthese perpendicular distances to each of the co-ordinate triples (x, y, z; control points) for each of theidentified hidden structures (transpose the coordinatetriplets to the hidden structure), the orientation of thehidden structures using the transposed coordinatetriples can be calculated.

Field Trials

The measurements for this research were conductedon small rocks cut along an outer road paralleling theInterstate 44 highway about just north of Rolla,Missouri, USA. The measurements were made in theRubidoux Sandstone, hard, competent sandstonewith near-horizontal bedding and two sub-verticaljoint sets that are approximately mutually orthogonalto each other and to the bedding orientation. Therock cut itself consist of abundant sub-verticalfracture surfaces stained red, with occasional white-colored, blast-induced fractures. For this research, thetesting was conducted only on sub-vertical disconti-nuity surfaces exposed in the rock cut. The rock cut(the study area) was divided into five stations. Eachstation was treated as an individual station on whichboth LIDAR and GPR measurements were con-ducted.

LIDAR MEASUREMENT

Overview

The idea of mapping discontinuities on rock massfaces using remote sensing techniques is conventional.Terrestrial stereo-photogrammetric techniques havebeen recently used through improvements in imagingand digital processing data techniques. These andLIDAR measurements can be used for many applica-tions in different disciplines, especially in the fields ofrock engineering, rock mechanics, and landslides(Roberts and Propat, 2000; Fasching et al., 2001;Gaich et al., 2006; Kemeny et al., 2006; Whitworth etal., 2006; Dunning et al., 2009; Sturzeinegger andStead, 2009; Abellan et al., 2010; Asahina and Taylor,2011; Garcıa-Selles et al., 2011; and Gonzalez-Jorgeet al., 2011).

Slob and Hack (2004) successfully used the semi-automated and automated approaches of three-di-mensional (3D) laser scanning survey to map a rockslope face composed of carboniferous meta-siltstoneand slate with well-developed discontinuity sets alonga secondary road in Catalonia, Spain. They found

that even though the two approaches can producehigh-resolution data suitable for mapping disconti-nuities and any other purpose in rock engineering, thefull-automatic method is capable of capturing moredata than are required for further statistical and/ormodeling analysis.

Moreover, another one of the most recent applica-tions of the LIDAR is in the art of forecastingpossible rock falls and rock mass slides, which aremainly controlled by the presence of discontinuitiesand their orientations and geometry (Abellan et al.,2010).

Alba and Scaioni (2010) have described how toextract change and rock mass deformation detectionbased on a LIDAR survey for the same rock face attwo different times or periods. Their analysis wasconducted by taking into account the multi-temporalpoint cloud georeferences and is built on three mainsteps: (1) vegetation filtering based on near infraredimagery; (2) detection of major changes such as lossmaterials; and (3) deformation analysis and testing.Moreover, the prediction of slope failures by moni-toring and understanding of ongoing, even millimeter,deformation, which is mainly controlled by thegeometrical and orientation characteristics of discon-tinuities, has been conducted by utilizing LIDARtechnology (Abellan et al., 2010).

It is not difficult to carry out a 3D laser scansurvey; however, it is quite a challenge to convert theLIDAR data to useful information that can directlybe used for the purpose of slope stability analysis orany other purpose in the rock engineering practice.Different methods or approaches have been used tohandle this issue, such as semi-automated or auto-mated methods, in some cases using a geologicalcompass 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 ofmanual compass measurements.

However, traditional and LIDAR measurementsare limited to exposed discontinuities on the rockslope, which excludes detection of hidden disconti-nuities that may have a significant effect on the rockslope stability analysis. For these reasons, it isdesirable to employ a geophysical tool that will beable to detect and delineate or map hidden disconti-nuities or fractures at shallow depths inside the slope.

Lidar Scanning

Lidar scans were taken using a Leica ScanStation IIscanner. The ScanStationII scanner is a time-of-flight,static, tripod-mounted system that deploys front andtop windows with an oscillating mirror design tocover the full field of view of 360u horizontally and

Measuring Orientations of Sub-Vertical Sandstone Discontinuities


270u vertically. It has a detection range of 90 m at 90percent reflectivity. It can scan 50,000 points persecond with an accuracy on the order of 3 t0 5 mm. Alaptop connected by Ethernet cable records data onrange, angles, and degree of reflectivity of returninglaser signals. The scanning system also collects opticalimages that are registered to the LIDAR dataautomatically using a built-in digital camera.

The measurement process is initiated by selectinga rock cut face to image. In preparation for imaging,non-colinear control points are manually markedusing chalk on near-planar face at each station. Thesepoints need to be enumerated and recorded either onthe face or in the field notes (Figure 1). For theLIDAR measurement calibration process, the strikeand dip angle of the rock face (or any planar verticalor sub-vertical orientation control element in thescan, such as a clipboard) is measured using a compassand the result recorded.

The LIDAR scanner is then set up across from therock face. No survey control is needed, but thescanner must be correctly leveled. Scanning at a lowresolution (5-mm spacing) is more than adequate formost purposes. Only a single scan is required.

Lidar Data Processing

Processing of the LIDAR data is simple. A LIDARviewer is used to view the resulting point cloud. Thecoordinate triplets (x, y, z) for each of the controlpoints are identified by mouse click on the pointcloud and their values recorded. Coordinate triplesmust also be identified and recorded for theorientation control element. These can be anyarbitrary non-colinear points on the control elementsurface. In some cases, as shown in Figure 1, the rockcut face itself is used as an orientation controlelement, then the coordinate triples of the controlpoints are used. No further processing of LIDARdata is required.

GPR MEASUREMENT

Overview

GPR has previously been used to identify andmeasure the horizontal depth to discontinuities ina rock mass. Maerz and Kim (2000) conducted a fieldinvestigation in a sandstone rock formation inMissouri using GPR with 400-MHz antenna for thepurpose of identifying the hidden vertical and/or sub-vertical discontinuities in a rock cut. The resultsshowed the ability of the GPR to detect and depictthe vertical discontinuities up to 2.5 m deep horizon-tally into the rock mass under dry conditions.

However, the strength of the reflected GPR pulsesincreases as the difference between relative dielectricpermittivities increases (Aqeel, 2012). Open disconti-nuities that are filled with water and/or clay areclearly more visible in the GPR radiogram than arethose discontinuities that are closed or that have nofilling material (Toshioka et al., 1995; Pernito, 2008;and Otoo et al., 2011). However, the GPR penetra-tion depth with appropriate resolution is less and thebackground noise is greater when water occurs indiscontinuity apertures, causing, to some extent, moredifficulty in identifying the hidden discontinuities.

GPR Soundings

Equipment

In this research and for rock slope engineeringpurposes, the main point behind using GPR is todetect and map discontinuities in rock cuts withindepths up to 4 m, since this is usually the range ofdepths in which discontinuities can play a major rolein causing rock failures. The deeper (rather thanshallower) discontinuities are not as likely to contrib-ute to slope failure as they are less likely to contributeto toppling-type failures.

The impulse GPR equipment used was manufac-tured by Geophysical Survey Systems, Inc. (GSSI)and utilized a 400-MHz monostatic antenna. Adistance measuring wheel was attached to the antennato acquire the GPR data in distance mode in order toproduce 3D images for the detected discontinuitiesusing both Radar Data Analyzer software package(RADANTM), which is manufactured by Geophys-ical Survey Systems, Inc., and ArcGIS 9.3, producedby Environmental Systems Research Institute, Inc.

The 400-MHz antenna was selected as a compro-mise between depth of penetration and minimumresolution (minimum aperture of discontinuity thatcan be detected). Higher frequency antennas willdetect smaller fractures at shallower depths, whilelower frequency antenna will penetrate to deeperdepths but detect only fractures with greater aper-tures. Using the 400-MHz antenna revealed measure-able discontinuities at depths of at least 2.5 m. Thetheoretical minimum resolution (minimum resolvablethickness) of a layer is considered to be one quarter ofthe antenna wavelength (Gonzalez-Jorge et al., 2011).

Measurement of GPR Pulse Velocity

The GPR pulse velocity can be measured either inthe field or in the lab; however, we preferred toestimate the velocity in the lab to ensure repeatabilityand accuracy. Therefore, a large rock sample was



collected from the study area to measure the velocityof GPR pulses that transmit through the sandstone ofthe study area. The rock sample was then trimmedinto two rectangular blocks. To simulate naturalconditions, the two rock blocks were positioned ontop of each other with a separation of 1.80 cmbetween them to act as a discontinuity plane(Figure 2). The top block had a thickness of 10.60cm, which means that the true perpendicular depthfrom the surface of the top block to that createddiscontinuity had to be 10.60 cm in the resultingradiograms of this test (Figure 3). Subsequently, thevelocity of GPR pulses was measured in thelaboratory utilizing a 1,500-MHz GPR monostaticantenna on the two rock blocks. The velocity of GPRsignals is fixed for each earth material regardless ofthe frequency of the utilized GPR antenna. Thereason a 1,500-MHz antenna was used is that it ismore compatible with the small sample in terms of thesheer size of the 400-MHz antenna and its penetrationdepth. The resulting radiogram image showed thatthe two-way travel time (t) of the GPR pulses was 2nanoseconds (ns), as illustrated in Figure 3. Since themeasured perpendicular (vertical) depth is 10.60 cm,the velocity (V) can be calculated as follows:

V~2d=t ð1Þ

where t 5 the two-way travel time 5 2 ns 5 t; d 5 theperpendicular (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 pulsevelocity is estimated, relative dielectric constant (e)can be accurately calculated as follows:

V~c=(e)1=2 ð2Þ

where e 5 the relative dielectric constant (e), which isdimensionless; and c 5 the speed of light in meters pernanosecond (0.3 m/ns).

Accordingly, the relative dielectric constant (e) ofthe sandstone is 8. As a result, the relative dielectricconstant in the GPR system during data acquisitioncan be set to 8. To test our results for this part of theresearch, the vertical axis that corresponds to thetravel time or the penetration depth of GPR signals ofthe radiogram was set in a depth mode in the monitorof the GPR system; the radiogram image showed thatthe perpendicular (vertical) depth to the discontinuitywas 10.60 cm, which reflects the accuracy and theprecision of our lab work (Figure 3).

Since the 400-MHz GPR antenna has a pulse periodof 2.5 ns, the pulse wave length in this research is theproduct of 0.106 m/ns and 2.5 ns, which is 0.265 m.Consequently, the minimum resolvable aperture of

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

GPR Field Measurement Methodology

Several parallel horizontal GPR profiles wereacquired on each rock cut face (station) (Figure 1).These profiles were conducted in a horizontal di-rection. Two of these profiles were located to passthrough the three fixed co-planar control pointsearlier identified for the LIDAR scanning. Thesetwo GPR profiles are called the index GPR profiles.Additional profiles were measured simply to increasethe definition of the hidden discontinuities. Thespacing between consecutive GPR profiles wasbetween 10 and 20 cm. The spacing between and thetrend of the GPR profiles were recorded in the field.

A 400-MHz GPR monostatic antenna was movedalong these profiles on the rock cut faces. Threepersons were used for the GPR data acquisition(Figure 4). The locations of the control points oneach rock cut were identified on the radiogram imagesutilizing the option of creating tick marks in the GPRdisplay. Next, the collected GPR data were processed.

GPR Data Processing

Visual Enhancement

GPR data processing was done for all of theacquired GPR data to enhance the visual quality of

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



the resulting two-dimensional (2D) and 3D radio-gram images. The GPR product radiogram is notonly an image but is also the recorded response of thesubsurface materials and structures to the GPR EMwaves. In most cases, radiograms are processed usingspecialized software to enhance their visual quality,and, thus, interpretation of the radiograms becomeseasier and more reliable. In this research, the RadarData Analyzer software package RADANTM wasused for this purpose. After applying a zero correc-tion (position or time-offset correction) for all GPRdata, high-pass and low-pass filtering were used to

remove instrument noise from data to improve thequality of the data (Reynolds, 1997; Annan, 2009;and Cassidy, 2009a). Infinite impulse responsefiltering was used. A 100-MHz vertical pass filterwas used to remove potential flat-laying ringingsystem noise, while an 800-MHz vertical low passfilter was used to remove high frequency.

Generally, filtering the GPR data processingmethod is sufficient to locate subsurface features inmany GPR applications (Reynolds, 1997; Annan,2009). However, the main goal in this research was todetect concealed discontinuities, map them, and

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 distancemeasured manually in the lab; e 5 8. The horizontal axis is the distance along the rock surface in 1025 m.



measure their orientations, up to of 4 m deep.Consequently, additional GPR data processing meth-ods were applied. Color transformation of value 17was applied to hide what little noise may remain inthe GPR radiograms. Moreover, automated rangeand display gain and deconvolution were applied tomaximize resolution and improve the visual quality.The main purpose behind using the deconvolutionmethod is to maximize bandwidth and reduce GPRpulse dispersion in order to ultimately maximizeresolution. In deconvolution, an operator length of31, 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 processedacquired GPR data from the study area.

In the study areas, 13 hidden sub-vertical disconti-nuities were manually identified on the radiograms atfive stations. Figures 5 through 7 display the resultingprocessed radiograms of the two index GPR profiles(of the first three stations) showing the detecteddiscontinuities at each rock cut (station). Three

hidden sub-vertical discontinuities were identified ateach of both stations 1 and 3, two hidden sub-verticaldiscontinuities were identified at station 2, fourhidden sub-vertical discontinuities were identified atstation 5, and only one hidden sub-vertical disconti-nuity was identified at station 4.

Apparent Horizontal Depths

Once the vertical axis of the processed GPRradiogram image is set into depth mode instead oftime mode on the GPR monitor, apparent depth(d) to any detected target in the radiogram image canbe estimated manually from this radiogram image.Apparent depth can be defined here as the distancemeasured manually from a specific point (controlpoint) on the rock surface to a discontinuity detectedon the radiogram image in a direction perpendicularto the plane of the discontinuity.

A migration process, as illustrated in Figure 8, isused to determine the true horizontal depth (d9)

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 toconduct such work.



perpendicular to the measurement plane. The appar-ent horizontal perpendicular depths (d), which weremeasured from the three control points to eachdetected hidden sub-vertical discontinuity the 2Dradiograms of each of the five stations (Figures 5 to7), are listed in Table 1.

Migration and True Horizontal Depths

Migration is a mathematical process and common-ly the final step in GPR data processing used torelocate and reconstruct detected targets to their truelocations and, thus, to their true geometry. Migrationwas applied in this research to determine the positionof the control points projected onto the hiddendiscontinuity to reconstruct the true geometry of thedetected discontinuities and, thus, the true perpen-dicular horizontal depths. Migration can be done

using specialized software or manually. GPR datamigration using software is usually successful inrelatively homogeneous environments, such as pave-ments and glacial environments. However, it tends tobe less successful for complex and heterogeneous sites(Cassidy, 2009b). Therefore, manual migration wasused in this research to avoid uncertainty resultingfrom variability in the inherent properties of thehidden discontinuities. Manual migration is explainedin many geophysical references (Kleyn, 1983; Jenyonand Fitch, 1985; Conyers, 2004; Lines and Newrick,2004; and Aqeel, 2012).

The GPR pulses along each profile can be imaginedas a perpendicular horizontal plane penetrating therock cut face and intersecting the planes of thedetected discontinuities. As a result, linear features(reflectors, interfaces, or events) will be recorded andshown in radiogram images (Figures 5 through 7).

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



Since the strike of a discontinuity can be defined asthe angle of the intersection of the discontinuity planeand a horizontal plane, the resulting linear feature inthe radiogram can be considered as “the strike line”of the detected discontinuity.

For a dipping discontinuity, the migration processresults in the apparent dip angle of the discontinuitybeing corrected to a steeper angle (Cassidy, 2009a,2009b). Consequently, migration will reconstruct theapparent “strike line” of the detected discontinuity toa steeper, “deeper” declination angle of “the strike line.”

Manual migration was done using the 2D GPRradiogram images and based on the followingequation from Aqeel (2012):

sin b~ tana ð3Þ

where b 5 the true declination angle of the strike lineof the detected hidden sub-vertical discontinuity anda 5 the apparent declination angle of the strike line ofthe detected hidden subvertical discontinuity.

Both the apparent perpendicular horizontal depths(distances) (d) and apparent declination angles (a) canbe estimated from the 2D radiograms (Figures 5through 7 and Table 1). Figure 8 explains how toestimate a and then calculate b based on Eq. 3. Thetrue horizontal depths (d9) can then be estimatedusing the following equation (Table 1):

d ’~d= cosb ð4Þ

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

vertical discontinuities can be generated (Figures 9through 11).

COMBINING LIDAR ANDGPR MEASUREMENTS

Introduction

Geospatial coordinate triplets (x, y, z) of any pointon a rock cut are determined by simply identifyingany particular point in the data set. The z-axis is inthe vertical direction, while y and z axes are arbitraryhorizontal axes, as defined by the LIDAR scanner.Selecting three control points on a planar object suchas a discontinuity surface defines the attitude of thatplane. Projecting these three control points horizon-tally and in a perpendicular direction from the rockcut face onto the hidden discontinuity plane willresult in three new coordinate triplets that define theorientation of the hidden discontinuity (x9, x9, z9).

Methodology

The dip direction (w) and dip angle (h) of each rockcut face were measured using a Brunton compass. Oneach rock cut face, the Cartesian coordinates of thethree control points were measured on the LIDARviewer and recorded (Table 2). Using the three-pointmethod, here the Cartesian coordinates converted tospherical (geographical) coordinates and, thus, thegeometry of the rock cut face measured using theLIDAR (Table 2).

Table 1. The apparent and true perpendicular horizontal depths measured from the three control points on the rock cut face to the detectedhidden sub-vertical discontinuities at each station in the study area.

Apparent PerpendicularHorizontal Depths (z) Measured

from the Control Points at St. 1 (cm)

Apparent (a) and True (b)Declination Angles

of the Strike (u)

True Perpendicular HorizontalDepths (d) Measured from the

Control Points at St. 1 (cm)

St. No.Discontinuity

No.Point 1(Blue)

Point 2(Red)

Point 3(Yellow) a b

Point 1(Blue)

Point 2(Red)

Point 3(Yellow)

1 1 148 133 170 11 11.2 151 136 1732 261 234 282 16 16.7 272 244 2943 325 344 314 10 10.2 330 350 319

2 1 188 198 177 16 16.7 196 207 1852 287 285 318 22 23.8 314 311 348

3 1 165 172 206 17 17.8 173 181 2162 208 223 290 35 44.4 291 312 4063 364 396 404 03 03 364 396 404

4 1 348 355 379 11 11.2 355 362 3865 1 187 189 197 7 7.1 188 190 198

2 224 227 240 7 7.1 225 229 2423 248 249 236 9 9.1 251 252 2394 336 323 320 5 5.1 338 324 321



The true depth vector (d) is always perpendicular tothe strike direction at each rock cut face. Sub-sequently, the true depth could be resolved to twocomponents x9 and y9. For instance, Figure 12 showsthat the true depth vector at station 2 was resolved tox9 5 x 2 d cos 63u and y9 5 y 2 d cos 27u. Sometimesthe true depth vector is in the same direction of thedip of the rock cut face. By substituting d values fromTable 1 into the resulting two components for eachrock cut (station), the Cartesian coordinates (x9, y9,z9) for each corresponding point on each detected

discontinuity plane can be mathematically calculated(Tables 3 through 5).

Using the three-point method (Maerz et al.,2012), the resulting Cartesian coordinates of thedetected hidden sub-vertical discontinuities wereconverted to spherical (geographical) coordinates,and, thus, their orientations were measured (Ta-bles 3 through 5). Field verification measurementswere conducted for those detected hidden disconti-nuities that had linear traces appearing on the rockcuts (Tables 3 through 5).




RESULTS AND VERIFICATION

Verification

Field verification (using a Brunton compass) of theLIDAR/GPR orientation measurements was possibleonly on those discontinuities that were exposed on thecut slope, and these were manually measured usinga Brunton compass (Figure 13). The “Results”section below shows that about one half of thediscontinuities measured by the LIDAR/GPR tech-

nique could be identified and measured elsewhere onthe rock cut because they projected to an open face.

The “Results” section also shows that typically theverification dip angle/dip direction measurementswere between 3u and 7u. Typically, rock discontinuitysurfaces have irregular and/or undulated surfaces thatcause a variation in the value of the measured dipangle/dip direction of the slope using a geologicalcompass from one part to another part on the samesurface. The LIDAR/GPR measurements can givemore reasonable and reliable measurements because




they measure the orientation of the discontinuity overa wider base than can be obtained when usinga compass, bridging over the irregularities.

Results

At station 1, the GPR technique was able toidentify three hidden sub-vertical discontinuities to

a depth of 320 cm (Figures 5 and 8). Both ofdiscontinuities 1 and 3 have a dip angle of 69u, withdip directions of 025u and 027u, respectively, whilediscontinuity 3 has a dip angle of 68u and a dipdirection of 024u. Manual measured field verificationmeasurements were conducted by measuring theorientation of those discontinuities that extend(daylight) out of the rock mass off to the side or atthe top of the cut, where they can be measured(Table 3). The difference between the measuredorientation using the LIDAR and geological compasswas within 3u for dip direction and within 4u for dipangle, which can be attributed to human and/ordevice errors, or it can simply be a function ofmeasuring the discontinuity in a different place.

Similarly, at station 2, the GPR technique was ableto identify two hidden sub-vertical discontinuities toa depth of 360 cm (Figures 5 through 7). The dipangle of discontinuity 1 is 78u, and the dip angle is 82ufor discontinuity 2, while the dip direction for thesetwo discontinuities is 029u and 035u, respectively(Table 4). These two mapped discontinuities haveexposed linear traces, which made field verificationseasy and accurate (Figure 13). The difference betweenthe measured geometry using the LIDAR and thatobtained using a geological compass was within 6u fordip direction and within 7u for dip angle.

At station 3, the GPR technique was able toidentify three hidden sub-vertical joints within a per-pendicular depth of 400 cm (Figures 6 and 10). Thedip angles for those three detected discontinuities arebetween 86u and 89u, while the dip direction isbetween 193u and 197u (Table 5). Verification was

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

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



possible for only one hidden discontinuity, the lineartrace of which was exposed (Figure 13). The differ-ence between the measured geometry using theLIDAR and that obtained using a geological compasswas within 3u for dip direction and within 3u for dipangle (Tables 2 and 5).

Only one hidden sub-vertical discontinuity wasidentified at a depth of 360 cm by the GPRinstrument at station 4. The dip direction and dipangle of this discontinuity are 021u and 89u, re-spectively. Verification was not possible for thisdiscontinuity, as it does not “daylight” anywhere.

Verification was also not possible for the fourdetected hidden discontinuities at station 5. The dipangles of these four detected discontinuities are 88u,89u, 88u, and 82u, respectively. Their dip directions are015u, 014u, 016u, and 015u, respectively.

Limitations and Application

A common limitation of using GPR technology isthe GPR data processing and interpretation, which isstill subjective and depends mainly on the interpreter’sskills and experience in interpreting radiograms. In





Table 2. The coordinates of the three control points and, thus, the geometry of the rock cut faces.

Station No. Cartesian Coordinates of the Three Control Points on the Rock Slope FaceGeometry of the Rock Slope Face (u)

In the Field By LIDAR

Point X Y ZDip

DirectionDip

AngleDip

DirectionDip

Angle

1 1 (Blue) 6458.09 16235.77 2324.57 028 73 026 0692 (Red) 7422.82 15878.50 2122.793 (Yellow) 5639.71 16751.02 2003.56

2 1 (Blue) 2991.82 15447.26 464.68 022 85 027 812 (Red) 2979.83 15417.44 260.113 (Yellow) 2194.30 15803.41 248.70

3 1 (Blue) 8966.93 15322.53 2936.98 202 87 199 852 (Red) 8977.10 15359.48 2486.483 (Yellow) 8170.74 15708.86 2394.12

4 1 (Blue) 4506.50 13003.10 21057.25 195 87 202 902 (Red) 4485.41 13012.93 2821.733 (Yellow) 2804.01 13626.68 2670.18

5 1 (Blue) 2644.17 13969.41 588.63 010 88 015 872 (Red) 2599.15 13986.50 741.643 (Yellow) 1629.43 14195.73 775.75

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 theThree Corresponding Control Points LIDAR (u) Field Verification (u)

Discontinuity No.Corresponding

Point No. X9 Y9 Z9

DipDirection

DipAngle

DipDirection

DipAngle

1 1 6391.89 16100.05 2324.57 025 69 025 NA2 7363.20 15756.26 2122.793 5563.87 16595.53 2003.56

2 1 6338.85 15991.30 2324.57 024 68 027 722 7315.85 15659.19 2122.793 5510.82 16486.77 2003.56

3 1 6313.42 15939.17 2324.57 027 69 026 712 7269.38 15563.92 2122.793 5499.86 16464.30 2003.56

*NA 5 Not Applicable

Table 4. LIDAR geometrical measurements for the detected hidden sub-vertical joints at St. 2.




Point No. X9 Y9 Z9

DipDirection

DipAngle

DipDirection

DipAngle

1 1 2903.42 15272.62 464.68 029 78 035 852 2886.47 15233.00 260.113 2110.87 15638.58 248.70

2 1 2850.21 15167.49 464.68 035 82 030 832 2839.57 15140.34 260.113 2307.35 15493.34 248.70



addition, this entire methodology is somewhat timeconsuming and requires lab testing to determine thevelocity of the rock, although this is likely to beconstant over large areas of the same rock type.

Another practical limitation of the method is thenecessity for a relatively flat near-vertical face. Whilethis seems to disqualify a large number of rock cuts, itis also true that because discontinuities tend to occurin parallel to sub-parallel sets, where, consequently,there is sub-vertical concealed discontinuity strikingparallel to a rock cut, often the rock cut itself isa product of a parallel or sub-parallel discontinuity.

A second practical limitation is that the GPSoperation requires physical contact between the GPSunit and the face. This means that equipment likea man-lift would be needed for access to anything thatcannot be reached from a safe standing spot below

the cut. Consequently, the height limitation wouldinvolve the limitations of the man-lift.

CONCLUSIONS

The results of this investigation show how accu-rately joint orientation data can be obtained onhidden sub-vertical discontinuities using a combina-tion of LIDAR and GPR.

Comparison of the orientation results shows thatthe manually (compass-) measured results are veryclose to the LIDAR measurements. Differences arelikely to be caused by the fact that discontinuities arenot perfectly planar, and variable measurements canbe expected depending on the part of the discontinu-ity in which the measurements are made.

Furthermore, the extension of linear traces of somedetected discontinuities measured using a geologicalcompass may differ from those measurements result-ing from GPR and/or LIDAR. This explains why thedifferences between the measured dip direction usinga geological compass and those measured usingLIDAR can be up to 7u.

Migration of the GPR data was a necessary step inorder to accurately estimate the true depths (trueperpendicular horizontal depths) to the detectedhidden discontinuities. Once this step has beenaccomplished,, the true geometry of those disconti-nuities can be measured using our new method.

Not only can hidden discontinuity orientations bemeasured, but any other specific hidden object can bedetected and mapped and then measured in terms ofits orientation in space using this method, whether forcivil or military purposes. Moreover, the results showthat three GPR survey lines conducted on a rockslope face can be enough to obtain GPR data, 2Dradiograms, which can be integrated with LIDARdata for hidden sub-vertical discontinuity orientation

Table 5. LIDAR geometrical measurements for the detected hidden sub-vertical joints at St. 3.




Point No. X9 Y9 Z9

DipDirection

DipAngle

DipDirection

DipAngle

1 1 8910.60 15158.96 2936.98 197 86 194 892 8918.17 15188.34 2486.483 8100.41 15504.63 2394.12

2 1 8872.18 15047.38 2936.98 193 88 NA NA2 8875.51 15064.48 2486.483 8038.55 15324.98 2394.12

3 1 8848.41 14978.36 2936.98 199 89 NA NA2 8848.16 14985.05 2486.483 8039.20 15326.87 2394.12

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.



measurements. Consequently, not only can accuracybe increased, but the labor cost and time-consumingfactors can be significantly reduced to a much greaterextent when 3D GPR radiograms are produced andused for such purposes.

ACKNOWLEDGMENTS

We would like to deeply thank both the GeologicalEngineering Program and Rock Mechanics & Explo-sive Research Center at Missouri University ofScience and Technology for their technical assistance.

REFERENCES

ABELLAN, 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 penetrat-

ing 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.

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, soilsand 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 ofElectrical Engineers, London, U.K, 726 p.

DONOVAN, J.; KEMENY, J.; AND HANDY, J., 2005, The application ofthree-dimensional imaging to rock discontinuity characteriza-tion, 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, Structuraland geomorphological features of landslides in the BhutanHimalaya derived from terrestrial laser scanning: Geomor-

phology, Vol. 103, pp. 17–29.

FASCHING, A.; GAICH, A.; AND SCHUBERT, W., 2001, Data

acquisition in engineering geology: An improvement ofacquisition methods for geotechnical rock mass parameters:

Felsbau, Vol. 19, No. 5, pp. 93–101.

GAICH, A.; POTSCH, M.; AND SCHUBERT, W., 2006, Acquisition and

assessment of geometric rock mass features by true 3Dimages. 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.

GARCIA-SELLES, D.; FALIVENE, O.; ARBUES, P.; GRATACOS, O.;TAVANI, S.; AND MUNOZ, J. A., 2011, Supervised identification

and reconstruction of near-planar geological surfaces fromterrestrial laser scanning: Computers Geosciences, Vol. 37,

pp. 1584–1594.

GONZALEZ-JORGE, H.; RIVERIO, B.; ARMESTO, J.; AND ARIAS, P.,

2011, Standard artifact of the geometric verification ofterrestrial laser scanning systems: Optics Laser Technology,Vol. 43, pp. 1249–1256.

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



GREGOIRE, C., 2001, Fracture Characterization by Ground-Pene-trating Radar: Unpublished Ph.D. Dissertation, KatholiekeUniversity of Leuven, Belgium.

HANEBERG, W., 2008, Using close range terrestrial digitalphotogrammetry for 3-D rock slope modeling and disconti-nuity mapping in the United States: Bulletin EngineeringGeology Environment, Vol. 67, No. 4, pp. 457–469.

JENYON, M. K. AND FITCH, A. A., 1985, Seismic ReflectionInterpretation: Stuttgart- Gebruder-Borntraeger, Berlin, Ger-many.

KEMENY, J.; NORTON, B.; AND TURNER, K., 2006, Rock slopestability analysis utilizing ground-based lidar and digitalimage processing: Felsbau, Vol. 24, No. 3, pp. 8–15.

KLEYN, A. H., 1983, Seismic Reflection Interpretation: ElsevierApplied Science Publishers LTD, London, England, 269 p.

KOVIN, O. N., 2010, Ground Penetrating Radar Investigations inUpper Kama Potash Mines: Unpublished Ph.D. Dissertation,Missouri University of Science and Technology, 182 p.

LINES, L. R. AND NEWRICK, R. T., 2004, Fundamentals ofGeophysical Interpretation: In Geophysical Monograph Series,No. 13: Society of Exploration Geophysicists, Tulsa, OK.

MAERZ, N. AND KIM, W., 2000, Potential use of groundpenetrating 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 discontinu-ity orientations from terrestrial lidar images: EnvironmentalEngineering Geoscience, Vol. 19, pp. 185–195.

OTOO, J.; MAERZ, N.; XIAOLING, L.; AND DUAN, Y., 2011, 3-Ddiscontinuity orientations using combined optical imagingand Lidar techniques. In Proceedings of the 45th U.S. RockMechanics Symposium: San Francisco CA, 26–29 June2011, 9 p.

OTOO, J. N.; MAERZ, N. H.; DUAN, Y.; AND XIAOLING, L., 2012,Verification of a 3-D lidar point cloud viewer for measuringdiscontinuity orientations. In The 46th U.S. Rock Mechanics/Geomechanics Symposium: San Francisco, CA, 24–27 June2012.

OTTO, J. C. AND SASS, O., 2006, Comparing geophysical methodsfor talus slope investigations in the Turtmann Valley (SwissAlps): Geomorphology, Vol. 76, pp. 257–277.

PERNITO, M., 2008, Rock Mass Slope Stability Analysis Based on3D Terrestrial Laser Scanner and Ground Penetrating Radar:Unpublished M.S. Thesis, ITC: International Institute forGeo-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 assessmentin 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 Pro-

specting, Vol. 49, pp. 300–308.

STURZEINEGGER, M. AND STEAD, D., 2009, Close-range terrestrialdigital 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: Interna-

tional 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.



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, Inter-layered Rock Units, Differential Weathering, Rockfalls

ABSTRACT

Designing cut slopes along Ohio highways depends onlocal stratigraphy and slope stability problems. Basedon stratigraphy and modes of failure, cut slopes in Ohiowere divided into three types: 1) those consisting ofstrong rock units (sandstones and limestones) thatexhibit discontinuity-related failures; 2) those consistingof weak rock units (shales, claystones, and mudstones)that exhibit raveling, gully erosion, and rotationalsliding; and 3) those consisting of interlayered strongand weak rock units where differential weatheringcauses undercutting-induced failures. Data regardinggeological, geotechnical, and geometrical parameterswere collected for 26 sites representing the three typesof slopes and were used to perform kinematic analysis,rockfall trajectory simulations, and global stabilityanalysis. This article focuses on the design of cut slopesin interlayered stratigraphy where differential weather-ing is the primary cause of slope instability. Based onstratigraphic variations, we categorized cut slopes in theinterlayered units into four types: Type I—thicksandstone underlain by thick shale or claystone/mudstone; Type II—sandstone interlayered with shaleor claystone/mudstone in nearly equal proportions;Type III—limestone interlayered with claystone/mud-stone in nearly equal proportions; and Type IV—claystone/mudstone interlayered with minor, thin lime-stone layers. Based on stability analyses and rockfallsimulations, we recommend cut slope designs for eachstratigraphic sequence that consider slope angles forundercut units to reduce rockfall potential, slope anglesfor undercutting units that are close to naturally stableangles, benches to reduce undercutting and contain

rockfalls, drainage to reduce erosion, and catchmentditches to contain rockfalls.

INTRODUCTION

Most cut slopes in Ohio consist of interlayeredsequences of strong and weak rock units of varyingthicknesses. These slopes are highly subject todifferential weathering and undercutting-induced fail-ures (Shakoor and Weber, 1988; Shakoor, 1995; andAdmassu and Shakoor, 2012). Currently availablemethods of rock slope design can be used for designingcut slopes in uniformly strong rocks or uniformlyweak rocks but they cannot be directly applied fordesigning cut slopes in interlayered sequences of strongand weak rocks. The purpose of this study, which wasfunded by the Ohio Department of Transportation(ODOT), is to develop a rational approach fordesigning cut slopes in sub-horizontal, interlayeredsequences of sedimentary rocks subject to differentialweathering. The approach is based on a detailedinvestigation of 26 cut slope sites representative ofvarying stratigraphic conditions present in Ohio.

GEOLOGIC SETTING OF OHIO

The geologic setting of Ohio is mainly a result ofPaleozoic sedimentation and Pleistocene glaciation.The oldest rocks in Ohio are Ordovician-age lime-stones deposited in a shallow, warm sea (Camp, 2006).The Acadian mountains, resulting from the collisionof the Baltica plate and the North American plate(,375 mega-annum [Ma]), supplied sediments for theDevonian-Mississippian-age sandstones and shales ofOhio. The latest Alleghenian orogeny (,318 Ma)resulted in the rise of the Appalachian Mountains thatprovided the source of Ohio’s Pennsylvanian-Permian-age sedimentary rocks. There is, however, little to nosedimentary record present in Ohio from the latePaleozoic, Mesozoic, and most of Cenozoic time1Corresponding author email: [email protected].


(between 290 million and 300,000 years) present inOhio.

As a result of the tectonic history, the bedrockgeology of Ohio consists of nearly flat-lying carbon-ate and siliciclastic sedimentary rocks from UpperOrdovician to Lower Permian. In southwestern Ohio,bedrock predominantly consists of Upper Ordovicianinterlayered limestones and shales or claystones/mudstones. Dolomites and shales of Silurian ageunderlie the western and west-central parts of thestate. The northwestern and central parts of Ohio areunderlain by Devonian marine carbonate rocks(limestones and dolomites interlayered with shales)as well as siliciclastic rocks (sandstones and siltstonesinterlayered with shales). Mississippian-age rocks,including sandstones, siltstones, conglomerates, andshales, with minor proportions of limestone, cover theeast-central part and northwestern corner of the state.The largest part of eastern Ohio is covered byPennsylvanian-age rocks, including sandstones, silt-stones, limestones, shales, claystones/mudstones, andsome coals. Lower Permian/Upper Pennsylvanian-agesandstones, siltstones, shales, claystones/mudstones,and minor coal seams cover southeastern Ohio. Theserocks were deposited as cyclothems in non-marine,deltaic, or estuarine environments (Chesnut, 1981;Bennington, 2002).

Most cut slopes in Ohio are located in the easternand southeastern parts of Ohio, which are character-ized by interlayered sandstones, limestones, shales,claystones, mudstones, and minor amounts of coal.Slopes in the southwestern part are characterized byinterlayered limestones, shales, claystones, and mud-stones. Only a few rock slopes are present in centralOhio.

Types of Rock Slopes and Modes of Failure in Ohio

Based on stratigraphy, cut slopes in Ohio can bedivided into three broad types: 1) those that comprisemostly (.90 percent) strong rock units (sands-tones, limestones); 2) those that comprise mostly(.90 percent) weak rock units (shales, claystones,mudstones); and 3) those that comprise interlayeredstrong and weak rock units. Cut slopes belonging totype 1 are usually less than 30 ft (9 m) high and makeup 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 Ohiobelong to type 3. These cut slopes are highly variablein height and may contain from a few to manyinterlayered strong and weak rock units of varyingthicknesses. For cut slope design purposes, these threetypes of slopes can be considered to have threedistinctly different design requirements.

The failures that may affect type 1 slopes includeplane, wedge, and toppling failures that are triggeredby an unfavorable orientation of discontinuities withrespect to orientation of the slope face (Admassu andShakoor, 2013a). The common modes of failureaffecting type 2 cut slopes are general degradationand raveling of weak rock due to weathering withraveled material accumulating at the bottom ofthe slope (Figure 1), gully erosion and mudflows(Figure 2), and, less frequently, rotational slides.Type 3 slopes experience failures that are typical ofboth strong rock units and weak rock units.Differences in durability of interlayered rock unitscauses differential weathering, resulting in undercut-ting of the stronger rock layers, creating overhangs.Once the depth of undercutting exceeds the jointspacing within the undercut layer, rockfalls begin tooccur (Figure 3). Undercutting also promotes plane,wedge, and toppling failures along discontinuitiesthat do not “daylight” on the original cut slope(Shakoor and Weber, 1988; Shakoor, 1995). Depend-ing on their location on the slope face, these failuresbecome rockfalls as they descend the slope (Shakoorand Weber, 1988). Thus, rockfalls are the dominantmode of failure affecting cut slopes subject todifferential weathering.

Kinematics of Undercutting-Induced Failures

Undercutting-induced failures are kinematicallypossible when at least three sets of intersecting

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

Admassu and Shakoor


discontinuities are present so that a rock block canmove freely when the depth of undercutting exceedsthe spacing between the discontinuities. The threecommon types of discontinuities present in Ohio arebedding planes, orthogonal joints, and valley stressrelief joints. Orthogonal joints are sub-vertical joints

that develop perpendicular to each other. Valleystress relief joints result from horizontal extension ofvalley walls as stream erosion removes lateral support(Ferguson and Hamel, 1981). The rock blocksreleased as rockfalls can be bounded either by thebedding and two sets of orthogonal joints or bybedding, a set of orthogonal joints, and a set of stressrelief joints. When the undercut blocks are firstreleased, the initial movement could be either in theform of a plane failure, a wedge failure, or a topplingfailure (Shakoor and Weber, 1988; Shakoor, 1995).Toppling failures, associated with undercutting, occurwhen the depth of undercutting extends beyond theblock’s center of gravity (Neimen, 2009). As a resultof the dominance of near-vertical discontinuities inOhio, toppling is a common mode of failure. Re-gardless of the initial mode of failure, all undercut-ting-induced failures become rockfalls.

RESEARCH METHODS

Site Selection

We performed a reconnaissance survey of 113 cutslope sites across the state of Ohio to select 26 sites(Appendix 1 and Figure 4) for detailed investigationsfor this study. Site designation in Appendix 1 andFigures 1 through 4 follows ODOT standard nota-tion, which uses the three-letter county code, thenumerical name of the road, and the mile markerfrom the county line. The sites were selected to ensurethat they are representative of different stratigraphicscenarios and geologic ages. Appendix 1 shows that ofthe 26 sites, six consist mostly of strong rock units,two mostly of weak rock units, and 18 mostly ofinterlayered rock units. The selected sites have

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

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

Cut Slope Design


a greater representation of slopes consisting ofinterlayered strong and weak rock units because suchslopes are the most common and most problematic interms of performance and hazard potential. Thisstudy focuses on design of cut slopes for theinterlayered stratigraphy based on the 18 sitesrepresentative of this type of stratigraphy.

Field Investigations

Field investigations consisted of collecting dataregarding slope geometry (slope angle, slope height,slope aspect, slope profile, bench width), slopestratigraphy, discontinuity characteristics (orienta-tion, continuity, spacing, nature of infilling material,

Figure 4. Location map of the 26 study sites.

Admassu and Shakoor


surface irregularities, groundwater conditions), depthof undercutting, and catchment ditch dimensionswidth and depth). A laser range finder was used toprepare the slope profiles, which were then used toprepare stratigraphic cross sections for each site.Fifteen of the sites were drilled, and borehole logsfrom drilling were used to refine the stratigraphicdetails. Field values of rock quality designation(RQD) for stronger rock units were estimated usingthe method proposed by Palmstrom (1982), which isbased on the number of joints within a cubic meter ofrock outcrop. The detailed line survey method (Piteauand Marin, 1977), the window mapping method(Wyllie and Mah, 2004), and random measurementswere used to collect the discontinuity data. Ameasuring tape and laser range finder (for inaccessi-ble layers) were used to measure the depth ofundercutting. The presence of pre-split blast-holemarkings on the undercut rock unit provideda reference to ensure that the undercut unit hadremained in place since the time of construction.Where pre-split markings were absent, original designplans for the cut were used for this purpose.

Laboratory Investigations

Laboratory tests were performed to determineunconfined compressive strength, second-cycle slakedurability index (Id2), friction angle, and densityvalues for both strong and weak rock units. Un-confined compressive strength and slake durabilityindex were determined using American Society forTesting and Materials (ASTM) methods D2968 andD4644, respectively (ASTM, 1996); density wasdetermined by measuring weights and volumes ofoven-dried core samples; and Stimpson’s method(Stimpson, 1981) was used for determining frictionangle. Laboratory data were used for various types ofstability analyses. Unconfined compressive strengthvalues ranged from 1,179 to 21,507 psi (8.13–148.32MPa), with a mean of 9,001 psi (62.10 MPa), forsandstone core samples (n 5 34); from 4,148 to 25,699psi (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 meanof 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 second-cycle slake durability index values for sandstones(n 5 34), limestones (n 5 18), shales (n 5 30), andclaystone/mudstone (n 5 26) samples were found to be93 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), respecti-vely. The mean basic friction angle along disconti-nuities was 36u for sandstones and 43u for limestones.

Stability Analysis

Undercutting-induced rockfalls, promoted by dif-ferential weathering, represent the most common typeof failure affecting cut slopes with interlayeredstratigraphy. A rational method for analysis anddesign of cut slopes subject to differential weatheringcannot be developed without understanding thefactors that cause undercutting. We used bivariateand multivariate statistical methods to identify thegeological and geotechnical factors that influence thedepth of undercutting. SPSS (Statistical Package forthe Social Sciences) and Microsoft Excel softwarewere used for statistical analysis. A detailed discus-sion of the factors affecting the depth of undercuttingcan be found in Admassu et al. (2012). The fate of therockfalls with respect to the effect of slope height,slope angle, and catchment ditch dimensions wasevaluated using rockfall simulation software, RocFall(Rocscience, 2006). RocFall determines the trajectoryand the landing space of a rockfall generated fromany point on the slope face.

We performed kinematic analysis, using DIPSsoftware (Rocscience, 2006), to evaluate the potentialfor plane, wedge, and toppling failures for thick(.10 ft/3 m), strong rock units within the interlayeredsequences. Kinematic analysis did not show anysignificant potential for plane or wedge failures dueto persistently steep dips of the discontinuities. Inaddition, we used the SLIDE software (Rocscience,2006) to determine factor-of-safety values againstrotational failures due to low rock mass strength forslopes consisting of interlayered rock units. Com-pressive strength and geologic strength index data(GSI) were used for this analysis. The GSI data wereobtained using the method and charts developed byMarinos and Hoek (2001) and Hoek et al. (2005). Theresulting factor-of-safety values ranged from 2.7 to40. We also observed naturally stable slope angles forweak rock units at numerous sites to aid in selectingappropriate cut slope angles for weak rock units.

Information about the factors that have the mostinfluence on the depth of undercutting (obtained fromstatistical analyses), the expected frequency, sizes, andfate of rockfalls (obtained from RocFall analysis),was used to suggest an appropriate cut slope design(slope angle, bench width and location, catchmentditch width, stabilization techniques) for the inter-layered strata.

Cut Slope Design


DESIGN APPROACH

Based on our detailed investigations of the 18 cutslopes in interlayered stratigraphic sequences in Ohio,we propose the following approach for designing cutslopes subject to differential weathering:

1. Categorize the site stratigraphy into one of fourcategories, as discussed later.

2. Identify the factors that influence the depth ofundercutting at the site and prepare a list ofmethods that may be used to reduce the effect ofthese factors.

3. Evaluate the appropriate slope angles for strongrock units, taking into account the number ofundercutting-induced rockfalls, using the two-dimensional (2D) numerical simulation softwareUDEC (UDEC, 2014).

4. Evaluate the ultimate stable angles for weak,undercutting rock units.

5. Evaluate the fate of rockfalls including bounceheights and roll out distances, using rockfalltrajectory simulation software (RocFall), for de-signing catchment ditches.

6. Design benches that take into account the site-specific stratigraphic conditions so that the poten-tial for undercutting is reduced and some of therockfalls may land on the benches, consideringrockfall roll out distances.

Categorizing Site Stratigraphy for Design Purposes

On the basis of variations observed within theinterlayered stratigraphic sequences in the Appala-chian plateau of Ohio, we categorized stratigraphyinto four types: Type I—thick sandstone layerunderlain by thick shale or claystone/mudstone layer;Type II—sandstone interlayered with shale or clays-tone/mudstone in nearly equal proportions; TypeIII—limestone interlayered with claystone/mudstonein nearly equal proportions; and Type IV—claystone/mudstone interlayered with minor limestone. Figure 5shows 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 followingfactors were studied for their possible contribution tothe depth of undercutting: 1) vertical distance of theundercut unit from the slope crest; 2) relative positionof the undercut unit from the slope crest, defined asthe vertical distance of the undercut unit from theslope crest divided by the total slope height.; 3) total

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

Admassu and Shakoor


thickness of the undercut unit; 4) spacing oforthogonal joints within the undercut unit; 5) slakedurability index value of the undercutting unit; 6)initial slope angle; and 7) age of the road cut.

Multiple regression analyses conducted by Ad-massu et al. (2012) showed that the seven factorsidentified above explained 61 percent of the variationin the depth of undercutting, with position of theundercut unit and spacing of joints in the undercutunit explaining the most variation. Undercut unitscloser to slope crest and highly jointed undercut unitshave the highest susceptibility to deeper undercutting.Admassu et al. (2012) attribute the unexplained39 percent of the variation to the amount of waterseeping along the contact between the undercut andthe undercutting units (i.e., surface runoff also

contributes to undercutting; Figure 6). Shakoor andRogers (1992) showed that the rate of undercutting isdependent on the durability of the undercutting rockunit. Admassu et al. (2012) and Neimen (2009)suggest that the rate of undercutting is not necessarilylinear and might slow down over time as theundercutting unit reaches a stable angle. The abovediscussion suggests that during cut slope design,special considerations should be given to undercutlayers close to the slope crest and to the highly jointedstrong rock layers because they will likely be deeplyundercut.

Effect of Slope Angles for Strong Rock Units on theNumber of Rockfalls

The main discontinuities within strong rock units(sandstones, limestones) of interlayered stratigraphicsequences are invariably sub-horizontal bedding planesand orthogonal joints (Admassu et al., 2012). Theorthogonal joints do not exhibit any preferred orien-tation but consistently show steep dips (average 79u)and steep lines of intersection (.70u). Kinematicanalysis, using DIPS software (Rocscience, 2006),shows that plane and wedge failures are not possiblebecause of the steepness of discontinuities. Therefore,undercutting-induced rockfalls, including toppling, isthe primary mode of failure in interlayered stratigra-phy. Using a 2D numerical modeling software,UDEC (UDEC, 2014), we investigated the effect ofvarying slope angles on the number of undercutting-induced rockfalls (Admassu and Shakoor, 2013a).The results of our investigation showed that strongrock units cut at 1H:1V (45u) angles resulted in thesmallest number of rockfalls. However, consideringthe large amount of excavation required to cut slopesat 1H:1V (45u), we recommend an angle of 0.5H:1V(63u), which would significantly reduce the number ofrockfalls without requiring too much excavation.

Stable Slope Angles for Weak, UndercuttingRock Units

One important question that needs to be addressed iswhether the weak rock units tend to reach a final stableangle beyond which further undercutting will notoccur. Angles of undercutting rock units that appearedstable, as indicated by the presence of vegetation, weremeasured for the 18 study sites as well as for someadditional sites. These angles show a normal distribu-tion, with an average value of 38u (Figure 7), which maybe considered as the final stable angle for undercuttingrock units. Therefore, weak units cut at 1.5H:1V (35u)are expected to stabilize sooner, consequently causingfewer rockfalls over time. However, where more

Figure 6. (a) Undercutting caused by groundwater flowing outof joints and (b) undercutting caused by surface runoff overa slope face.

Cut Slope Design


durable silty shales form the weaker rock units, 1H:1V(45u) slope angles may be acceptable.

Fate of Rockfalls and Design of Benches andCatchment Ditches

An important consideration when designing cutslopes prone to undercutting-induced failures is tostudy the trajectories and volume of released rock-falls. This information is vital for designing benchesand rockfall catchment ditches. We investigated thefollowing aspects of rockfalls to address this issue:

1. Effect of rockfall shape and type of rock compri-sing the rockfalls on the fate of rockfalls.

2. Effect of slope angle, slope height, and slopestratigraphy 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 simulationsoftware RocFall (Rocscience, 2006). This informa-tion is essential for designing benches and width,depth, and slope angle of catchment ditches.

Field observation and measurement of rockfalldimensions revealed that the travel behavior of

rockfalls is governed mainly by the ratio of beddingthickness to joint spacing for the undercut units.Equi-dimensional rockfalls with a ratio of 1 tend toroll, whereas rock units for which this ratio is lessthan 1 result in flat rockfalls that tend to slide andmay remain on the slope instead of landing in thecatchment ditch (Admassu and Shakoor, 2013b).Rockfalls generated from undercut limestone units,with an average bedding thickness to joint spacingratio of 0.96, tend to roll and, therefore, have greatertrajectories. On the other hand, rockfalls comprisedof sandstone, with an average bedding thickness tojoint spacing ratio of 0.53, have flatter shapes andare more likely to stay on the slope face. The sizes ofundercutting-induced rockfalls also vary substantial-ly, ranging from less than 0.81 ft3 (0.03 m3) to110.70 ft3 (4.1 m3).

In addition to investigating rockfall shape–con-trolled trajectories, we performed simulations, usingRocFall software (Rocscience, 2006), for the differentrockfall weights (average weight for sandstone rock-falls 5 764 lb/361 kg; average weight of limestonerockfalls 5 68 lb/31 kg), stratigraphic configurations,slope angles, and catchment ditch slopes. Thesimulation considered pre-split blasting that is usedto create a smooth slope surface during slopeexcavation. Pre-split blast-holes can be drilled toa maximum depth of 40 ft (12 m) before providinga 1.5 ft (0.5 m) offset for the next phase of drilling.The maximum roll out distances of rockfalls releasedfrom different heights, different slope angles, anddifferent catchment ditch slopes were recorded foreach type of stratigraphic configuration (Appendix 2).The widths of benches and catchment ditches arebased on the rockfall roll out distances. In order torecommend catchment ditch width and bench widthfor different slope angles, a relationship betweenrockfall roll out distances to slope height wasestablished 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 Option 2 Option 1 Option 2

Slope1 Width (ft) Slope Width (ft) Slope Width (ft) Slope Width (ft)

Flat 0.4 3 H + 3 Flat 0.3 3 H Flat 0.35 3 H + 5 Flat 0.4 3 H + 76:1 0.4 3 H 2 2 6:1 0.3 3 H 2 1 6:1 0.3 3 H + 5 6:1 0.25 3 H + 93:1 0.1 3 H + 2 3:1 0.15 3 H + 1 3:1 0.2 3 H + 5 3:1 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. Catchmentditches in Ohio are usually designed with slope angles of 6:1, 3:1, or zero (flat).

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

Admassu and Shakoor


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 0.8 3 H + 15 Flat 1.1 3 H + 7 Flat 1.25 3 H + 4 Flat 1.1 3 H + 86:1 0.6 3 H + 4 6:1 0.55 3 H + 6 6:1 0.55 3 H + 4 6:1 0.55 3 H3:1 0.35 3 H + 5 3:1 0.35 3 H + 5 3:1 0.35 3 H + 3 3:1 0.25 3 H + 2

DESIGN RECOMMENDATIONS

Our recommendations for designing cut slopes ininterlayered stratigraphic sequences subject to differ-ential weathering take into account the four types ofstratigraphic configurations, the associated slope in-stability problems, and the anticipated trajectories ofrockfalls. The recommendations are aimed at 1)reducing the number of undercutting-induced rockfallsby selecting appropriate cut slope angles for thick,strong rock units, based on modelling results usingUDEC software; 2) providing stabilization measuresfor undercut rock units that have the highest potentialfor the deepest undercutting, as suggested by thepreviously discussed factors influencing the depth ofundercutting; 3) reducing the rate of undercutting byselecting slopes for undercutting rock units that areclose to their naturally stable angles, providing benchesalong the contacts between undercut and undercuttingunits when feasible, and providing drainage to in-tercept water seeping from the fractures in theundercut, strong rock units onto the undercuttingunits; and 4) reducing the rockfall hazard to theroadways by providing adequate catchment ditchesbased on rockfall trajectories. In the following sections,we provide specific recommendations for each of thefour types of stratigraphic scenarios. Table 2 providesa summary of the design recommendations.

Type I Stratigraphy

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

reduce rockfalls due to toppling failures and cuttingshale or claystone/mudstone at 38u or less. Based onrockfall simulations, a bench should be provided alongthe contact between the two units to delay the processof 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 usedfor the sandstone layer, and the maximum allowablesingle slope should not exceed 40 ft (12 m), which is themaximum depth for an offset to be provided duringpre-split blasting. Catchment ditch width should bebased on the guidelines provided in Table 2. In order toreduce surface runoff on the slope of the undercut unit,a drainage ditch, filled with rip rap, should be providedbehind the crest of the sandstone slope (Figure 8). Inaddition, the seepage along the sandstone-shale con-tact should be collected by a rip rap–filled ditchconstructed on the bench. Both drainage ditches shouldbe 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 beprovided, as rockfalls will have shorter trajectoriesfrom the steeper slope. The underlying unit can be cutat 1.5H:1V (34u), gentler than 38u, so that degradationof the underlying slope will not compromise the widthof the bench. The catchment ditch design should bebased on the guidelines provided in Table 2.

Type II Stratigraphy

Type II stratigraphy consists of thin (,3 ft/1 m)sandstone interlayered with shale or claystone/mud-stone in variable proportions. We recommend usinga uniform slope angle for this type of stratigraphythat would result in fewer rockfalls. Pre-split blastingshould be used, and the maximum slope height shouldnot exceed 40 ft (12 m). Benches should be providedat a maximum slope height of 40 ft (14 m). Twodesign options are shown in Figure 9. Option 1

Cut Slope Design


involves using a uniform slope of 1H:1V (45u), whichis close to the average natural slope angle and not toogentle for pre-split blasting. Catchment ditches orbenches can be designed based on the rockfallsimulation results shown in Table 2. A drainage ditchshould be provided on the slope crest to reducesurface runoff. For option 2, in order to reduce theamount of excavation, the slope can be cut at a steeperangle of 0.5H:1V (63u), which can significantly reducethe number of undercutting-induced rockfalls. Thetop half of the slope should be stabilized usingshotcrete to prevent the weathering of weak layers.Perforated drain pipes should be installed through theshotcrete so that groundwater seepage is not blocked.Catchment ditches and benches should be designedaccording to the results of rockfall simulation, assummarized in Table 2.

Type III Stratigraphy

Type III stratigraphy involves limestone interlay-ered 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

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

StratigraphicGroup Slope Angle

Bench Design:Width (ft)

Catchment Ditch Design

Stabilization Drainage2Slope1 Width (ft)

Type I(Figure 8)

Option 1 0.5H:1V forsandstone/1H:1V forshale

0.5 3 H + 2 (H . 40 ft) Flat6:13:1

0.4 3 H + 30.4 3 H 2 20.1 3 H + 2

Slope crest, bench,and catchment ditchdrainage should beprovided

Option 2 0.25H:1V forsandstone/1.5H:1Vfor shale

0.45 3 H 2 2 (H . 40 ft) Flat6:13:1

0.3 3 H0.3 3 H 2 10.15 3 H + 1

Rock boltsshould be usedto stabilize theupper half ofthe slope

Slope crest, bench,and catchment ditchdrainage should beprovided

Type II(Figure 9)

Option 1 Uniformangle of1H:1V

0.35 3 H + 5 (H . 40 ft) Flat6:13:1

0.35 3 H + 50.3 3 H + 50.2 3 H + 5

Slope crest andcatchment ditchdrainage shouldbe provided

Option 2 Uniformangle of0.5H:1V

0.4 3 H + 7 (H . 40 ft) Flat6:13:1

0.4 3 H + 70.25 3 H + 90.25 3 H + 6

Shotcrete(concretespraying) thetop half of theslope


Type III(Figure 10)


0.8 3 H + 15 (H . 40 ft) Flat6:13:1

0.8 3 H + 150.6 3 H + 40.35 3 H + 5



1.1 3 H + 7 (H . 40 ft) Flat6:13:1

1.1 3 H + 70.55 3 H + 60.35 3 H + 5

Shotcrete(concrete spra-ying) the tophalf of the slope


Type IV(Figure 11)


1.25 3 H + 4 (H . 40 ft) Flat6:13:1

1.25 H + 40.55 3 H + 40.35 3 H + 3

Erosion-controlmatting

Slope crest, mid-slope,and catchment ditchdrainage should beprovided


1.1 3 H + 8 (H . 40 ft) Flat6:13:1

1.1 3 H + 80.55 3 H0.35 3 H + 2

Erosion-controlmatting

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 thevertical. Catchment ditches in Ohio are usually designed with slope angles of 6:1, 3:1, or zero (flat).2Slope crest drainage refers to the provision of a drainage ditch on top of the backslope of a bench. Bench drainage refers to the provision ofa drainage ditch on the bench. Mid-slope drainage refers to the provision of a drainage ditch in the middle of a slope. Catchment ditchdrainage refers to the provision of a drainage ditch within the catchment ditch at the toe of the slope (Figures 8–11).

Admassu and Shakoor


should be provided. Benches, when necessary(slopes . 40 ft/12 m high), and catchment ditchesshould be designed using Table 2. The catchment ditchwidths, based on the trajectories of simulated rockfalls(Table 2), tend to be extremely large. Therefore,narrower catchment ditches with catchment fencesshould be considered. Option 2 utilizes a cut slope of0.5H:1V (63u), which will require less excavation than

option 1 but at the same time will reduce undercutting-induced failures to a large extent. The upper part of theslope can be stabilized with well-drained shotcrete.Slope crest drainage and bench drainage should beprovided (Figure 10). Catchment ditch/bench widthsare summarized in Table 2.

Type IV Stratigraphy

Type IV stratigraphy is characterized by claystone/mudstone units with minor layers of limestone. Pre-split blasting is not necessary, as these slopes consistmainly of weak rocks. Slopes cut in Type IVstratigraphy also generate cubical rockfalls that canhave long trajectories. Since catchment ditches tocontain such rockfalls can be excessively wide,catchment fences should be considered. The best

design approach for such slopes is to use a gentleslope angle that is close to the natural stable angle of38u and to provide slope crest drainage to reducesurface erosion. Since such slopes are highly prone tosurface erosion, a mid-slope drain, lined with rip rapand connected to the back slope drain, may benecessary for high cut slopes. These highly erodibleslopes should also be covered by erosion-controllingmats made from biodegradable materials. Figure 11provides 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 theslope is higher than 20 ft (6 m), a mid-slope drainageditch connected to the slope toe drain should beprovided. If there is a layer of limestone close toa mid-slope section, the mid-slope drain should followthe limestone layer to siphon away seeping ground-water as well as to collect surface water (Figure 11).Erosion-control matting should be installed. Catch-ment ditch/bench design should be based on rockfallsimulation, as summarized in Table 2. Rockfallcatchment fences should be used as part of the ditchdesign to reduce the width of catchment ditches. If thelimestone layers represent , 20 percent of the slopeface, a steeper slope angle of 1H:1V (45u) can be usedas option 2. Mid-slope and slope crest drainageditches and erosion-control matting should be

Figure 10. Design options for Type III stratigraphy.

Figure 8. Recommended slope design for Type I stratigraphy.

Figure 9. Design options for Type II stratigraphy. Figure 11. Design options for Type IV stratigraphy.

Cut Slope Design


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

CONCLUSIONS

Based on the results of this study, the followingconclusions can be drawn:

1. Slope stability problems in Ohio depend on theprevalent stratigraphy, which often consists ofstronger, durable rock units (sandstones, lime-stones) alternating with weaker, non-durable rockunits (shales, claystones, mudstones). This type ofstratigraphy is highly prone to differential weath-ering, which results in undercutting and formationof unsupported overhangs. Undercutting leads tovarious types of slope failures, such as rockfalls,toppling failures, plane failures, and wedge failures.

2. Cut slopes in Ohio can be divided into threedistinctly different types: 1) those consisting mostly(.90 percent) of strong rock units; 2) thoseconsisting mostly (.90 percent) of weak rockunits; and 3) those consisting of interlayered strongand weak rock units, each ranging in proportionfrom more than 10 percent to 90 percent. Most cutslopes in Ohio are located in the interlayeredstratigraphic sequences.

3. Slope stability problems affecting the interlayeredrock units are primarily undercutting-inducedfailures (plane failure, wedge failure, topplingfailure). Regardless of the mode of failure, allundercutting-induced failures become rockfalls.

4. Cut slope design for interlayered sequences iscomplex and depends on stratigraphic variations.Four stratigraphic variations, designated as TypesI through IV, are recognized within the interlaye-red sequences, as follows: Type I—thick (.7–10 ft/2–3 m) sandstone or limestone underlain by shaleor claystone/mudstone; Type II—thin to medium-thick (,3 ft/1 m) sandstone units interlayered withshale or claystone/mudstone units in variableproportions; Type III—thin to medium-thicklimestone (,3 ft/1 m) units interlayered withclaystone/mudstone units in variable proportions;and Type IV—thin to medium-thick (,3 ft/1 m)limestone units interlayered with claystone/mud-stone units in usually minor proportions. The cutslope design for Type I stratigraphy combinesdesign principles for strong and weak rocks, withthe provision for a bench along the contact betweenthe two rock types. The sandstone can be cut at0.5H:1V (63u) and the shale at 1H:1V (45u). For

stratigraphic variations II and III, cut slopes cangenerally be designed at 1H:1V (45u) or 0.5H:1V(63u), and for variation IV they can be designed at1H:1V (45u). Rockfall simulation is required fordesign of drainage ditch widths and bench widths.Table 2 in the text provides a summary of designrecommendations.

ACKNOWLEDGMENT

The authors would like to thank the Ohio De-partment of Transportation (ODOT) and the FederalHighway Administration (FHWA) for the financialsupport of this research project.

DISCLAIMER

The authors are solely responsible for the contentsof this this article, including the accuracy of the data.The contents do not reflect the official views orpolicies of the Ohio Department of Transportation orthe Federal Highway Administration.

REFERENCES

ADMASSU, Y. AND SHAKOOR, A., 2013a, Cut slope designrecommendations for sub-horizontal hard sedimentary rockunits in Ohio, USA: Geotechnical Geological Engineering,Vol. 31, pp. 1207–1219.

ADMASSU, Y. AND SHAKOOR, A., 2013b, Computer simulation-based evaluation of rock fall roll-out distances for catchmentditch 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, Evaluatingselected factors affecting the depth of undercutting in rockssubject 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 lossof marine biofacies with increasing proximity to detritalsource: 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 ofPetroleum Geologists, Memoir 19, Ontario, Canada,pp. 12–21.

CAMP, M. J., 2006, Roadside Geology of Ohio: Mountain PressPublishing Company, Missoula, MT, 412 p.

CHESNUT, D. R., 1981, Marine zones of the Upper Carboniferousof 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 AmericaCoal Division Field Trip, Kentucky Geological Survey,Lexington, KY, pp. 57–66.

FERGUSON, H. F. AND HAMEL, J. V., 1981, Valley stress relief inflat lying sedimentary rocks: Proceedings International

Admassu and Shakoor


Symposium Weak Rock, September 21–24, Tokyo, Japan,

pp. 1235–1240.

HOEK, E.; MARINOS, P. G.; AND MARINOS, V. P., 2005, Characteriza-

tion 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 Engi-

neering 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., Minneapo-

lis, 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 Lithology Slope Type Geologic Age Formation or Group Name

ADA-32-12 Limestone underlain by claystone/mudstone

Interlayered competent/incompetent rock

Upper and Lower Silurian Peebles Dolomite

ADA-41-15 Limestone interlayered with claystone/mudstone


Lower Silurian Drowning Creek Formation

ATH-33-14 Sandstone Mostly competent rock Upper Pennsylvanian Conemaugh GroupATH-50-22 Red claystone/mudstone (redbeds)

interlayered with limestoneInterlayered competent/

incompetent rockUpper Pennsylvanian Conemaugh Group

BEL-470-6 Limestone and sandstoneinterlayered with green shale


Upper Pennsylvanian Monongahela Group

BEL-70-22 Sandstone interlayered with shale Interlayered competent/incompetent rock

Lower Permian/UpperPennsylvanian

Dunkard Group

BEL-7-10 Limestone and sandstoneinterlayered with green shale



CLA-4-8 Limestone Mostly competent rock Upper and Lower Silurian Cedarville, Springfield FormationCLA-68-6.9 Limestone Mostly competent rock Upper and Lower Silurian Cedarville, Springfield FormationCLE-275-5.2 Limestone interlayered with

claystone/mudstoneInterlayered competent/

incompetent rockUpper Ordovician Kope Formation

COL-7-5 Sandstone interlayered with shale Interlayered competent/incompetent rock

Middle/Lower Pennsylvanian Allegheny and Pottsville Groups

FRA-270-23 Shale Mostly incompetent rock Upper Devonian Ohio ShaleGUE-22-6.9 Sandstone interlayered with shale Interlayered competent/

incompetent rockMiddle/Lower Pennsylvanian Allegheny and Pottsville Groups

GUE-77-8.2 Sandstone underlain by coal withminor interlayers with siltstone/shale

Mostly competent rock Middle/Lower Pennsylvanian Allegheny and Pottsville Groups

HAM-74-6.4 Claystone/mudstone interlayeredwith minor limestone


Upper Ordovician Grant Lake Formation,Miamitown Formation,Fairview Formation

HAM-126-12 Claystone/mudstone interlayeredwith minor limestone


Upper Ordovician Grant Lake Formation,Miamitown Formation,Fairview Formation

JEF-CR77-0.38 Sandstone interlayered with shale Interlayered competent/incompetent rock

Upper Pennsylvanian Conemaugh Group

LAW52-11 Sandstone interlayered with shale Interlayered competent/incompetent rock


LAW-52-12 Sandstone interlayered with shale Interlayered competent/incompetent rock


LIC-16-28 Sandstone Mostly competent rock Lower Mississippian Black Hand Member of theCuyahoga Formation

Cut Slope Design


APPENDIX 2

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

Stratigraphic Type I II III IV

Slope Height (ft) 100 80 60 40 20 100 80 60 40 20 100 80 60 40 20 100 80 60 40 20

0.25H:1VSlope Angle

CatchmentDitch Slope

3H:1V 1 1 1 1 1 69 50 44 15 11 71 54 46 16 13 1 1 1 1 14H:1V 1 1 1 1 1 75 68 50 21 12 82 54 56 21 14 1 1 1 1 16H:1V 5 3 4 2 2 86 76 56 25 14 104 74 56 21 20 7 7 4 3 2

Flat 29 23 19 15 8 115 88 66 27 16 140 109 90 41 24 35 27 22 16 90.5H:1V

Slope AngleCatchment

Ditch Slope3H:1V 3 1 2 1 1 69 39 41 16 11 72 51 45 19 12 3 2 2 2 14H:1V 6 6 5 3 2 40 31 26 12 9 85 54 51 24 12 7 7 7 6 36H:1V 17 14 11 17 4 90 61 51 19 14 94 66 60 28 17 18 17 16 11 7

Flat 46 36 27 19 12 101 76 59 23 15 117 116 100 49 28 60 50 45 29 191H:1V



Flat 41 36 21 19 11 95 49 45 19 12 130 127 76 47 31 100 86 57 50 291.5H:1V



Flat 26 19 14 12 6 53 36 30 12 6 131 85 96 54 31 138 91 68 54 29

Site Lithology Slope Type Geologic Age Formation or Group Name

MEG-33-6 Red claystone/mudstoneinterlayered with sandstone



MEG-33-15 Red claystone/mudstoneinterlayered with sandstone


Lower Permian/UpperPennsylvanian

Dunkard Group

MUS-70-11 Sandstone interlayered with shale Interlayered competent/incompetent rock


RIC-30-12.5 Sandstone Mostly competent rock Upper and LowerMississippian

Logan and Cuyahoga Formations

STA-30-27 Shale with minor siltstone Mostly incompetent rock Middle/Lower Pennsylvanian Allegheny and Pottsville GroupsWAS-7-18 Red claystone/mudstone

interlayered with sandstoneInterlayered competent/

incompetent rockLower Permian/Upper

PennsylvanianDunkard Group

Appendix. 1. Continued.


Admassu and Shakoor

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, LeakageProblem, Grout Curtain

ABSTRACT

The Kowsar Dam is constructed on the Kheyrabad Riverat the northern limb of the Duck Anticline, close to its NWplunge, SW Iran. The dam is built on the karstic AsmariFormation, and the reservoir is in direct contact with thisformation from the dam body to the upstream impermeableGachsaran Formation. After reservoir impounding, severalnew springs emerged from the Asmari Formation, adjacentto the old small springs at the southern limb of the DuckAnticline. The discharge of these downstream springs wasnot reduced despite grout curtain treatment works. Themain water leakage route is not below or through the groutcurtain, as shown by considering rock permeability in pilotand check holes, cement consumption in grouting bore-holes, borehole water levels, spring locations, and dischargeof the dam galleries. The Duck Anticline is hydrogeologi-cally connected to the adjacent Dill and Pahn Anticlines,comprising the Asmari Karst Aquifer, which is divided intothree karst sub-aquifers. The general flow direction of karstwater in each sub-aquifer was determined using hydro-geological analysis and water balance calculations. Twoalternative models are proposed for water flow at thenorthern limb of the Duck Anticline. The main leakageroute toward the downstream springs is most probablythrough a relict karst conduit system, developed alongbedding planes of the Asmari Formation at the NW plungeof the anticline, according to the first proposed model. Inthis case, the water leakage can be significantly reduced byextension of the grout curtain further into the upstreamimpermeable Gachsaran Formation.

INTRODUCTION

Karst is defined as a terrain with distinctivelandforms and hydrology that is developed on

especially soluble rocks such as limestone, marble,dolomite, halite, and gypsum (Ford and Williams,2007). Karst covers approximately 20 percent ofEarth’s surface and is characterized by fluted andpitted rock surfaces, sinking streams, enclosed depres-sions, caves, sinkholes, shafts, springs, and subsurfacedrainage systems (White, 1988; Ford and Williams,2007). Karst aquifers are characterized by three typesof porosity: matrix (or inter-granular) porosity,fracture porosity, and conduits (White, 1969, 1977;Ford and Williams, 2007). Depending on whether theflow path is through matrix, fractures, conduits, ora combination, there are different types of flowsystems in karst aquifers (Martin and Screaton, 2001).Diffuse flow systems occur predominantly withinmatrix and fracture porosity, while conduit flowoccurs within conduits (Pitty, 1968; Paterson, 1979).Development of conduits within karst aquifers resultsin high permeability and allows rapid transfer ofgroundwater (Martin and Screaton, 2001). Due toa variety of intrinsic geological and hydrologicalfeatures, karst systems are among the most vulnerablesettings in the world for design and development ofengineering projects, including dams (Gutierrez et al.,2014; Parise et al., 2014).

Leakage from karst dam sites has been reported allover the world. The difficulties involved in construct-ing a dam on a karstified bedrock were firstdocumented at the Hales Bar Dam, which was builton the Tennessee River between 1905 and 1913(Donnelly et al., 2009). Milanovic (2004) reviewedin detail the leakage problems of several damsconstructed on karst regions all over the world; mostof them work successfully, but a few have problemswith unacceptably heavy leakage from the reservoir.In many dams, huge leakages were reduced aftercomplicated sealing programs. Leakage appears atpreexisting springs or as new resurgences. Leakageroutes are often karst conduits developed alongbedding planes, joints, and/or their intersections.The leakage rate can vary from a few liters to severalcubic meters per second. Even with extensive grout-ing, sometimes leakage is serious from the beginning

1Email address: [email protected] author email: [email protected].


of impoundment and may increase greatly with time(Palmer, 1988).

Most papers on the topic focus on remediationmethods. An overview of the articles shows that oneor several of the following techniques have been usedto identify the source(s) and route(s) of leakage:response of spring discharge and borehole water levelto reservoir water-level changes (Sahuquillo, 1985;Pantzartzis et al., 1993; Turkmen, 2003; Unal et al.,2007; Bonacci and Rubinic, 2009; and Bonacci andBonacci, 2013), tracing tests (Quinlan, 1985; Turkmenet al., 2002; and Mozafari et al., 2011), isotopicstudies (Hansen and Teter, 1970; Gunay et al., 1995;Crilley and Torak, 2002; and Laksiri, 2007), hydro-chemistry (Gunay et al., 1995; Qingzhi et al., 1998;Montoroi et al., 2001; Ghobadi et al., 2005; Torak etal., 2006; and Al-Omosh et al., 2008), exploratorydrilling, rock permeability, and grout curtain charac-terization (Zogovic, 1993; Jarvis, 2003; Turkmen,2003; Milanovic, 2004; Schaefer, 2009; Milanovicet al., 2010; and Mozafari et al., 2011), and geo-physical methods (Al-Saigh et al., 1993; Ginther andCharlton, 2009; and Bedrosian et al., 2012). Moham-madi et al. (2007) proposed three steps for leakagestudy at a karst dam site, including: (1) preparation ofthe hydrogeological map; (2) delineation and func-

tional analysis of the karst system by means ofstructure and functioning approaches; and (3) assess-ment of the leakage potential using the results of steps1 and 2 and determination of the most probableleakage zones.

The concrete gravity Kowsar Dam, with a height of144 m and a reservoir capacity of 580 million cubicmeters (MCM), was constructed by the IranianMinistry of Power in SW Iran (Figure 1) to annuallysupply 200 and 300 MCM of drinking and irrigationwater, respectively. The dam is built on the Kheyr-abad River, on the karstified Asmari Formation atthe northern limb of the Duck Anticline (Figure 1).Here, the Kheyrabad River, with an average dis-charge of 21.5 m3/s, flows almost perpendicular tothe anticline axis near NW plunge, developing theV-shape Kheyrabad Valley in the Asmari Formation.The reservoir normal water level (RNWL) is 625 mabove sea level (a.s.l.), about 125 m above theKheyrabad River floodplain. The reservoir is in directcontact with the Asmari Formation vertically fromthe riverbed to the RNWL, and on the sides from thedam body to the upstream impermeable GachsaranFormation. Soon after impounding, leakage occurredin the downstream preexisting springs and at severalnew springs, located on the Asmari Formation at the

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

Mozafari and Raeisi


southern limb of the Duck Anticline. The maximumtotal leakage stabilized at 5.61 m3/s at the RNWL.The main objective of this paper is to identify themain leakage route of the reservoir by means ofhydrogeological analysis of the Asmari Karst Aquifer(AKA), rock permeability tests in pilot and checkholes, cement consumptions in grouting boreholes,borehole water levels, springs locations, and dischargeof the dam galleries. The paper explains the role of ageneral hydrogeological study and evaluation of relictkarst development mechanisms in delineation of thereservoir leakage route. The results would be usefulto determine potential leakage routes in other karstdam sites in Iran and elsewhere.

HYDROGEOLOGICAL SETTING

The study area is located in the Simply FoldedZone of the Zagros Orogenic System in SW Iran,where the sedimentary rocks have been folded intoparallel NW-SE anticlines and synclines since theMiocene (Falcon, 1961). In the Simply Folded Zone,most of the karst formations are sandwiched betweentwo impervious formations, forming independentkarst aquifers (Raeisi, 2008). The study area iscomposed of five anticlines parallel to the generalstructure of the Simply Folded Zone, namely, theDuck, Dill, Pahn, Mish, and Khami Anticlines(Figure 1). The structural characteristics and stratig-raphy of the Zagros sedimentary sequence have beendescribed in detail by Stocklin and Setudehnia (1977)and Alavi (2004). The geological formations in thestudy area in decreasing order of age are (Figure 1):Khami Group carbonates and marls (Jurassic toLower Cretaceous), Bangestan Group carbonates andshale (Upper Cretaceous), Pabdeh-Gurpi shale andmarl (Cretaceous–Tertiary), Asmari Formation, mar-ly limestone and marl (Oligocene–Miocene), Gach-saran marl and evaporites (Tertiary), and Quaternaryalluvium. The distribution of the geological forma-tions in the Dill, Duck, Pahn, Mish, and KhamiAnticlines is presented in Figure 2. Due to action ofthrust on the Mountain Front Fault (MFF) (Fig-ure 1), the older geological formations of the Mishand Khami Anticlines are located next to the youngerformation of the Dill Anticline. The Khami Groupconstitutes the core of the Mish and Khami Anticli-nes, overlain by the exposed Bangestan Group. ThePabdeh-Gurpi Formation outcrops only at the MishAnticline and makes an important aquiclude unitbeneath the Asmari Formation in the core of theDuck, Dill, and Pahn Anticlines. The AsmariFormation outcrops at the top of the Dill, Duck,and Pahn Anticlines where the impermeable overlyingGachsaran Formation has been eroded, and it is

mostly exposed at the foot of the anticlines or buriedunder a thin alluvium on the adjacent plains. TheGachsaran Formation is divided into salt and non-salt equivalents (Bahrudi and Koyi, 2004). The SaltGachsaran Formation outcrops at the study area.Stocklin and Setudehnia (1977) divided it into sevenmembers based on a type section from wells in theGachsaran oil field. Member 1, which consists ofabout 40 m of inter-bedded anhydrite and limestoneassociated with shale, is known as an importantsealing unit over the Asmari Formation oil reservoirs.

The Kowsar Dam is built on the Asmari Formationat the northern limb of the Duck Anticline (Figure 3).Here, the Asmari Formation is classified into threemembers: Lower Asmari (LAs), Middle Asmari(MAs), and Upper Asmari (UAs) (Figure 3), basedon detailed geological mapping and borehole logs(Fars Regional Water Authority, 1997b). The LAsunit, which extends vertically from 70 m above and toat least 100 m below the valley floor, based onborehole log data, consists of 4- to 8-m-thick beds ofdense crystalline massive limestone with rare marlylimestone inter-beds. The MAs unit is composed ofmore than 100 m of 1.5- to 3-m-thick beds ofcrystalline limestone inter-bedded with marl and marlylimestone layers (marls represent about 7 to 8 percentof the MAs unit). A 1.5-m-thick marly layer is locatednear the lower boundary of the MAs unit (Figure 3).The UAs unit consists of 150 m of medium- to thin-bedded, crystalline limestone with relatively moreinter-bedded marly limestone and marls. Caves up to0.8 m were observed in the exposed LAs and MAsunits, filled partly with clay materials. In addition,a few open caverns were discovered in the AsmariFormation during borehole drilling and galleryexcavation. At the dam axis, the lithologic beds dipvery gently from the left abutment to the rightabutment, but in the right abutment, they havea sleeper slope. In addition, by moving from the damaxis toward the reservoir, lithologic beds dip upstream.

Before the dam construction, the groundwater levelmeasured in the six pilot holes along the dam axis waslocated about 6 to 10 m below the riverbed (Figure 4).There was no spring at the northern limb of the DuckAnticline, and just a few small springs emerged onboth the banks of the Kheyrabad River, from theMAs unit in the southern limb of the anticline, about1.2 km downstream of the dam body (Figure 3).There, a near-vertical minor fault (F1 fault inFigure 3) extended from the springs toward theUAs unit. Unfortunately, there are no data fordischarge of the springs before the dam construction,but a few measurements indicated that in addition tothe springs, there was about 0.5 m3/s direct seepage

Understanding Karst Leakage, Kowsar Dam, Iran


from the AKA into the riverbed at the southern limbof the anticline.

Grout Curtain

The water-tightness system of the Kowsar Damincludes a hanging grout curtain in the LAs and lowerparts of the MAs units at both abutments (Figures 3and 5). The hanging grout curtain was designed usingthe results of a FILTER software model based onmeasured rock permeabilities obtained from the eightpilot boreholes (Gidrosproekt, 1996) (Figure 4). Themodel predicted seepage less than 0.4 m3/s at theRNWL, by considering a 65- to 150-m-deep groutcurtain with 280 and 230 m extensions into the rightand left abutments, respectively. The as-built groutcurtain is 100 m longer in the right abutment and30 m deeper than the pre-built model (Tuzhikhin andKolichko, 2001). The grout curtain was constructedin one, two, or three rows with a distance interval of

1.7 m and by a three-level gallery system. The finalgrouting borehole spacing was in average 1.5 m, but itwas reduced at the zones with high cement consump-tion. The diameter of the grouting boreholes was0.076 m, and grouting was carried out in 5-m-longsections. A two-level drainage gallery system wasdesigned and built parallel and 25 m downstream ofthe grout curtain. The water transport gallery wasdesigned and built at the right abutment.

METHODOLOGY

The geologic map of the dam site (Figure 1) isbased on 1:100,000 (National Iranian Oil CompanyExploration and Production [NIOC], 1959) and1:5,000 (Fars Regional Water Authority, 1997a)geologic maps. Reservoir and borehole water levels,hydrochemistry, and discharge of springs weremeasured by the Fars Regional Water Authority(2010). Reservoir and boreholes water levels were

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

Mozafari and Raeisi


measured monthly during a 7 year period. Thedischarge of every spring was measured individuallyby volumetric and weir methods for about 2 years.Leakage from the dam galleries was measured usingweirs monthly for 7 years. A hydrometric station wasconstructed at the end of Kheyrabad valley, down-stream of the springs (Figure 3). Here, the total

discharge of springs and dam galleries, and the AKAdirect seepage into the riverbed were measured whenall of the dam outlets were closed. The water electricalconductivity (EC) was measured monthly in thereservoir and largest springs on both banks of theKheyrabad River during the spring to summer 2009.The following equation was used to determine the

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

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



mean annual recharge of each aquifer during onehydrological year (Q) in m3/s:

Q 5 API/t, (1)

where A is the aquifer area (m2), P is the 20 year meanannual rainfall (m), I is the recharge coefficient, and tis the 1 year period in seconds. The mean rainfall overthe aquifer surface was calculated based on therelationship between the elevations and 20 year meanrainfall from adjacent climatologic stations.

ANALYSIS

The AKA includes the Asmari Formation at theDuck, Dill and Pahn Anticlines, which is underlain bythe impermeable Pabdeh-Gurpi Formation, sur-rounded by the impermeable Gachsaran Formationand thin alluvial deposits; it is limited to the NE bythe MFF thrust fault (Figures 1 and 2). The AKAand the Bangestan Group are separated by the 800-m-thick impermeable Pabdeh-Gurpi Formation(Figure 2A and B). The impermeable Member 1 ofthe Gachsaran Formation prevents any hydrauliccommunication between the AKA and the adjacentMish and Khami Anticlines. The Gachsaran Forma-tion is covered by alluvial deposits in some parts

around the AKA. With the exception of the Lishteralluvium at the southwest of the Pahn Anticline,which was exploited by more than twenty 30- to 75-m-deep pumping wells (Figure 1), the other overlyingalluvium deposits are thin. The water level in theAKA is most probably lower than the contact of theGachsaran Formation and those alluvium deposits(Figure 2); therefore, karst water cannot flow intothem. In addition, since alluvial aquifers are oftenexploited by pumping wells in Iran, the lack ofpumping wells in those alluviums suggests that thereis no water recharge from the AKA.

The AKA recharge source is direct rainfall on thekarst aquifer body. There is no recharge from thesurrounding impermeable Gachsaran Formation oralluvium because those are located at lower elevationsthan the AKA outcrops (Figure 2). Recharge fromthe southern limbs of the Mish and Khami Anticlinesto the AKA is unlikely because they are structurallyand hydraulically disconnected by the impermeableGachsaran Formation and fractured zone of theMFF thrust fault (Figure 2A and B). In addition, theinfiltrated water of the southern limb of the MishAnticline discharges into the downstream Emamza-deh Jafar Aquifer, located about 10 km to the SE ofthe study area (Sharifi, 2009). The AKA waterdischarge zones are the Kheyrabad River and the

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

Mozafari and Raeisi


Lishter alluvium, controlling general flow direction inthe karst aquifer. The proposed discharge zones areconfirmed based on the following reasons: (1) Thereare no springs at the contact of the AKA and thesurrounding Gachsaran Formation and alluviums. (2)Before the dam construction, there were a few springson both banks of the Kheyrabad River at thesouthern limb of the Duck Anticline. In addition,a few measurements indicated that there was about0.5 m3/s of direct seepage from the AKA into theKheyrabad River at the southern limb of the DuckAnticline. (3) The underlying impermeable Pabdeh-Gurpi Formation prevents any downward water flowfrom the AKA into the Bangestan Group. (4) Thesurrounding impermeable Gachsaran Formation andthe fractured zone of the thrust MFF prevents anywater flow from the AKA to the adjacent anticlines.(5) The contact elevations of the Asmari Formationand the Gachsaran Formation are located about 150to 800 m above the riverbed all around the AKA,except at the Kheyrabad Valley inlet and outlet. Thetransverse Kheyrabad River is the main base level oferosion and the main discharge zone of the AKAbecause it has the lowest elevation within the AKA.

Based on the hydrogeological setting and thecatchment areas of the discharge zones, the AKA isdivided into three sub-aquifer areas (Figure 1): thesouthern limb of the Pahn Anticline sub-aquifer (SPsub-aquifer), which discharges into the LishterAquifer; the northern limbs of the Dill and DuckAnticlines sub-aquifer (NDD sub-aquifer), whichdischarges into the Kheyrabad River; and thenorthern limb of the Pahn and southern limb of theDill and Duck Anticlines sub-aquifer (NPSDD sub-aquifer), which discharges into the Kheyrabad River.Since there is no spring discharge measurement andtherefore no exact data on discharge of the AKAbefore the dam construction, the discharge of theAKA sub-aquifers during one hydrological year wasestimated based on Eq. 1 and using the estimatedrecharge coefficient according to previous studies onthe karst regions of Iran (Pezeshkpour, 1991; KarstResearch Centre of Iran, 1993; Karimi, 2003; andKarimi et al., 2005) (Table 1).

A conceptual model of the general flow direction isproposed for the three AKA sub-aquifers (Figure 1).

The elevation of the impermeable Pabdeh-GurpiFormation under the crest of the Pahn and DillAnticlines is higher than the contact of the AsmariFormation and the surrounding Gachsaran Forma-tion or alluvium at the foot of anticlines (Figure 2B),disconnecting the hydraulic connectivity of bothlimbs. Therefore, every limb becomes an independentsub-aquifer with catchment area limited to the crest ofthe anticline. In each sub-aquifer, the water flowsinitially along bedding planes until it reaches the footof the limb, where it flows parallel to the strike of thefolds. This hypothetical general flow direction alongthe fold strike has been demonstrated by Ashjari andRaeisi (2006) in most of the Zagros anticlines.

In the SP sub-aquifer, water initially flows alongbedding planes and finally discharges into theadjacent Lishter Aquifer (Figure 1). The LishterAquifer annual recharge by rainfall is about 4MCM, while its annual discharge by pumping wellsis 9.1 MCM (Table 1). Therefore, the missing 5.1MCM/yr must be supplied by inflow from theadjacent SP sub-aquifer, because: (1) the Lishteralluvium is surrounded by the impermeable Gach-saran Formation in all directions except adjacent tothe Pahn Anticline; and (2) the Pahn Anticline is thenearest karstic limestone to the Lishter Aquifer, andits elevation is higher than the Lishter alluvium.Based on Eq. 1, the required karst catchment area forproviding 5.1 MCM/yr water is about 16.5 km2. Thesouthern limb of the Pahn Anticline (SP sub-aquifer)is a feasible catchment area, since its exposed area isabout 25 km2.

The contact elevation of the Pabdeh-Gurpi For-mation with the Asmari Formation in the SE plungeof the Dill Anticline is about 650 m higher than in theNW plunge of the Duck Anticline. Therefore, at theNPSDD and NND sub-aquifers, a main conduitsystem has probably developed parallel to the foldstrike, starting from the SE plunge of the DillAnticline and going toward the NW plunge of theDuck Anticline. The NPSDD sub-aquifer collectsgroundwater coming from the northern limb of thePahn Anticline and southern limb of the DillAnticline. This karst water flows along the anticlinestrike toward the southern limb of the Duck Anticlineand finally emerges into the Kheyrabad River

Table 1. Annual recharge of the AKA sub-aquifers and Lishter Aquifer, calculated based on Eq. 1.

Aquifer orSub-Aquifer Area (km2)

RechargeCoefficient

Recharge byRainfall (MCM)

Discharge by Springs,Seepage, or Wells (MCM)

SP 25 0.4 6.3 —NPSDD 69 0.4 21.2 17NDD 49 0.4 15.3 46Lishter1 36 0.2 4 9.1



through the springs on the left bank or direct seepageinto the riverbed. In the NDD sub-aquifer, ground-water flows along the foot of the northern limb of theDill Anticline toward the northern limb of the DuckAnticline. Two alternative schematic models areproposed for the water flow route at the northernlimb of the Duck Anticline (Figure 6). In the firstmodel, water flows through a conduit system de-veloped most probably along bedding planes at theNW plunge and finally emerges into the KheyrabadRiver as the right bank springs. The NDD sub-aquifer water may be flowing below the KheyrabadRiver and emerging at the right bank springs, since thegroundwater level in the dam site was 6 to 10 m lowerthan riverbed before dam construction. The fact thatthere were no springs in the Kheyrabad Valley at thenorthern limb supports this model. Oberlander (1965)analyzed the origin of drainages and streams at theZagros Range in detail and showed a model of Zagrosstream superposition. The Kheyrabad River was mostprobably superimposed upon the newly exposed crestof the Duck Anticline, eroding it and leading to theformation of the Kheyrabad valley. Once riverincision started, and bearing in mind that the riverbedat the southern limb was lower than at the northernlimb, part of the river water may have seeped into thejoints and bedding planes to develop a new karstconduit system in the NW plunge, ending at the rightbank springs. As the river continued cutting down into

the bedrock and water base level continued to drop,new conduits formed, others enlarged, and the highestrelict conduits were dewatered and abandoned. Theimpermeable Gachsaran Formation prevents emer-gence of karst water from the plunge nose. At thepresent time, the elevation of the downstream springsis at least 140 m below the contact elevation of theAsmari Formation and the Gachsaran Formation atthe NW plunge nose of the Duck Anticline. Twocavities with a height of 3 m were observed in boreholeTD9 (Figure 3) at elevations of 508 and 561 m a.s.l.,probably confirming the relict conduit development inthe proposed model route. In the alternative model,the karst water flows perpendicular to the DuckAnticline axis, parallel to the Kheyrabad River andemerges through the downstream springs (Figure 6).Hydraulic connection between the northern andsouthern limbs of the Duck Anticline is possible sincethe elevation of the downstream springs is higher thanthe contact of the Asmari Formation and the Pabdeh-Gurpi Formation at the anticline axis (Figure 2C).This second flow route model seems to be unlikelysince water should flow perpendicular to the limestonebedding planes, and marly inter-beds up to 1.5 m thick(Figure 3) may create an impermeable barrier andprevent conduit development perpendicular to thebedding plane. In addition, some of the thick non-karstified limestone layers of the LAs unit maycreate an impermeable barrier to water and prevent

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

Mozafari and Raeisi


conduit development perpendicular to bedding plane,an idea supported by the lack of spring or seepage atthe LAs unit.

Leakage History and Remedial Efforts

Reservoir impounding started in November 2002.By the time that the reservoir water level reached546.1 m a.s.l. in February 2003, the discharge of thedownstream springs located in the southern limb ofthe Duck Anticline started to increase, and severalnew springs emerged adjacent to them. No new springor seepage was observed in the northern limb ofthe anticline, and water just emerged from some ofthe dam galleries, especially from those located at theright abutment. As explained already, the totaldischarge of the downstream springs, dam galleries,and the AKA direct seepage into the riverbed wasmeasured at the hydrometric station when all of thedam outlets were closed. In addition, the discharge of

every spring was measured individually until the endof 2004. Figure 7A presents the time series of the totaldischarge of the hydrometric station, right banksprings, left bank springs, dam galleries, and theAKA direct seepage into the riverbed until the end of2004. The AKA direct seepage into the riverbed wasestimated by subtracting the discharges of the springsand dam galleries from the discharge of the hydro-metric station. Results show that the discharges of thesprings and dam galleries correlated with the reservoirwater-level changes, but the AKA direct seepage intothe riverbed was most affected by the seasonalvariations. At the reservoir water level of 613 ma.s.l, the discharge of the hydrometric station was 3.4m3/s, and the apportioning of the right bank springs,left bank springs, galleries, and the AKA directseepage was 46, 20, 15, and 19 percent, respectively.The time series of the reservoir water level and thedischarge of the hydrometric station for a period of10 years are presented in Figure 7B. The figure shows

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



a tight relationship between the reservoir water leveland the changes in the discharge of the hydrometricstation. The discharge of the hydrometric stationreached 5.61 m3/s, more than one fourth of the meanannual discharge of the Kheyrabad River (21.5 m3/s),at the RNWL in 2005.

In order to reduce reservoir water leakage, water-tightness treatments were applied by constructinga new 90-m-long and 100-m-deep grout curtain alongthe transport gallery at the right abutment andadditional grouting in some parts of the as-builtgrout curtain (Figure 5). The treatment works weresuccessful in reducing nearly 60 percent of dischargefrom the dam galleries (Fars Regional Water Au-thority, 2010), but they did not affect the discharge ofthe downstream springs. The discharge of thehydrometric station was about 4 m3/s at a reservoirwater 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 thedownstream springs, mainly through those oneslocated on the right bank of the Kheyrabad River.The potential leakage route could be below and/orthrough the grout curtain, or through a relict conduitsystem at the NW plunge, based on the first proposedmodel of the general flow direction in the northernlimb of the Duck Anticline. The grouting workquality was evaluated based on the rock permeabilityin the pilot and check holes (Figures 4 and 8) andcement consumption in different stages of grouting.Before grouting, rock permeability was more than5 Lu in 92 percent of the measured sections in thepilot holes, and it was even more than 100 Lu in 13percent of sections. The permeability was reduced

significantly by grouting, since after grouting, it wasless than 5 Lu, 5 to 10 Lu, and 10 to 18 Lu in 76, 21,and 3 percent of the measured sections of the checkholes, respectively. The total cement consumption inthe grouted sections is presented on Figure 9. Thecement consumption decreased by stages of grouting;the average cement take was more than 1000 kg/m atthe first stage, but it was decreased to less than 40 kg/mat the last stage. In spite of high permeability beforegrouting, the grouted sections in the middle part of theleft abutment had low total cement consumptions(Figure 9) due to transfer of the grouting mix duringgrouting of the upper sections. The grout curtainworks properly, based on the borehole water levels,discharge of the dam galleries, and location of thesprings. The difference between boreholes water levelsupstream and downstream of the grout curtain was atleast 60 m at the RNWL (Figure 10). The linearregression coefficient between the reservoir water leveland borehole water levels upstream of the groutcurtain was 0.99, while at the downstream, it was lessthan 0.75, and mainly less than 0.74 at the left andright banks, respectively. The total discharge of thedam galleries was about 0.5 m3/s at the RNWL, whichis not far from the predicted leakage by the FILTERmodeling (0.4 m3/s). The Kheyrabad River is parallelto the dam axis about 300 m downstream of the groutcurtain (Figure 3), but no spring or seepage wasobserved on the river banks at this area. The damoutlets were closed several times, and the main leakagezone was at the downstream springs area. Theseevidences confirm that the main leakage route is notbelow or through the grout curtain.

The main leakage route is most probably froma relict conduit system that developed along beddingplanes of the Asmari Formation at the NW plunge ofthe Duck Anticline (Figure 3). The reservoir is in

Figure 8. Rock permeabilities in the check holes.

Mozafari and Raeisi


direct contact with the LAs, Mas, and UAs units.However, since the hydraulic relation between thereservoir and the underlying LAs unit is disconnectedby the grout curtain (Figures 3 and 5), reservoir watercan leak into the MAs and UAs units and then flowthrough the relict conduit system and finally emergeas the right bank springs. The karst water in the UAsand MAs units could be directed toward the springsby the F1 fault.

Borehole P38 is located 330 m from the reservoir, inthe UAs and MAs units and inside the proposedleakage route (Figure 3). The water level of thisborehole has a similar response to reservoir water-level variations (Figure 10). The linear regression

coefficient between the water levels of the reservoirand borehole P38 is 0.99, indicating a tight correla-tion and therefore a strong hydraulic connectionbetween the reservoir and the proposed leakage route.The borehole P38 water level was 12.8 m lower thanRNWL, and the reservoir water flowed toward it witha hydraulic gradient of 4 percent.

The water EC of the largest spring on the rightbank was about 80 to 115 ms/cm higher than thereservoir during March to July 2009, when thereservoir water level was near 595 m a.s.l. (Figure 11).The EC of the reservoir water must be increased bymoving through the proposed leakage route towardthe downstream springs.

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

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



The total discharge of the right bank springs was 46MCM during 2004 (Table 1), which is much morethan the expected annual discharge of the NDD sub-aquifer (15.3 MCM). Even considering that all NDDsub-aquifer water emerged at the right bank springsafter reservoir impoundment, the reservoir waterleakage through the right bank springs is calculatedto be 30.7 MCM.

The total annual discharge of the left bank springswas 16.6 MCM during 2004 (Table 1). Twoassumptions can be proposed for the water sourceof these springs. Based on the first assumption, thespring water is completely supplied by the karstwater of the NPSDD sub-aquifer area. The expectedannual discharge of the NPSDD sub-aquifer is 21.2MCM, i.e., 4.6 MCM higher than the annualdischarge of the left bank springs (16.6 MCM),which can be seeping directly into the riverbed. Thelower water EC of the largest spring on the left bankcompared to the reservoir (Figure 11) supports thisassumption. Based on the second assumption, thesource of the springs could be mixing of the NPSDDsub-aquifer karst water and the reservoir water.At the springs area, the limestone layers are repeatedat both the banks of the Kheyrabad River, sincethey are eroded almost perpendicular to the beddingplanes by the Kheyrabad River. Some reservoirwater could be transferred from the right bank tothe left bank of the river by a probable conduit

system in the limestone layers below the riverbed.The crossing of a karst conduit below a riverbedwithout any hydrogeological relationship has beenproved by a dye tracing test at the Doosti Dam siteby Mozafari et al. (2011). The transferred reservoirwater could be mixed with the water of the NPSDDsub-aquifer and emerge as the left bank springs. TheEC of the transferred reservoir water can be reducedby mixing with the karst water of the NPSDD sub-aquifer.

CONCLUSIONS

The reservoir water leakage route toward thedownstream springs is not below or through thegrout curtain. The proper function of the groutcurtain is evidenced by the evaluation of rockpermeability in the pilot and check holes, cementconsumptions of the grouting boreholes, boreholewater levels, springs locations, and discharge ofthe dam galleries. Understanding the hydrogeologyof the AKA acts as an effective tool to determinewater leakage route toward the downstream springs.The hydraulic relation between the AKA andadjacent aquifers, except the Lishter Aquifer, isdisconnected from the underlying impermeable Pab-deh-Gurpi Formation, surrounding impermeableGachsaran Formation, and function and fracturedzone of the MFF thrust fault. The AKA recharge

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

Mozafari and Raeisi


source is direct rainfall on the karst aquifer, andthere is no recharge from the surrounding alluvium,impermeable Gachsaran Formation, and adjacentMish and Khami Anticlines. The AKA consists ofthree sub-aquifer areas, SP, NPSDD, and NDD.While the SP sub-aquifer area is discharged into theLishter alluvial aquifer, the Kheyrabad River is themain discharge zone of the NPSDD and NDD sub-aquifers, controlling the general flow direction inthese sub-aquifers. The proposed conceptual model ofgeneral flow direction in the three AKA sub-aquifersis in agreement with the general flow direction in mostof the anticlines of the Zagros Mountain Range inIran. A relict conduit system has likely been de-veloped along bedding planes in the NW plunge ofthe Duck Anticline by karst water flow of the NDDsub-aquifer and action of the Kheyrabad River. Thereservoir water leakage route is mainly through thisrelict conduit system. By raising the reservoir waterlevel, water seeps into the Asmari Formation at thenorthern limb, converges into the relict conduit systemin the MAs and UAs units, flows along the plunge ofthe anticline, and finally emerges through the rightbank springs at the southern limb. The as-built groutcurtain and the treatment works in the right abutmentwere not able to disconnect the hydraulic relationbetween the reservoir and the envisioned relict conduitsystem.

The hydrometric station mean annual discharge isabout 95.6 MCM. Considering 37.5 MCM annualdischarge of the NDD and NPSDD sub-aquifers, theannual water leakage of the reservoir is calculated tobe 58.1 MCM. This volume of water is about 20 and30 percent of the proposed annual irrigation anddrinking water, respectively. The irrigation diversioncanal is a few kilometers downstream of the springs;therefore, the leakage water is used for agriculturalpurposes. The Kowsar Dam is located in an arid andsemiarid part of Iran, and it is the most reliable sourceof drinking water for the downstream arid areas.Because of low precipitation in the Kheyrabad Rivercatchment area, provision of proposed irrigation andeven drinking water could be associated with anuncertainty during drought periods. Therefore, pre-vention of the current reservoir leakage seems to bereasonable. Construction of a new grout curtainperpendicular to the bedding planes in the upper partsof the MAs and UAs units and toward the upstreamimpermeable Gachsaran Formation at the rightabutment would be effective in reducing of the waterleakage amount. Dye tracing test is recommended fordetection of specific leakage route and subsequentlythe correct path for the probable additional groutcurtain.

ACKNOWLEDGMENTS

The authors gratefully appreciate the sincere co-operation of the Fars Regional Water Authority ofIran for providing useful data. The authors alsowould like to thank Mr. Ahmadi for information andDr. J.P. Galve, Dr. J. Guerrero, and Prof. F.Gutierrez for their reviews and comments.

REFERENCES

ALAVI, M., 2004, Regional stratigraphy of the Zagros Folds-Thrust Belt of Iran and its proforeland evolution: AmericanJournal of Science, Vol. 304, pp. 1–20.

AL-OMOSH, H.; AL FARAJAT, M.; AND ZUNIC, F., 2008, Leakage inBayer 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-potentialmethod: Engineering Geology, Vol. 37, No. 2, pp. 115–121.

ASHJARI, J. AND RAEISI, E., 2006, Influences of anticlinal structureon regional flow, Zagros, Iran: Cave and Karst Science,Vol. 68, No. 3, pp. 118–129.

BAHRUDI, A. AND KOYI, H. A., 2004, Tectonic-sedimentary frame-work of the Gachsaran Formation in the Zagros forelandbasin: 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, Geophysicalinvestigations of geology and structure at the Martis CreekDam, Truckee, California: Journal of Applied Geophysics,Vol. 77, pp. 7–20.

BONACCI, O. AND RUBINIC J., 2009, Water losses from a reservoirbuilt 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 negativeconsequences of underground dam and reservoir construc-tion and operation in coastal karst areas: An example of theHydro-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 hydro-chemical evidence of lake leakage and assessment of karstfeatures in the vicinity of Lake Seminole, southwesternGeorgia and northwestern Florida. In American GeophysicalUnion Fall Meeting Abstracts: American GeophysicalUnion, Washington, DC, abstract H72C-0864.

DONNELLY, C. R.; HINCHBERGER, S.; AND MOHAMMADIAN, E., 2009,The design of foundation treatment measures for dams onkarst foundations: Canadian Dam Association Bulletin,Vol. 20, No. 4, pp. 20–27.

FALCON, N. L., 1961, Major earth-flexing in the Zagros Mountainsof southwest Iran: Quaternary Journal of Geological Societyof London, Vol. 117, pp. 367–376.

FARS REGIONAL WATER AUTHORITY, 1997a, Water Supply Projectof Boushehr, Geological Map of Kowsar Dam Site: FarsRegional Water Authority, Shiraz, Iran, scale: 1:5,000, 1sheet.

FARS REGIONAL WATER AUTHORITY, 1997b, Water Supply Projectof Boushehr, Engineering Geology of Kowsar Dam Site: FarsRegional Water Authority, Shiraz, Iran [Report in Persian],275 p.

FARS REGIONAL WATER AUTHORITY, 2010, Kowsar Dam SiteCondition Reports: Fars Regional Water Authority, Shiraz,Iran [Report in Persian], 98 p.



FORD, D. C. AND WILLIAMS, P. W., 2007, Karst Geomorphology andHydrology: 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 ShahidAbbaspour Dam, southern Iran: Engineering Geology,Vol. 82, pp. 119–126.

GIDROSPROEKT, 1996, Model Studies of Seepage and CutoffArrangement: Tange Duck Water Development Report, Iran,Kowsar Dam: Gidrosproekt, Tehran, Iran 1913-NSI-1-15702.

GINTHER, C. H. AND CHARLTON, J. C., 2009, Comprehensivefoundation rehabilitation at Bear Creek Dam. In USSD 29thAnnual Conference: United States Society on Dams, Nash-ville, TN, 4/20/09.

GUNAY, G.; EKMEKCI, G. M.; AND TEZCAN, L., 1995, Hydro-geochemical and isotopic evaluation of the hydrogeologicalsystem at Ceyhan-Berke Dam site: Application of tracers inarid zone hydrology. In Proceedings of the Vienna Sympo-sium, August 1994: International Association of HydrologicalSciences Publication 232, pp. 395–408.

GUTIERREZ, F.; PARISE, M.; DE WAELE, J.; AND JOURDE, H., 2014,A review on natural and human-induced geohazards andimpacts in karst: Earth-Science Reviews, Vol. 138, pp. 61–88.

HANSEN, R. L. AND TETER, G. A., 1970, Use of Radioisotopes inTracing Reservoir Leakage at Anchor Dam: U.S. Departmentof the Interior, Bureau of Reclamation, Office of ChiefEngineer, 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 inthe United States: Oklahoma Geological Survey Circular 109,pp. 271–278.

KARIMI, H., 2003, Hydrogeological Behavior of the Alvand BasinKarst Aquifers, Kermanshah: Ph.D. thesis, Shiraz, Universityof Shiraz, 281 p.

KARIMI, H.; RAEISI, E.; AND ZARE, M., 2005, Physicochemicalparameters time series of karst spring as a tool to differentiatethe source of spring water: Carbonates and Evaporites,Vol. 20, No. 2, pp. 138–147.

KARST RESEARCH CENTRE OF IRAN, 1993, Comprehensive Study andResearch in Water Resource of the Maharlu Basin, Fars: KarstResearch Centre of Iran Report, Volumes 1–4.

LAKSIRI, L. B. K., 2007, Investigation of Water LeakageMechanism in the Karstic Dam Site, Samanalawewa, Srilanka:Ph.D. thesis, Department of Engineering Systems andTechnology, Graduate School of Science and Engineering,Saga University, Saga, Japan, 136 p.

MARTIN, J. B. AND SCREATON, E. J., 2001, Exchange of Matrix andConduit Water with Examples from the Floridan Aquifer: U.S.Geological Survey Karst Interest Group Proceedings, WaterResources 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 lossesrisk assessment: An example from Carpathian karst: Envi-ronmental Earth Sciences, Vol. 60, pp. 817–827.

MOHAMMADI, Z.; RAEISI, E.; AND BAKALOWICZ, M., 2007, Methodof leakage study at the karst dam site. A case study:Khersan 3 Dam, Iran: Environmental Geology, Vol. 52,pp. 1053–1065.

MONTOROI, J. P.; GRUNBERGER, O.; AND NASRI, S., 2001,Hydrochemical behaviour of waters in a Tunisian hillreservoir watershed and reservoir impact on alluvial aquifer.In: HYDROMED: seminaire international les petits barragesdans le monde mediterraneen: recueil des resumes. Paris(FRA); Tunis: IRD; INRGREF, 2001, 2 p. Multigr. Les.

Petits Barrages dans le Monde Mediterraneen: SeminaireInternational, Tunis (TUN), 2001/05/28–31.

MOZAFARI, M.; RAEISI, E.; AND ZARE, M., 2011, Water leakagepaths in the Doosti Dam, Turkmenistan and Iran: Journal ofEnvironmental 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 inPersian): NIOC, Tehran, Iran scale: 1:250,000, 1 sheet.

OBERLANDER, T., 1965, The Zagros Streams: A New Interpretationof Transverse Drainage in an Orogenic Zone: SyracuseGeographical Series No. 1, Syracuse University Press, NewYork, 168 p.

PALMER, A. N., 1988, Solutional enlargement of opening in thevicinity of hydraulic structures in karst regions. In Proceed-ings of the 2nd Conference on Environmental Problems inKarst Terranes and Their Solutions: Association of Ground-water 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, Proceedingsof the Antalya Symposium and Field Seminar: InternationalAssociation 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 Geologyfor Society and Territory. Volume 5—Urban Geology,Sustainable Planning and Landscape Exploitation. SpringerInternational Publishing, Switzerland, 2015, pp. 479–482.

PATERSON, K., 1979, Limestone springs in the OxfordshireScarplands: The significance of spatial and temporal varia-tions in their chemistry: Zeitschrift fur Geomorphologie N.F.,Vol. 32, pp. 46–66.

PEZESHKPOUR, P., 1991, Hydrogeological and HydrochemicalEvaluation 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 relationto 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 character-istics and seepage of regional karst water: Geological Scienceand Technology Information, Vol. 17, No. 2, pp. 9–14.

QUINLAN, J. F., 1985, Application of dye-tracing to dam-siteevaluation in a Kentucky karst area, U.S.A. In Proceedings ofthe International Symposium on Karst Water Resources:Ankara, Turkey, July 1985: International Association ofHydrological Sciences Publication 161, pp. 533–544.

RAEISI, E., 2008, Groundwater storage calculation in karstaquifers with alluvium or no-flow boundaries: Journal ofCave and Karst Studies, Vol. 63, pp. 62–70.

SAHUQUILLO, A., 1985, Spanish experience in karst water resources.In Proceedings of the International Symposium on KarstWater Resources, Ankara, Turkey, July 1985: InternationalAssociation of Hydrological Sciences Publication 161,pp. 133–147.

SCHAEFER, J. A., 2009, Risk evaluation of dams on karstfoundations. In Proceedings of the 29th Annual USSDConference, April 20–24, 2009: United States Society onDams, Nashville, TN, pp. 541–579.

SHARIFI, M., 2009, Evaluation of Groundwater Contamination inthe Emamzadehjafar Aquifer, Gachsaran City: M.Sc. thesis,University of Shiraz, Shiraz, Iran, 298 p.

STOCKLIN, J. AND SETUDEHNIA, A., 1977, Stratigraphic Lexicon ofIran: Geological Survey of Iran Report 18–1971, 370 p.

Mozafari and Raeisi


TORAK, L. J.; CRILLEY, D. M.; AND PAINTER, J. A., 2006, Physical andHydrochemical Evidence of Lake Leakage near Jim WoodruffLock and Dam and Ground-Water Inflow to Lake Seminole andan Assessment of Karst Features in and near the Lake,Southwestern Georgia and Northwestern Florida: U.S. Geo-logical Survey Scientific Investigations Report 2005-5084, 89 p.

TURKMEN, S., 2003, Treatment of the seepage problems at the Kale-cik 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 founda-tion of the Kalecik Dam (south Turkey): EngineeringGeology, Vol. 63, pp. 147–157.

TUZHIKHIN, G. G. AND KOLICHKO, A. V., 2001, Cementationoperations 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 ofleakage 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 Niksic

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.



Documents

Environmental & Engineering Geoscience