Environmental & Engineering Geoscience

Environmental & Engineering Geoscience AUGUST 2016

VOLUME XXII, NUMBER 3

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

Environmental & Engineering Geoscience (ISSN 1078-7275) is published quarterly by the Association of Environmental & Engineering Geologists (AEG) and the Geological Society of America (GSA).

Periodicals postage paid at AEG, 1100 Brandywine Blvd, Suite H, Zanesville, OH 43701-7303 and additional mailing offices.

EDITORIAL OFFICE: Environmental & Engineering Geoscience journal, Department of Geology, Kent State University, Kent, OH 44242, U.S.A. phone: 330-672-2968, fax: 330-672-7949, ashakoor@kent.edu.

CLAIMS: Claims for damaged or not received issues will be honored for 6 months from date of publication. AEG members should contact AEG, 1100 Brandywine Blvd, Suite H, Zanesville, OH 43701-7303. Phone: 844-331-7867. GSA members who are not members of AEG should contact the GSA Member Service center. All claims must be submitted in writing.

POSTMASTER: Send address changes to AEG, 1100 Brandywine Blvd, Suite H, Zanesville, OH 43701-7303. Phone: 844331-7867. Include both old and new addresses, with ZIP code.

Canada agreement number PM40063731. Return undeliverable Canadian addresses to Station A P.O. Box 54, Windsor, ON N9A 6J5 Email: returnsil@imexpb.com.

DISCLAIMER NOTICE: Authors alone are responsible for views expressed in articles. Advertisers and their agencies are solely responsible for the content of all advertisements printed and also assume responsibility for any claims arising therefrom against the publisher. AEG and Environmental & Engineering Geoscience reserve the right to reject any advertising copy.

SUBSCRIPTIONS: Member subscriptions: AEG members automatically receive digital access to the journal as part of their AEG membership dues. Members may order print subscriptions for $60 per year.

GSA members who are not members of AEG may order for $60 per year on their annual GSA dues statement or by contacting GSA. Nonmember subscriptions are $295 and may be ordered from the subscription department of either organization. A postage differential of $10 may apply to nonmember subscribers outside the United States, Canada, and Pan America. Contact AEG at 844-331-7867; contact GSA Subscription Services, Geological Society of America, P.O. Box 9140, Boulder, CO 80301.

Single copies are $75.00 each. Requests for single copies should be sent to AEG, 1100 Brandywine Blvd, Suite H, Zanesville, OH 43701-7303. © 2016 by the Association of Environmental and Engineering Geologists All rights reserved. No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or by any information storage and retrieval system, without permission in writing from AEG.

EDITORIAL BOARD ROBERT H. SYDNOR JEROME V. DEGRAFF USDA Forest Service Consulant THOMAS J. BURBEY CHESTER F. WATTS (SKIP) Virginia Polytechnic Institute Radford University SYED E. HASAN University of Missouri, Kansas City ASSOCIATE EDITORS JOHN W. BELL PAUL M. SANTI Nevada Bureau of Mines and Colorado School of Mines Geology ROBERT L. SCHUSTER U.S. Geological Survey RICHARD E. JACKSON (Book Reviews Editor) ROY J. SHLEMON R. J. Shlemon Geofirma Engineering, Ltd. & Associates, Inc. JEFFREY R. KEATON AMEC Americas GREG M. STOCK National Park Service PAUL G. MARINOS National Technical University RESAT ULUSAY Hacettepe University, Turkey of Athens, Greece CHESTER F. “SKIP” WATTS JUNE E. MIRECKI U.S. Army Corps of Radford University Engineers TERRY R. WEST Purdue University PETER PEHME Waterloo Geophysics, Inc NICHOLAS PINTER Southern Illinois University SUBMISSION OF MANUSCRIPTS Environmental & Engineering Geoscience (E&EG), is a quarterly journal devoted to the publication of original papers that are of potential interest to hydrogeologists, environmental and engineering geologists, and geological engineers working in site selection, feasibility studies, investigations, design or construction of civil engineering projects or in waste management, groundwater, and related environmental fields. All papers are peer reviewed.

The editors invite contributions concerning all aspects of environmental and engineering geology and related disciplines.

Recent abstracts can be viewed under “Archive” at the web site, “http://eeg.geoscienceworld.org”. Articles that report on research, case histories and new methods, and book reviews are welcome. Discussion papers, which are critiques of printed articles and are technical in nature, may be published with replies from the original author(s). Discussion papers and replies should be concise.

To submit a manuscript go to http://eeg.allentrack.net. If you have not used the system before, follow the link at the bottom of the page that says New users should register for an account.

Choose your own login and password. Further instructions will be available upon logging into the system. Please carefully read the “Instructions for Authors”.

Authors do not pay any charge for color figures that are essential to the manuscript. Manuscripts of fewer than 10 pages may be published as Technical Notes.

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

THIS PUBLICATION IS PRINTED ON ACID-FREE PAPER EDITORS ABDUL SHAKOOR Department of Geology Kent State University Kent, OH 44242 330-672-2968 ashakoor@kent.edu

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

Cover photo A panoramic view of the 2011 Cedar Canyon landslide. See article on page 245.

Photo courtesy of William Lund, Utah Geological Survey.

Environmental & Engineering Geoscience Volume 22, Number 3, August 2016 Table of Contents

173

A Comparison of Three Hydraulic Parameter Optimization Schemes for Las Vegas Valley Meijing Zhang and Thomas J. Burbey

195

Tracers of Groundwater Mixing in the Hueco Bolson Aquifer, Ciudad Jua´rez, Mexico Christopher J. Eastoe, Alfredo Granados Olivas, and Barry J. Hibbs

209

Landscape Change as Recorded by the Ocean Shore Railroad Matthew A. Thomas and Keith Loague

225

Landslide Risk Reduction in the United States—Signs of Progress Jerome V. De Graff, William J. Burns, and Vicki McConnell

245

Characterization of Failure Parameters and Preliminary Slope Stability Analysis of the Cedar Canyon Landslide, Iron County, Utah Ashley Tizzano, Abdul Shakoor, and William R. Lund

259

Investigating Movement and Characteristics of a Frozen Debris Lobe, South-Central Brooks Range, Alaska Jocelyn M. Simpson, Margaret M. Darrow, Scott L. Huang, Ronald P. Daanen, and Trent D. Hubbard

A Comparison of Three Hydraulic Parameter Optimization Schemes for Las Vegas Valley MEIJING ZHANG1 Department of Geosciences, Virginia Tech, Blacksburg, VA 24061 Email: zhangmj@vt.edu Address: A207 Levine Science Research Center, Duke University, Durham, NC 27708-0328

THOMAS J. BURBEY Department of Geosciences, Virginia Tech, 4044 Derring Hall (0420), Blacksburg, VA 24061 Email: tjburbey@vt.edu

Key Terms: Inverse Parameterization, Transmissivities, Elastic and Inelastic Skeletal Storage Coefficients, The Adaptive Multi-Scale Method, DREAM MCMC, Subsidence

ABSTRACT Las Vegas Valley has had a long history of groundwater development and subsequent surface deformation. Much research has been done to estimate parameters within the Las Vegas Basin, but this research represents the first effort to use parameter estimation techniques to inversely calibrate hydrogeologic aquifer parameters for the principal aquifer of the entire basin.

Three different inversion strategies are invoked to determine the most accurate and computationally efficient method for estimating transmissivities T and elastic and inelastic skeletal storage coefficients (Ske and Skv) at the basin scale: the zonation method (ZM), the adaptive multi-scale method, and the differential evolution adaptive metropolis Markov chain Monte Carlo scheme (DREAM MCMC). The three inversion methods are compared and contrasted based on quantitative measurements of model fit, computational efficiency, and user flexibility. The results indicate that overall, the adaptive multi-scale method, which is able to efficiently reconstruct the T, Ske, and Skv zones while providing more flexibility and accuracy than the other two methods, is the best strategy for calibrating optimal model parameters and providing a framework for developing an accurate hydrogeologic model for Las Vegas Valley.

1 Corresponding author phone: 540-231-2404; Fax: +1-540-231-3386; email: zhangmj@vt.edu.

INTRODUCTION Inverse modeling involving history matching of model outputs to field measurements has been widely used in the field of hydrogeology for several decades (Hill and Tiedeman, 2007) and has largely superseded trial-and-error methods. The objective of inverse modeling is to calibrate optimal model parameters, and then models can be used to generate credible predictions. Parameters such as transmissivity, hydraulic conductivity, storage coefficient, aquitard leakage, and porosity are often important parameters that are calibrated using inverse models in groundwater investigations. Although groundwater levels are the most popular type of observational data used to calibrate groundwater models, they alone are usually insufficient to obtain adequate and unique results (Hill, 1998; Hill and Tiedeman, 2007; Yan and Burbey, 2008; and Zhang et al., 2013). Other measurement data types that are useful for calibrating a model may include fluid concentration values, flow rates, recharge rates, and land subsidence.

The inversion approach includes both deterministic and stochastic estimation. Deterministic estimation aims to find a single optimal parameter using available observed data. One widely used deterministic estimation approach is the zonation method (ZM), which involves dividing the entire study area into a number of zones, while unknown parameters are treated as uniform over each zone. This approach is commonly used in popular groundwater inversion codes such as UCODE_2005 (Poeter et al., 2005). However, when using the ZM method, the shape and the number of the zones have to be defined by the user, which is somewhat subjective. The multi-scale method is an approach that provides criteria to discretize the zones.

In multi-scale parameterization, the parameter estimation problem is solved through successive approximations by refining the zone domains during the inverse

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 173–194

173

Zhang and Burbey

procedure. When the zone refinement is complete, the refinement no longer induces a significant decrease of the objective function (Chavent and Liu, 1989; Liu, 1993; Ameur et al., 2002; Hayek and Ackerer, 2007; and Hayek et al., 2008). However during the refinement process, the number of degrees of freedom is increased, which can lead to over-parameterization if they exceed the number of available observations (Ameur et al., 2002). To overcome this drawback, Ameur et al. (2002) developed an algorithm that uses both refinement and coarsening indicators to decide whether addition or removal of some degrees of freedom is advantageous.

Another important inversion approach applied in groundwater model calibration is the pilot points method (PiPM) (De Marsily et al., 1984; Lavenue and Marsily, 2001; and Doherty, 2003), which involves the perturbation of hydraulic properties at a small number of selected “pilot point” locations in an effort to better match observational data (Alcolea et al., 2006). The PiPM is used in the parameter estimation software PEST (Doherty, 2004). The PiPM is designed to represent the hydrogeologic heterogeneity under either deterministic or stochastic conditions, while kriging interpolation is used to generate the spatial distribution of hydraulic properties from estimated values at the pilot points. One inherent weakness in this method is that the number and location of pilot points are somewhat subjective, and a considerable degree of nonuniqueness still exists. Stochastic estimation techniques such as simulated annealing (Rao, 2003), genetic algorithms (Prasad, 2001), and the Markov chain Monte Carlo (MCMC) method (Fu and Jaime Gómez-Hernández, 2009; Vrugt et al., 2009) rely on randomness and re-trials to estimate the probability distributions of the unknown parameters and to sample the parameter space in searching for an optimal solution. The deterministic estimation may become “trapped” in local minima if the initial guess is far from the optimal solution, while the stochastic estimation techniques allow for the estimation to jump out of the local minimum. With the development of high-performance computing and high-throughput computing technology, new technologies, such as Grid computing (Foster et al., 2001), makes the computationally long running of stochastic calculations tractable (Renard and De Marsily, 1997); however difficulties still exist, such as defining how many iterations are required to obtain an accurate statistic distribution (Ballio and Guadagnini, 2004; Renard, 2007). Among the adaptive MCMC sampling approaches, the differential evolution adaptive metropolis (DREAM) MCMC scheme significantly improves the efficiency of MCMC simulations (Vrugt et al., 2009).

174

In this work, we aim to investigate the advantages and disadvantages of three different inverse strategies applied to a groundwater flow model at the basin scale: (1) the ZM, (2) the adaptive multi-scale method, and (3) DREAM MCMC. We apply each of these methods to estimate the distribution of transmissivity (T) and elastic and inelastic skeletal storage coefficients (Ske and Skv, respectively) from observations of hydraulic head and land subsidence measurements for Las Vegas Valley between 1912 and 2010. A secondary objective is to find the best strategy for evaluating the model parameters for making credible predictions, and to provide a framework for developing an accurate hydrogeologic model for Las Vegas Valley.

BACKGROUND Las Vegas Valley has had a long history of groundwater development (Harrill, 1976) and subsequent surface deformation (Bell, 1981; Bell et al., 2002). The complex hydrogeological conditions and large amount of available temporal and spatial groundwater hydraulic head and land subsidence measurements since the 1930s make the Las Vegas Valley area an ideal site to conduct inverse parameterization research. Harrill (1976) was the first to systematically estimate basinwide transmissivities from aquifer test data. Morgan and Dettinger (1994) developed the first valley-wide numerical groundwater model, which covered the period 1912–1981, using the Trescott, Pinder, and Larson groundwater model (Trescott et al., 1976). In their model, pre-calibration transmissivity was estimated using regression analysis from pumping tests and lithologic data interpreted by Plume (1984). The final transmissivities were manually calibrated by adjusting transmissivities and storage values so that the difference between simulated heads and observed heads was minimized as much as possible through trialand-error methods and a minimum grid spacing of 2.5 km. Mean inelastic specific storage values were calculated based on groundwater level declines, clay thicknesses, and subsidence data measured at four benchmarks. The model yielded a rather generalized distribution of simulated drawdown and land subsidence, but it did not capture the likely complex patterns of subsidence we know exist today and likely existed at the time this model was developed. Jeng (1998) later converted the Morgan and Dettinger model to the MODFLOW model (McDonald and Harbaugh, 1984) without changing any final simulated parameters from Morgan and Dettinger. Yan (2007) later updated Jeng’s model by extending the simulation period to 2005 using mostly the original parameter estimates of Morgan and Dettinger as well as the same coarse grid spacing. Pavelko (2004)

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 173–194

Three Hydraulic Parameter Optimization Schemes for Las Vegas Valley

Figure 1. (a) Generalized surficial geologic map of Las Vegas Valley showing distribution of coarse- and fine-grained deposits, and principal Quaternary faults and fissures. (b) Geologic crosssection (A-A9) as modified from Bell et al. (2002) and schematically illustration of the stratigraphic and fault relations interpreted from well log data (Bell et al., 2002). (c) The conceptual model layer.

developed a one-dimensional groundwater and subsidence model in conjunction with UCODE (Poeter and Hill, 1998) to calibrate hydraulic parameters at the Lorenzi site, where the U.S. Geological Survey

(USGS) installed an extensometer in 1994. Hoffmann et al. (2001) and Bell et al. (2008) used interferometric synthetic aperture radar (InSAR) and persistent scatterer (PS) InSAR combined with groundwater level declines to evaluate elastic and inelastic skeletal storage coefficients in Las Vegas Valley at several individual locations. Until now, no research has been done on calibration of the transmissivities and elastic and inelastic skeletal storage coefficients of Las Vegas Valley at the basin scale.

In order to compensate for the deficiencies of these earlier models, we invoke three different inversion strategies to determine the best method for calibrating transmissivities and elastic and inelastic skeletal storage coefficients at the basin scale with the aim of achieving both accuracy and computational efficiency. We have successfully extended the Las Vegas Valley model to year 2010 using MODFLOW-2005 (Harbaugh, 2005) with the subsidence (SUB) package (Hoffmann et al., 2003) while implementing a 600 m 6 600 m grid cell size (Figure 1a). Observations include available hydraulic head data and land subsidence measurements (from 1992 to 2010). However, due to the immense size of the model and large number of cells used, the estimation of parameters for each of these three strategies with the large number of observations (time and space) used is extremely time-consuming (especially with the accelerating DREAM MCMC method). In order to find the optimal inversion strategy for the Las Vegas Valley model, a shorter-term model is used in this study that extends from 1912 to 1987, with the period from 1912 to 1981 used as the calibration period and the period from 1981 to 1987 representing the evaluation period.

The shorter-term model is used in this investigation due to the time-dependent delayed drainage of finegrained clay inter-beds, which play an important role in the complicated development and distribution of land subsidence in Las Vegas Valley. The inter-bed is assumed to be areally far more extensive than its thickness, and the hydraulic conductivity of the inter-bed is considerably lower than the aquifer. The time constant, which represents the time required for the inter-bed to reach 93 percent of its compaction for a given head change, ranges from approximately 100 years to perhaps more than 1,000 years for Las Vegas Valley (Pavelko, 2004). Thus, any model that is to adequately simulate subsidence and hydraulic heads must coincide with the commencement of pumping in the basin. On the other hand, the subsidence rates of recent years are largely influenced by delayed drainage, making it difficult for inverse modeling techniques to accurately calibrate parameters (Yan, 2007). Therefore, a shorter simulation period is used in this

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 173–194

175

Zhang and Burbey

investigation that commences with the start of pumping in the year 1912, before the onset of subsidence in the valley. It is anticipated that this time period is sufficient for investigating the advantages and disadvantages of the various inverse methods previously described as applied to the basin scale.

FIELD SITE AND FORWARD MODEL

exist near the central part of the valley. The maximum total subsidence in 1980 was in the northwest bowl, which was measured to be 78 cm (Figure 1a). For the period 1963–1987, the three principal localized subsidence bowls became more widespread and extensive.

Observed subsidence of more than 1.5 m had occurred in the northwest bowl by 1987 (Figure 1a) (Bell et al., 2002).

Las Vegas Valley Study Area

Forward Model

Las Vegas Valley encompasses an area of about 4,150 km2 in Clark County, southern Nevada. The northwesttrending valley is bounded on all sides by various mountains. It is located in a structurally controlled alluvial basin. Series of north- to northeast-trending, east-dipping Quaternary faults cut the valley floor (Figure 1a).

A thick accumulation of inter-bedded and inter-fingered coarse- and fine-grained sediment fills the structural basin. In the northern, western, and southern parts of the valley, sediment consists of mainly coarse-grained sand and gravel. However, sediment becomes thicker and finer grained toward the central and eastern parts of the valley (Figure 1a and b). The principal aquifer, composed mainly of coarse-grained deposits with some fine-grained low-permeable inter-beds, occurs at depths ranging from 60 to 300 m below land surface (Malmberg, 1965; Harrill, 1976; and Morgan and Dettinger, 1994). In the vertical direction, permeability tends to decrease with depth (Maxey and Jameson, 1948). Overlying the principal aquifer, there is a 30-mto 90-m-thick sequence of clay, sand, and gravel, which is often referred to as the near-surface reservoir, and which is not pumped nor used as a potable water source.

Groundwater has been pumped since 1905 in the Las Vegas Valley area. Due to declining water levels, decreasing pore-water pressures within the aquifer system have led to significant increases in effective stress, which account for large-scale compaction of mostly fine-grained sediments (Terzaghi, 1925; Poland and Davis, 1969; Poland et al., 1972; and Helm, 1975).

Because pumping has exceeded natural recharge for many years now, intensive groundwater pumping in Las Vegas has led to highly varying degrees of land subsidence, which have resulted in surface fissures and socioeconomic problems (Bell, 1981). According to benchmark surveys, Bell plotted an early subsidence map for the period 1935–1963 (Bell et al., 2002), which shows that subsidence occurred as a singular bowl located near downtown Las Vegas (central bowl), with a maximum subsidence during that period of 67 cm. For the ensuing period 1963–1980, an updated subsidence map shows that three localized subsidence bowls—the northwest, central and southern bowls—

We updated Yan’s (2007) model by extending the simulation period to 2010 using MODFLOW-2005 (Harbaugh, 2005) with the subsidence (SUB) package (Hoffmann et al., 2003) while implementing a 600 m 6 600 m grid cell size, by far the finest-scale model developed for the valley to date. In total, 41 stress periods are used to simulate the flow and subsidence in the new model, which extends from March 1912 to October 1987. The new conceptual model consists of four model layers: Layers 1 and 2 represent the nearsurface aquifers; layer 3 represents the developedzone aquifer; and layer 4 represents the deep-zone aquifer (Figure 1c). All four layers are assumed to be confined aquifers. Nearly all of the groundwater supply in the valley comes from the developed-zone (or principal) aquifer, which is 200–300 m below the land surface (Maxey and Jameson, 1948; Morgan and Dettinger, 1996). Only the developed-zone (or principal) aquifer is pumped for groundwater.

Groundwater levels under pre-development conditions approximated by Harrill (1976) are treated as initial groundwater levels in the developed-zone aquifer (Harrill, 1976). Initial groundwater levels for the other layers are the same as those used by Morgan and Dettinger in their model (Morgan and Dettinger, 1996). The well package is used to simulate groundwater pumping and natural recharge that occurs along the western and northern boundaries of the model. All the other boundaries are assumed to be no-flow boundaries. Secondary recharge is simulated with the recharge package.

The groundwater flow equation of our model is given as:

176

Ss b

X dhij @ha Kvj ¼ r ðTrha Þ þ W; @t dz j

(1)

– where ha is simulated hydraulic head of the aquifer, h i is hydraulic head in the inter-bed, b is thickness of the aquifer system, T is transmissivity, Ss is specific storage of the aquifer, Kv is vertical conductivity of the aquifer, and W is a source term.

The inter-bed is assumed to be areally far more extensive than its thickness, and the hydraulic

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 173–194

Three Hydraulic Parameter Optimization Schemes for Las Vegas Valley

conductivity of the inter-bed is considerably lower than that of the aquifer, so the direction of groundwater flow within the inter-bed can be treated as vertical. Groundwater flow from the inter-bed to the aquifer occurs when the head in the aquifer declines, with the head change in the inter-bed lens lagging that of the aquifer and described by the diffusion equation i

Sk

@h @ @ i ¼ ðKvj hj Þ; @t @z @z

(2)

where Sk is skeletal storage coefficient of the inter-bed.

The SUB package (Hoffmann et al., 2003) was used to calculate subsidence at each model cell from a single inter-bed using the equation: "

# 1 expð p2 t Þ 8 X 4 sk sðtÞ ¼ Sk Dh 1 2 ; p k¼0 ð2k þ 1Þ2 i

(3)

sk ¼

K 0 v ð2k þ 1Þ2

;

(4)

where s is inter-bed compaction (land subsidence), t is time, b0/2 is one-half the thickness of the inter-bed, Ssk is the skeletal specific storage of the inter-bed, and K9v is the vertical hydraulic conductivity of the inter-bed. The time constant τ0 (where k 5 0 for Eq. 4) represents the time required for the inter-bed to reach 93 percent of its compaction for a given head change, Δh.

The land subsidence of the entire system can be calculated by multiplying the value from Eq. 3 by the factor nequiv.

N X 1 nequiv ¼ sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi bi : N P 2 i¼1 1 bi N

Sk ¼ Ssk b0 :

(6)

The updated model we developed uses all the parameters from Yan (2007). However, simulation results show that the simulated drawdown and subsidence do not fit well with the observational data; hence, an inverse model, which is designed to calibrate the parameters, is required to accurately describe and predict the drawdown and subsidence patterns in Las Vegas Valley.

INVERSION MODEL AND METHOD Inversion Model, Observations, Parameters, and Weighting Simulation Period

with τk defined as ðb20 Þ2 Ssk

between skeletal storage coefficient and skeletal specific storage is given as:

(5)

i¼1

Laboratory consolidation tests indicate that the skeletal specific storage, which describes the compressibility, can vary greatly depending on whether the effective stress exceeds the previous maximum effective stress, which is termed as the pre-consolidation stress (Johnson et al., 1968; Jorgensen, 1980). Inelastic skeletal specific storage, Sskv, is used when the water level in the inter-bed is below its previous minimum value, whereas elastic skeletal specific storage, Sske, is invoked when the water level in the inter-bed is above the previous minimum values. The relationship

A shorter model covering the period 1912–1987 is used in this investigation. The shorter model still covers nearly the entire pumping and subsidence history.

The actual calibration period used here covers the period from 1912 to 1981, while the test evaluation period is from 1982 to 1987.

Observations Selected measurements (including 357 hydraulic head values covering from year 1939 to 1981 and 113 land subsidence measurements collected from year 1963 to 1980) from the calibration period are treated as observational data in the analysis. The evaluation period is from 1982 to 1987, during which 4,000 hydraulic head values and 26 land subsidence measurements (from year 1987) were collected. The locations of the data are shown in Figure 2. All the land subsidence measurements are taken from Figure 4 of Bell and Amelung’s paper (Bell et al., 2002).

Observational Weighting For the ZM and adaptive multi-scale strategy, the weightings used in our analyses are proportional to the inverse of the square of the observational values.

The precise definition of weights generally is not required for observations. Rather, it is more important that the sensitivity analysis results and parameter estimates be reasonably consistent over the plausible range (Foglia et al., 2009). It has been shown that choosing the weight as being proportional to the inverse of the square of the observational value is a good way to obtain accurate estimates of hydraulic parameters (Yan, 2007).

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 173–194

177

Zhang and Burbey

ZM, (2) the adaptive multi-scale method, and (3) DREAM MCMC. The ZM method and the adaptive multi-scale method are deterministic estimations aimed at finding a single optimal parameter using available observed data. The DREAM MCMC is a stochastic estimation method based on using random parameter values and running re-trials on these parameters to estimate their probability distributions.

Zonation Method (ZM) The first inversion strategy applied in this study was the ZM. Transmissivities (T) and elastic and inelastic skeletal storage coefficients (Ske and Skv) of layer 3 were calibrated at the basin scale. UCODE-2005 (Poeter et al., 2005) was used to minimize the weighted least-squares objective function defined as a leastsquares misfit to the data with respect to the parameter values using a modified Gauss-Newton method. The weighted least-squares objective function can be written as: JðaÞ ¼

ND X i¼1

Figure 2. Groundwater and land subsidence monitoring network used as observations in the model (red color points represent observations for the period 1912–1980).

For the DREAM method, the weighting is a diagonal matrix with the weight for each observation equal to 1/σi2, where σi2 is the variance of the associated error.

Parameters In this investigation, we are interested in calibrating transmissivities (T) and elastic and inelastic skeletal storage coefficients (Ske and Skv) of layer 3 (the developed-zone aquifer) in Las Vegas Valley. The principal aquifer is capable of transmitting significant quantities of groundwater (Maxey and Jameson, 1948; Malmberg, 1965; Harrill, 1976; and Morgan and Dettinger, 1994) and contributing to virtually all of the observed land subsidence. Seventy-two T parameters, and six Ske and Skv parameters were considered in Yan’s model (Yan, 2007), which were treated as initial estimates for both zones and parameters of layer 3 in our study. The zones and parameters of the other layers were kept the same as Yan’s model (Yan, 2007).

Inversion Method Three different inverse strategies were applied to a groundwater flow model at the basin scale: (1) the

178

xi ½yobsðiÞ yi ðaÞ 2 þ

NPR X

xp ½PpriorðpÞ Pp ðaÞ 2 ;

p¼1

(7) where a ¼ ðT; Ske ; Skv Þ;

(8)

and this represents the parameter vector to be optimized, where ND is the number of observations, NPR is the number of prior information values, ωi is the weight for the ith observation, ωp is the weight for the pth prior estimate, yobs(i) represents the ith observation, yi(a) represents the simulated value that corresponds to the ith observation, Pprior(p) represents the pth prior estimate, and Pp(a) represents the pth simulated value.

Prior information was used for parameters T and Ske and Skv on the basis of the distribution map of T and Ske and Skv provided by Morgan and Dettinger (Morgan and Dettinger, 1994). The parameters were log transformed. We assumed that the standard deviation is 0.349, which was used to weight the prior information. The approximate reasonable range of values for the parameter is between one fifth to five times the values provided by Morgan and Dettinger (1994) with a 95 percent probability. The convergence criterion was set to 0.01, which means that the regression converges if the relative change of the objective function is less than 0.01 for three sequential iterations or the fractional change for all parameters is less than 0.01 for all parameters.

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 173–194

Three Hydraulic Parameter Optimization Schemes for Las Vegas Valley

parameterization a (Figure 3a). After minimization during the first parameterization a0, the gradient of the objective function J0* becomes

0¼

Figure 3. (a) Zone before refinement. (b) Zone after a cutting refinement.

Composite scaled sensitivities (CSS) are used to measure the overall sensitivity of the observations to the parameters. If parameters have CSS that are less than about 0.01 times the largest CSS, it is likely that the regression will not converge (Hill, 1998), so when minimizing the objective function using UCODE2005, we chose to omit those parameters that had CSS lower than 0.01. The parameter non-uniqueness can be detected using parameter correlation coefficients (PCC). Parameters with PCC values larger than 0.85 were omitted due to their non-uniqueness (Foglia et al., 2009).

Adaptive Multi-Scale Strategy The second inversion strategy applied in this study was the adaptive multi-scale algorithm, which provides criteria to reconstruct the zones by refining or coarsening the current zones. T, Ske, and Skv of layer 3 were calibrated at the basin scale.

In this strategy, we invoked the same objective function (Eq. 7) for minimization. UCODE-2005 (Poeter et al., 2005) was used to minimize the weighted leastsquares objective using a modified Gauss-Newton method. Both CSS and PPC analyses were conducted as previously described. The convergence criterion was the same as previously described.

Refinement Indicators.—In order to obtain more realistic zonations based on actual hydrogeological conditions, the parameter estimation problem was solved through a series of successive approximations by refining the zone domains during the inverse procedure.

The zone refinement was considered complete when it no longer induced a significant decrease of the objective function (Chavent and Liu, 1989; Liu, 1993; Ameur et al., 2002; Hayek and Ackerer, 2007; and Hayek et al., 2008). For the sake of simplicity, we assumed that only one parameter needed to be identified for each zone. The situation where additional parameters are used will be discussed later. Note that a0* and J0* represent the optimal parameter and objective function, respectively, corresponding to

Nz @Jða 0 Þ X @Jða 0 Þ ;

@a @ai i¼1

(9)

where Nz is the number of sub-zones.

For our example, we split the single zone Z (Figure 3a) into two zones Z1 and Z2 (Figure 3b). A large number of tentative cuts can be tested (dashed lines C1, C2, C3, and C4 in Figure 3b). The refinement indicator associated with splitting the single zone Z (Figure 3a) into two zones (Figure 3b) is defined as: I0 ¼ j

@Jða 0 Þ @Jða 0 Þ j¼j j: @a1 @a2

(10)

For zones that need to be split, the indicators associated with a large number of tentative cuts can be calculated. It can be seen from Eq. 9 and Eq. 10 that indicators provide first-order information on the minimization of the optimal objective function, so it is recommended to select not only the highest indicator Imax, but also indicators for which values are greater than τ*Imax (Hayek and Ackerer, 2007; Hayek et al., 2008). Here, we used τ 5 0.85.

After minimizing the objective function for all selected cuttings (indicators that are more than τ*Imax), the partition with the smallest objective function is the optimal solution.

For the case where more than one parameter needs to be identified for each zone, we use the definition provided by Hayek (Hayek et al., 2008), where the corresponding refinement indicator is defined as: Ci ¼

NP X

Ik;i ; max i Ik;i k¼1

(11)

where Ik,i is the indicator corresponding to the parameter ak at splitting partition i, and NP is the number of parameters that need to be identified for each zone.

Coarsening Indicators.—During the refinement process, the number of degrees of freedom is increased, which can lead to over-parameterization if the number of parameters multiplied by the number of zones exceeds the number of available observations (Ameur et al., 2002). To overcome such a drawback, Ameur et al.

(2002) presented an algorithm that uses both refinement

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 173–194

179

Zhang and Burbey

Foglia et al., 2009). The algorithm for this analysis is described in the following steps:

Figure 4. (a) Zone before coarsening. (b) (c) Zone after coarsening.

and coarsening indicators to decide whether addition or removal of some degrees of freedom is advantageous.

Consider the situation shown in Figure 4, where there are two zones, an interior zone Z1 and an exterior zone Z2 (Figure 4a). After minimization using Eq. 7, using the modified Gauss-Newton method, optimal parameter values a*1 and a*2 are obtained for the interior zone Z1 and an exterior zone Z2. Then Z1 is refined using the scheme provided in the previous section. Suppose that the refinement corresponding to the cut (dashed line) shown in Figure 4a is selected. Before choosing this new zonation (zone Z1,1, zone Z1,2, and zone Z2), one must first make sure that zone Z1,1 or zone Z1,2 can be combined with zone Z2, which will assure that there will not be an increase in the number of degrees of freedom. According to Eq. 9, this leads to: @Jða 1 ; a 2 Þ @Jða 1 ; a 2 Þ þ ¼ 0: @a1;1 @a1;2

(12)

Multiplying Eq. 12 by (a*2 – a*1) yields: DJ1;1 þ DJ1;2 ¼ 0:

(13)

It can be seen from Eq. 13 that, in theory, at least one ΔJ1,i value is negative, which indicates that combining zones Z2 and Z1,j for which ΔJ1,j , 0 will lower the objective function. However, from eq. 9 we know that ΔJ1,j only provides a first-order estimate of the decrease in the optimal objective function.

Thus, the decrease of the objective function cannot be guaranteed, but this provides an alternative where the degrees of freedom can be unchanged during the zone reconstruction. The term ΔJ1,j is the coarsening indicator.

The Adaptive Multi-Scale Algorithm.—In this study, we provide an alternative to using zone refinement and coarsening indicators. This alternative involves first identifying those zones where hydrogeological analysis (such as geological sediment distribution maps and inter-bed thickness maps) indicates where refinement or coarsening may be required, and then identifying those zones that are important to fitting the observations through a sensitivity analysis (Grimstad et al., 2003;

180

1. Choose an initial parameter zonation Z distribution (which contains Nz sub-zones).

2. For the current zonation Z, minimize the objective function J using UCODE, and compute the optimal parameter values a*.

3. At this point, Z contains Nz sub-zones. However, based on the hydrogeological and sensitivity analysis, choose a new zone Zchosen that needs restructuring.

4. Choose a set of cuts that split the zone Zchosen.

Compute all the refinement indicators I (or Г for multi-dimensional refinement indicator) corresponding to the chosen set of cuts. Compute Imax (or Гmax), the largest value of all computed refinement indicators. Select all cuts for which refinement indicators are larger than 85 percent of Imax (or Гmax).

5. If Zchosen is totally inside another zone Zout (as shown in Figure 4), then the corresponding coarsening indicator. Combine the sub-domains where coarsening indicators allow it.

Else go directly to step 6.

6. Minimize the objective function J using UCODE for all selected discontinuities (refinement cuts or coarsening). Notice that the number of zones is now Nz_new 5 Nz + 1 for the refinement situation and Nz_new 5 Nz for the coarsening situation. The best parameterization a*new corresponds to the smallest calculated objective function Jnew.

7. If the updated objective function Jnew is sufficiently small such that a (near) perfect fit of the observed data occurs, or if the maximum number of zones is reached, or if the refinement and coarsening no longer induces a significant decrease of the objective function, then stop iteration.

Else J 5 Jnew, a* 5 a*new, Nz 5 Nz_new and return to step 3.

End if The DREAM MCMC Method The third inversion strategy involves the applica‐ tion of the DREAM MCMC method, which was explained by Vrugt et al. (Vrugt et al., 2009). DREAM is a widely used program for Bayesian uncertainty analysis. This scheme involves performing multiple diff‐ erent “chains” simultaneously in order to estimate the probability distributions of the unknown parameters.

Assuming multi-Gaussian distributed errors in the observations, the likelihood function for this approach is defined as (Lu et al., 2014):

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 173–194

Three Hydraulic Parameter Optimization Schemes for Las Vegas Valley

Lðajyobs Þ ¼ð2pÞ ND=2 jCj 1=2 1 expð ðyobsðiÞ yi ðaÞT C 1 ðyobsðiÞ yi ðaÞÞÞ 2 (14) where C is the covariance matrix of the measurement error. Measurement errors are taken to be uncorrelated. UCODE-2014 (Lu et al., 2014; Poeter et al., 2014), which invokes the DREAM MCMC method, is used for estimating parameters.

Although the DREAM MCMC scheme significantly improves the efficiency of the MCMC simulation (Vrugt et al., 2009), compared to the ZM and adaptive multi-scale methods, it requires significantly more computational time. In our study, UCODE-2014 takes 4.5 minutes of computational time to run one iteration with high performance computing (HPC) facilities at Virginia Tech. In general, DREAM MCMC takes thousands to hundreds of thousands of iterations to converge. Furthermore, additional iterations are required if more parameters are calibrated, as in our case. However, with the same HPC, it takes 70 minutes to run one iteration of the modified Gauss-Newton method, which is not as efficient as the DREAM MCMC, but for our case, this method converges within 50 iterations.

DREAM MCMC may not be practical for highdimensional models, because it requires expensive computational time (Keating et al., 2010). Therefore, we were only able to use the DREAM MCMC method to estimate posterior probability distributions of T at selected zones due to the large computational time requirements.

Measures Used for Evaluation of Model Fit For the ZM and the adaptive multi-scale method, the Nash-Sutcliffe efficiency index (NS) (Nash and Sutcliffe, 1970), mass balance error (m), the Schulz criterion (D) (Schulz et al., 1999), and sum of squared weighted residuals (WSSR) are used to evaluate the goodness and fit of the hydrologic model.

NS is defined as (Nash and Sutcliffe, 1970): n P

NS ¼ 1

n P i¼1

ðyi yobsðiÞ Þ2

i¼1

y2 obsðiÞ 1n ð

n P

;

(15)

yobsðiÞ 2

i¼1

where yobs(i) represents the ith observation, and yi represents the simulated value.

When the observations are perfectly fitted by the simulated values, NS 5 1; NS . 0.8 represents

excellent fitting; 0.6 , NS , 0.8 represents very good fitting; 0.4 , NS , 0.6 represents good fitting; and 0.2 , NS , 0.4 represents sufficient fitting; while NS , 0.2 represents insufficient fitting.

Another indicator, m, is used to evaluate whether the simulated values are generally less than or greater than the observed values, and it is calculated as: n P

m ¼ 100 i¼1

ðyi yobsðiÞ Þ n P

:

(16)

yobsðiÞ

i¼1

If the observations are perfectly fitted by the simulated values, m 5 0. When m . 0, this indicates that the simulated values are greater than the observed values on average, while for the case when m , 0, the simulated values are lower than the observed values on average.

Another fitting parameter, D, is calculated as (Schulz et al., 1999): n P

D ¼ 200

jyi yobsðiÞ jyobsðiÞ

i¼1

nðyobsðiÞ;max Þ2

;

(17)

where yobs(i),max represents the maximum of all the observations. When 0 , D , 3, the model fit is considered to be very good; when 3 , D , 10, the model fit is classified as good; when 10 , D , 18 the model fit is sufficient; and when D . 18, the model is fit is considered to be insufficient.

The standard error of regression s is a quantitative measure of overall model fit to the weighted observations (Hill, 1998; Hill and Tiedeman, 2007) and is calculated as: s¼ð

1 WSSR Þ =2 ; ND þ NPR p

(18)

where ND is the number of observations, NPR is the number of prior information values, p is the number of estimated parameters, and WSSR is the sum of squared weighted residuals calculated for the diagonal weight matrix and is calculated as: WSSR ¼

NDþNPR X

xi ½yobsðiÞ yi ðaÞ 2 :

(19)

i¼1

If the model fit is consistent with the data accuracy as reflected in the weighting, then s is 1.0.

For the DREAM MCMC strategy, the predictions were evaluated at the 95 percent credible interval using

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 173–194

181

Zhang and Burbey

Figure 5. (a) Transmissivity zones and values from Yan (2007). (b) Transmissivity zones and values calibrated with ZM. (c) Transmissivity zones and values calibrated with DREAM MCMC.

the equal-tailed method (Casella and Berger, 2002).

For each prediction, the samples were ordered from the smallest to the largest values, and the 95 percent credible interval was determined at the 2.5 percent and 97.5 percent percentiles of model predictions.

RESULTS Calibrated Transmissivity Zones and Values The ZM Method Transmissivities (T) of layer 3 ere calibrated at the basin scale with the ZM. Seventy-two T parameters described in Yan’s model (Yan, 2007) are shown in Figure 5a. Calibrated T values obtained from the ZM are shown in Figure 5b. The results indicate that most zones have similar T values to those from Yan (2007), except for zones E and S, located in the far eastern part of the basin, where relatively higher T values were calculated using the ZM (Figure 6a).

Coarser-grained alluvial deposits are known to occur near point A9 (Figure 1b) and in the southern part of the basin (Figure 1a), which may explain the higher simulated T values for zones E and S. In order to accurately describe how T is distributed in these zones, a much finer zone distribution based on hydrogeological conditions is required. Sensitivity analysis shows that all parameters have CSS values larger than 0.04 times the largest CSS (Figure 6a), which indicates that the model and the observed data provide sufficient information to estimate the parameters. The PCC analysis shows that no parameters have correlation coefficients

182

larger than 0.83, indicating that the estimated parameters are unique given the information provided by the observations.

The Adaptive Multi-Scale Strategy In order to obtain more realistic zonations based on actual hydrogeological conditions, we applied the adaptive multi-scale algorithm to reconstruct the T zones of layer 3 at the basin scale. The parameter estimation problem was solved through 13 iterations of successive approximations by refining the zone domains (especially zone E and zone S in the eastern part of the basin) during the inverse procedure. The final zone distributions and zone values are shown in Figure 5c. Most zones have similar T values to that from Yan (2007), except at zones E and S (Figure 6b). The sensitivity analysis shows that all parameters have CSS value larger than 0.03 times the largest CSS (Figure 6b), indicating that the model and the observed data provide sufficient information to estimate the parameters. A PCC analysis shows that no parameter has a correlation coefficient larger than 0.80, which indicates that the estimated parameters are unique given the information provided by the observations.

The DREAM MCMC Method Due to the significant computational time required for the global optimization method, we were only able to use this method to calibrate T at selected zones.

For our test, we calibrated T at zones a, b, c, and d

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 173–194

Three Hydraulic Parameter Optimization Schemes for Las Vegas Valley

Figure 6. A comparison of transmissivity values (a) from Yan (2007) and calibrated with ZM; (b) from Yan (2007) and calibrated with the adaptive multi-scale strategy.

(Figure 7). Uniform prior information (T ranges from 1 to 15,000 m2/d) was used for all parameters. In total, four Markov chains were used to generate candidate samples. Figure 8 illustrates the numerical evolution of the sampled T values at zones a, b, c, and d. The four chains converged to a limiting distribution within 400 evaluations. The convergence was then evaluated by the Gelman-Rubin values, which were shown to be continuously less than 1.2 after 400 iterations. The final T values from Yan (2007), the ZM and adaptive multi-scale methods are also shown in Figure 8 for comparison. T values calibrated with the ZM and the adaptive multi-scale method are within one order of magnitude of the values obtained with DREAM MCMC. This outcome is reasonable because T values were calibrated at only four selected zones with DREAM MCMC, but for the entire basin with the other two methods. With the sufficient parameter samples after chain convergence, the last 50 percent of the

samples in the four chains (0.5 6 4 6 9,000 5 18,000 samples) were used to construct histograms for each individual parameter (Figure 9). The T values at zones a and d have only one mode, which are Gaussian-like and exhibit negligible skewness, while the simulated T at zone b has one mode, which is Gaussian-like with positive skewness. The posterior distribution of T at zone c has two modes, which may indicate that the formal likelihood Gaussian function assumption may not hold. This is reasonable, because only the covariance matrix of the measurement error is considered in Eq. 14; however the Las Vegas Valley model is extremely non-linear and complex, with multiple sources of uncertainty including model structural error, input error, etc. The PCC between two parameters is shown in Figure 10. All the PCC values are smaller than 0.7, indicating that the parameters are weakly correlated.

This is consistent with the result obtained from the ZM method and the adaptive multi-scale method.

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 173–194

183

Zhang and Burbey

where initial estimates were largely unknown) during the inverse procedure. The final calibrated zone distributions and values are shown in Figure 11c. Sensitivity analysis shows that all parameters have CSS values larger than 0.18 times the largest CSS, with Skv exhibiting the largest CSS, which suggests this parameter is highly sensitive to observations and important in the calibration process. PCC analysis shows that no parameter has correlation coefficients larger than 0.83, indicating that the estimated parameters are unique and relatively unaffected by the model given the information provided by the observations.

Evaluation of Model Fit

Figure 7. Transmissivity zones calibrated with DREAM MCMC in test 1 and test 2 (zones with slash) and test 3 (zones with gray).

Calibrated Ske and Skv Zones and Values The parameters Ske and Skv of layer 3 were calibrated at the basin scale using the ZM and the multiscale strategy methods. Six pairs of Ske and Skv parameters described in Yan’s model (Yan, 2007) are shown in Figure 11a. The Ske and Skv parameters that Yan used are from Morgan and Dettinger (1994). The inter-bed thicknesses that were used to calculate Ske and Skv are inferred for many locations within the basin where data are limited, especially in the northwest part of the basin (see Figures 3.3.4-1 and 3.3.4-2 of Morgan and Dettinger, 1994); hence, in order to obtain a better model fit, reconstructing Ske and Skv zones and calibrating parameter values based on actual hydrogeological conditions are preferred methods. Calibrated Ske and Skv values using the ZM are shown in Figure 11b. The zones are the same as in Yan’s model. However, Ske and Skv values are lower than those calibrated by Yan. In order to obtain more realistic zonations based on actual hydrogeological conditions, the adaptive multi-scale algorithm was applied to reconstruct the Ske and Skv zones of layer 3. The parameter estimation problem was solved through four iterations of successive approximations by refining or coarsening the zone domains (especially in the northwest part of the basin,

184

We used the final T, Ske, and Skv values from the ZM and the adaptive multi-scale strategy methods to calculate simulated hydraulic heads and basin-wide land subsidence. For the DREAM MCMC strategy, the predictions were evaluated, and their 95 percent credible interval limits were determined using the equal-tailed method (Casella and Berger, 2002) for each of the 18,000 converged parameter samples (Figure 9).

Figure 12 shows plots that compare observed versus simulated hydraulic heads for the calibration period from the ZM and adaptive multi-scale strategies.

The results show that the ZM and adaptive multiscale methods produce a better model fit than Yan’s model. A quantitative assessment of the model fit is listed in Table 1. The NS criterion indicates that the fit is excellent with the ZM and the adaptive multiscale methods. The m criterion indicates that, on average, the simulated hydraulic heads are larger than the observed heads. The D criterion indicates that the fit is very good with both the ZM and adaptive multi-scale inversion strategies. The s criterion is shown to be decreasing with the ZM and the adaptive multi-scale methods, which represents an improved overall model fit to the weighted observational data. Based on the quantitative assessment results (Table 1), the adaptive multi-scale method is shown to be superior to the ZM inverse method. Figure 13 shows plots comparing observed versus simulated hydraulic heads at sites h1–h4 (Figure 2), where more than two observations were available for the calibration period. The results show that the adaptive multi-scale method produces the best model fit at sites h1–h4 (Figure 13).

Observed and simulated land subsidence values for the calibration period using the ZM and the adaptive multi-scale strategies are shown in Figure 14. It can be seen from Figure 14a and b that the ZM and the adaptive multi-scale methods fit the observed values more closely than Yan’s model, especially where the

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 173–194

Three Hydraulic Parameter Optimization Schemes for Las Vegas Valley

Figure 8. Numerical calibration of sampled T values for zones a, b, c, d, e, f, and g using DREAM MCMC.

observed land subsidence values are small. Quantitative assessment of the model fitting is listed in Table 1. The NS criterion indicates that the fit is very good for both the ZM and the adaptive multi-scale inversion methods. The m criterion indicates that, on average, simulated subsidence values are smaller than the observed ones. The D criterion indicates that the fit is good with both inversion strategies. The s criterion is decreasing with the ZM and the adaptive multi-scale methods, suggesting that the overall model fit to the

weighted observational data has improved. The adaptive multi-scale method appears to be superior to the ZM method based on NS, D, and s criteria (Table 1).

Observed versus simulated hydraulic heads for the evaluation period for the ZM and the adaptive multiscale strategy are shown in Figure 15. The results indicate that the ZM and adaptive multi-scale methods yield superior results (better fitting) when compared with Yan’s model. The quantitative assessment of model fit is listed in Table 1. The NS criterion indicates

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 173–194

185

Zhang and Burbey

Figure 9. The simulated marginal posterior distributions of the four transmissivities.

that the fit is excellent for the ZM and the adaptive multi-scale inversion methods. The m criterion indicates that, on average, the simulated hydraulic heads are lower than the observed heads for the ZM, while simulated heads are higher than observed heads for the other two methods. The D criterion indicates that the fit is very good for both the ZM and the adaptive multi-scale inversion strategies. The s criterion is decreasing with the ZM and the adaptive multi-scale methods. Based on quantitative assessment results (Table 1), the adaptive multi-scale method is shown to be superior to the ZM method based on these criteria. Figure 13 shows that the adaptive multi-scale method produces the best model fit at sites h1–h4. Figure 16 shows plots that compare the observed versus predicted hydraulic heads at sites h3 and h4 (Figure

186

2) using the DREAM strategy. The results show that the 95 percent credible intervals are too narrow to include in the observed data. The mean predictions (the center of the intervals) are close to the observation data at site h3, but they are far from the observation data at site h4.

Observed and simulated land subsidence values for the evaluation period using the ZM and the adaptive multi-scale inversion strategies are shown in Figure 17. Results indicate that the greatest fit relative to Yan’s model occurs with the ZM (Figure 17b) and the adaptive multi-scale (Figure 17b) methods. A quantitative assessment of model fit is listed in Table 1 and indicates that none of the three calibrated models is sufficient to accurately predict land subsidence because all three of the

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 173–194

Three Hydraulic Parameter Optimization Schemes for Las Vegas Valley

Figure 10. The parameter correlation coefďŹ cients (PCC) between two parameters.

inversion methods produce simulated subsidence values that are lower than the observed values on average. Figure 18 shows plots that compare observed versus predicted land subsidence. The results show that the 95 percent credible intervals

include several observations, but the intervals are too narrow to include all the observed data. The mean predictions (the center of the intervals) are shown to be far from the observations. One reason for the insufficient prediction is that only 113 land

Figure 11. Elastic and inelastic skeletal storage coefďŹ cient zones and values (a) from Yan (2007); (b) calibrated with the ZM; (c) calibrated with the adaptive multi-scale strategy.

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 173–194

187

Zhang and Burbey

basin; thus, many areas where subsidence is known to have occurred (based on more recent InSAR data) lack observed values.

DISCUSSION The results of this study provide an opportunity to compare three inversion strategies that were invoked for calibrating T, Ske, and Skv for Las Vegas Valley at the basin scale with the aim of achieving both accuracy and computational efficiency.

Parameter Estimation

Figure 12. Observed hydraulic head vs. simulated hydraulic head for the calibration period with (a) ZM, and (b) adaptive multi-scale method.

subsidence measurements were used for the evaluation period from 1963 to 1980 in the calibration process (Figure 2), which evidently is insufficient to adequately reflect the complex pattern of the land subsidence at the basin scale. Furthermore, the observed land subsidence values are not random, but they occur in two transects through the

The parameters T, Ske, and Skv of the developed zone aquifer (principal aquifer, layer 3) were calibrated for Las Vegas Valley at the basin scale. Many investigations and regression analyses have been used to determine T, Ske, and Skv values by various researchers (Harrill, 1976; Plume, 1984; Morgan and Dettinger, 1994; Hoffmann et al., 2001; Pavelko, 2004; and Bell et al., 2008). In order to constrain the range of the parameters for the ZM and adaptive multi-scale method, prior information obtained from the previous investigations was used for the parameters T, Ske, and Skv. Parameters described in Yan’s model (Yan, 2007) were treated as initial guesses for both zones and parameters in our study (Figures 1a and 6a). Calibrated T values that we obtained have similar values to those from Yan, except at zones E and S (Figure 6a and b), where coarsegrained sediments occur in the eastern part of the basin (close to point A9 in Figure 1b and in the southern part of the basin near Henderson shown in Figure 1a).

Calibrated Ske and Skv values are generally lower than those from Yan (Figure 11a–c), but they are among the reasonable range for the clay inter-bed.

Because the inter-bed thicknesses were inferred for a large portion of the basin, especially in the northwest

Table 1. Measures of model fit between observed and simulated hydraulic head and land subsidence.

Hydraulic Head Model YAN C P ZM C P Adaptive multi-scale method C P

NS

m

0.73 0.9

Land Subsidence

D

s

1.09 0.84

2.02 2.5

2.01 2.48

0.86 0.93

0.54 0.12

1.48 2.12

0.88 0.93

0.27 −0.06

1.33 2.08

NS

m

D

s

0.61 0.16

−16.93 −32.67

5.26 22.85

5.56 5.28

1.39 2.02

0.73 0.01

−22.09 −43.43

4.74 26.74

3.35 4.12

1.29 1.98

0.77 −0.05

−24.16 −48.78

4.65 27.77

3.13 4.05

NS 5 Nash-Sutcliffe efficiency index; m 5 mass balance error; D 5 Schulz criterion; s 5 quantitative measure of overall model fit to the weighted observations; C 5 calibration period; P 5 prediction period; YAN 5 Yan (2007); ZM 5 zonation method.

188

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 173–194

Three Hydraulic Parameter Optimization Schemes for Las Vegas Valley

Figure 13. Observed hydraulic head vs. simulated hydraulic head at (a) site h1, (b) site h2, (c) site h3, and (d) site h4 for the three evaluated calibration strategies.

part of the area (see Figures 3.3.4-1 and 3.3.4-2 of Morgan and Dettinger, 1994), Ske and Skv zones had to be reconstructed using the adaptive multi-scale method. Calibrated T, Ske, and Skv zones and values reflect more realistically the hydrogeological conditions of the basin. A CSS analysis indicates that the information provided by the observations and the model is sufficient to calibrate all T, Ske, and Skv parameters in our study, with Skv exhibiting the largest CSS, which suggests that this parameter is highly sensitive to observations and important in the calibration process. Pavelko (2004) found that the inelastic skeletal specific storage has a relatively large CSS compared to other parameters. The PCC analysis suggests that no parameter is prevented from being uniquely estimated (Foglia et al., 2009).

For the DREAM MCMC method, we only simulated marginal posterior distributions of the four parameters in the model. The results shown in Figure 9 indicate that the formal likelihood Gaussian function assumption may not hold due to the complexity and different sources of uncertainty that exist in the model.

The PCC analysis shows that transmissivity parameters are weakly correlated. This is consistent with

the result obtained from the ZM and adaptive multiscale methods.

Model Fit For the ZM and the adaptive multi-scale methods, the measures NS, m, and D were used to evaluate the overall model fit, and these are independent of the observation weights, while s is a measure of overall model fit to the observations related to error-based weighting. The s criterion decreases with the ZM and the adaptive multi-scale method (Table 1), which suggests that the overall model fit to the weighted observational data has improved. For the DREAM method, the prediction uncertainties are quantified using their 95 percent credible intervals. The results shown in Table 1 indicate that the calibrated model fit, based on NS, D, and m criteria, is excellent or very good for observations of hydraulic head during the calibration and evaluation periods. The measures of model fit shown in Table 1 indicate that the model fit is very good or good for observations of land subsidence during the calibration period, but it is insufficient for land subsidence during the evaluation

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 173–194

189

Zhang and Burbey

Figure 14. Observed vs. simulated land subsidence for the calibration period with (a) ZM, and (b) adaptive multi-scale method.

Figure 15. Observed vs. simulated hydraulic heads for the evaluation period using (a) ZM, and (b) adaptive multi-scale method.

period based on weighting independent analysis. The prediction uncertainties shown in Figures 16 and 18 indicate that the 95 percent credible intervals are too narrow to include all the observed data, and the mean predictions (the center of the intervals) are not close to the observations. One likely reason for the insufficient prediction of the land subsidence during the evaluation period is because of the complex factors influencing subsidence such as inter-bed thickness, the location of basin-fill faults, delayed inter-bed drainage, and the time constant that contributes to the occurrence of land subsidence. Another reason for the insufficient prediction of the land subsidence is that the observational data used during the calibration period were taken from a set of transects from years 1963 and 1980, which, as pointed out by Bell (2002), does not encompass the entire spatial extent of the basin, and the transects miss major subsidence areas, and there are complex interactions among many of the faults in the region. However, the objective of this paper is to find the optimal inversion strategy for the Las Vegas Valley model, so we incorporated a shorter model for the sake of minimizing computational time and maximizing efficiency. In fact, InSAR and global

positioning system (GPS) data have revealed new spatial patterns of subsidence that were not evident on earlier conventional subsidence maps based on the data used in this investigation (Hoffmann et al., 2001; Bell et al., 2002, 2008). We are in the process of incorporating high-spatial-resolution InSAR data for the period from 1992–2010 and anticipate greatly improved calibration results when these new InSAR data are included in the inversion model. Nonetheless, the data used in this investigation reveal that the multiscale method is the best inversion method to pursue for the full Las Vegas model.

190

Comparison of the Three Inversion Strategies Three different inversion strategies were investigated, including ZM, the adaptive multi-scale method, and DREAM MCMC. The ZM is a fast inversion method but is somewhat subjective when defining actual parameter zones. The multi-scale method is an approach that provides criteria to reconstruct the zones by refining or coarsening the current zones based on actual hydrogeological conditions, which removes the subjectivity in identifying zones that is

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 173–194

Three Hydraulic Parameter Optimization Schemes for Las Vegas Valley

Figure 16. Observed hydraulic head vs. simulated hydraulic head at (a) site h3 and (b) site h4 for the DREAM strategy.

inherent in the other methods. In this study, we first identifies those zones where hydrogeological analysis shows that the refinement or coarsening may be required. We then identifies those zones that are important to fitting the observations through a sensitivity analysis (Grimstad et al., 2003; Foglia et al., 2009). Both refinement and coarsening indicators provide first-order information on the minimization of the optimal objective function, so the decrease of the objective function is not guaranteed all the time, but this provides an alternative to reconstructing the current zones by saving time from trial-and-error guesses for zone structures. Overall, 102 parameters were estimated using 470 observations, which means that overparameterization is not a concern in this investigation.

Due to the complex hydrogeological conditions and large amount of available temporal and spatial groundwater hydraulic head and land subsidence measurements in Las Vegas Valley area, a model with finer cells and more parameter zones is required to reflect the actual hydrogeological conditions and to better predict the hydraulic head and land subsidence patterns. This study reveals that the adaptive multi-scale method represents an effective, flexible, and computationally efficient methodology for reconstructing the T, Ske, and Skv zones according to hydrogeological and sensitivity analyses.

Figure 17. Observed vs. simulated land subsidence for the evaluation period using (a) ZM, and (b) the adaptive multi-scale method.

The DREAM MCMC method is a scheme that significantly improves the efficiency of MCMC simulations (Vrugt et al., 2009). It can be used to estimate the posterior probability density function of model parameters in high-dimensional and multi-modal sampling problems. However, the computational time is excessive and impractical for such large models as the Las Vegas model developed in this study. Because many sources of uncertainty (such as measurement error, input error, and structure error) exist in the model processing, the commonly used formal likelihood function (Eq. 14), which is based on the assumption that the residual term follows a multi-variate Gaussian distribution, is not practical for complex applications (Beven et al., 2008). Thus, the DREAM MCMC method could not be used with the formal likelihood function to calibrate all the parameters for the Las Vegas model, and it is better suited for smaller, less complex, simulation problems. However, in general, DREAM itself is not limited to the Gaussian likelihood function, although the DREAM code of UCODE-2014 only uses the Gaussian likelihood function. For example, the formal generalized likelihood

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 173–194

191

Zhang and Burbey

Figure 18. Observed land subsidence vs. predicted land subsidence for the DREAM strategy.

function of Schoups and Vrugt (2010) is an alternative approach for calibrating the parameters for a problem containing non-Gaussian residuals (Shi et al., 2014).

The quantitative model fit measurement results show that, in general, the adaptive multi-scale method is superior to the ZM and Yan’s model (Table 1 and Figure 13). The adaptive multi-scale method is more efficient than the DREAM MCMC method. Therefore, it is suggested that the adaptive multi-scale method represents the best and most efficient strategy for calibrating the optimal model parameters for making credible predictions, for providing a framework for developing an accurate hydrogeologic model for Las Vegas Valley, and for other similar highly complex non-linear problems.

CONCLUSIONS This paper explores three inverse strategies, (1) the ZM, (2) the adaptive multi-scale method, and (3) DREAM to estimate the distribution of transmissivity (T) and elastic and inelastic skeletal storage coefficients (Ske and Skv, respectively) from observations of hydraulic head and land subsidence measurements for Las Vegas Valley between 1912 and 1987. The key conclusions can be defined as follows: 1. The calibrated T values that we obtained from ZM and the adaptive multi-scale method are similar in magnitude and distribution to those of Yan (2007), except at zones E and S (Figure 5b and c).

Calibrated Ske and Skv values are lower than those from Yan (Figure 11a–c), but they still fall within a reasonable range for the clay inter-bed. The Ske and Skv zones were reconstructed with the adaptive multi-scale method, where the inter-bed thicknesses were inferred from Morgan and Dettinger (1994).

Calibrated T, Ske, and Skv zones and values appear to reflect more realistically the hydrogeological conditions of the basin.

192

2. Because many sources of uncertainty (such as measurement error, input error, and structure error) exist in the Las Vegas model, the DREAM MCMC is not compatible with UCODE-2014 using a formal likelihood function (Eq. 14), which is based on the assumption that the residual term follows a multi-variate Gaussian distribution.

Therefore, DREAM, with its formal likelihood function, is not a good choice for calibrating parameters for this study. However, in general, DREAM itself is not limited to the Gaussian likelihood function. For example, the formal generalized likelihood function of Schoups and Vrugt (2010) is an alternative approach for calibrating the parameters for a problem containing nonGaussian residuals.

3. The adaptive multi-scale method is able to quickly and efficiently reconstruct the T, Ske, and Skv zones while providing more flexibility and accuracy than the other two methods. The model fit results show that overall, the adaptive multi-scale method is superior to the ZM method. The adaptive multiscale method is more efficient than the DREAM MCMC method, and it can be used to successfully calibrate parameters at the basin scale. Thus, the adaptive multi-scale method is the best strategy for calibrating optimal model parameters and providing a framework for developing an accurate hydrogeologic model for Las Vegas Valley.

4. The measures NS, m, D, and s were used to evaluate model fit for the ZM and the adaptive multiscale method in this study, among which s is related to the error-based weighting. The measures of model fit indicate that the model fit is excellent or very good for observations of hydraulic head during the calibration and evaluation periods, and the fit is very good or good for observations of land subsidence during the calibration period, but it is insufficient for land subsidence during the evaluation period. For the DREAM method, with its formal likelihood function, results show that the 95 percent credible intervals are too narrow to include all the observed data, and the mean predictions are not close to the observation data. Better calibration results are expected once high-spatialresolution and high-temporal-resolution InSAR data (basin-wide subsidence data) are included in the inversion model.

5. A CSS analysis indicates that information provided by observations is sufficient to calibrate all T, Ske, and Skv parameters in our study. A PCC analysis shows that no parameter is prevented from being uniquely estimated using the multi-scale method and ZM.

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 173–194

Three Hydraulic Parameter Optimization Schemes for Las Vegas Valley

ACKNOWLEDGMENTS The authors would like to thank David Donovan of the Las Vegas Valley Water District for providing water-level data and John Bell of the Nevada Bureau of Mines and Geology, who provided the subsidence data. In addition, the authors wish to thank Mary Hill of the U.S. Geological Survey and Dan Lu of the Oak Ridge National Laboratory for providing UCODE-2014, beta version. We also appreciated the help from the Advanced Research Computing team at Virginia Tech.

REFERENCES ALCOLEA, A.; CARRERA, J.; AND MEDINA, A., 2006, Pilot points method incorporating prior information for solving the groundwater flow inverse problem: Advances in Water Resources, Vol. 29, No. 11, pp. 1678–1689.

AMEUR, H. B.; CHAVENT, G.; AND JAFFRÉ, J., 2002, Refinement and coarsening indicators for adaptive parametrization: Application to the estimation of hydraulic transmissivities: Inverse Problems, Vol. 18, No. 3, pp. 775–802.

BALLIO, F. AND GUADAGNINI, A., 2004, Convergence assessment of numerical Monte Carlo simulations in groundwater hydrology: Water Resources Research, Vol. 40, No. 4, p. W04603.

BELL, J. W., 1981, Subsidence in Las Vegas Valley: Mackay School of Mines, University of Nevada, Reno, NV. BELL, J. W.; AMELUNG, F.; FERRETTI, A.; BIANCHI, M.; AND NOVALI, F., 2008, Permanent scatterer InSAR reveals seasonal and long‐term aquifer‐system response to groundwater pumping and artificial recharge: Water Resources Research, Vol. 44, No. 2, p. W02407.

BELL, J. W.; AMELUNG, F.; RAMELLI, A. R.; AND BLEWITT, G., 2002, Land subsidence in Las Vegas, Nevada, 1935–2000: New geodetic data show evolution, revised spatial patterns, and reduced rates: Environmental & Engineering Geoscience, Vol. 8, No. 3, pp. 155–174.

BEVEN, K. J.; SMITH, P. J.; AND FREER, J. E., 2008, So just why would a modeller choose to be incoherent?: Journal of Hydrology, Vol. 354, No. 1–4, pp. 15–32.

CASELLA, G. AND BERGER, R., 2002, Statistical Inference: Duxbury, Pacific Grove, CA. CHAVENT, G. AND LIU, J., 1989, Multiscale parametrization for the estimation of a diffusion coefficient in elliptic and parabolic problems. In 5th IFAC Symposium on Control of Distributed Parameter Systems: (Perpignan, France, 26–29 June 1989) ed.

M Amouroux and A El Jai, Oxford: Pergamon p. 193–202.

DE MARSILY, G.; LAVEDAN, G.; BOUCHER, M.; AND FASANINO, G., 1984, Interpretation of interference tests in a well field using geostatistical techniques to fit the permeability distribution in a reservoir model. In Verly, G., David, M., Journel, A. G., and Maréchal, A. (Editors), Geostatistics for Natural Resour‐ ces Characterization, Part 2: Reidel, Dordrecht, Netherlands, pp. 831–849.

DOHERTY, J., 2003, Ground water model calibration using pilot points and regularization: Ground Water, Vol. 41, No. 2, pp.

170–177.

DOHERTY, J., 2004, PEST Model-Independent Parameter Estimation User Manual: Watermark Numerical Computing, Brisbane, Australia.

FOGLIA, L.; HILL, M.; MEHL, S.; AND BURLANDO, P., 2009, Sensitivity analysis, calibration, and testing of a distributed hydrological

model using error‐based weighting and one objective function: Water Resources Research, Vol. 45, No. 6, p. W06427.

FOSTER, I.; KESSELMAN, C.; AND TUECKE, S., 2001, The anatomy of the grid: Enabling scalable virtual organizations: International Journal of High Performance Computing Applications, Vol. 15, No. 3, pp. 200–222.

FU, J. AND JAIME GÓMEZ-HERNÁNDEZ, J., 2009, Uncertainty assessment and data worth in groundwater flow and mass transport modeling using a blocking Markov chain Monte Carlo method: Journal of Hydrology, Vol. 364, No. 3, pp. 328–341.

GRIMSTAD, A.-A.; MANNSETH, T.; NÆVDAL, G.; AND URKEDAL, H., 2003, Adaptive multiscale permeability estimation: Computational Geosciences, Vol. 7, No. 1, pp. 1–25.

HARBAUGH, A. W., 2005, MODFLOW-2005, the US Geological Survey Modular Ground-Water Model: The Ground-Water Flow Process: U.S. Geological Survey Techniques and Methods 6-A16.

HARRILL, J. R., 1976, Pumping and Depletion of Ground-Water Storage in Las Vegas Valley, Nevada, 1955–74: Carson City, Nevada, State of Nevada, Division of Water Resources, Water Resources Bulletin no. 44.

HAYEK, M. AND ACKERER, P., 2007, An adaptive subdivision algorithm for the identification of the diffusion coefficient in two-dimensional elliptic problems: Journal of Mathematical Modelling and Algorithms, Vol. 6, No. 4, pp. 529–545.

HAYEK, M., LEHMANN, F.; AND ACKERER, P., 2008, Adaptive multiscale parameterization for one-dimensional flow in unsaturated porous media: Advances in Water Resources, Vol. 31, No. 1, pp. 28–43.

HELM, D. C., 1975, One‐dimensional simulation of aquifer system compaction near Pixley, California: 1. Constant parameters: Water Resources Research, Vol. 11, No. 3, pp. 465–478.

HILL, M. C., 1998, Methods and Guidelines for Effective Model Calibration: U.S. Geological Survey Water-Resources Investigations Report 98-4005.

HILL, M. C. AND TIEDEMAN, C. R., 2007, Effective Groundwater Model Calibration: With Analysis of Data, Sensitivities, Predictions, and Uncertainty: John Wiley & Sons, Inc., New York, New York, 455 p.

HOFFMANN, J.; LEAKE, S.; GALLOWAY, D.; AND WILSON, A. M., 2003, MODFLOW-2000 Ground-Water Model—User Guide to the Subsidence and Aquifer-System Compaction (SUB) Package: U.S. Geological Survey Open-File Report 03-233.

HOFFMANN, J.; ZEBKER, H. A.; GALLOWAY, D. L.; AND AMELUNG, F., 2001, Seasonal subsidence and rebound in Las Vegas Valley, Nevada, observed by synthetic aperture radar interferometry: Water Resources Research, Vol. 37, No. 6, pp. 1551–1566.

JENG, D. I., 1998, A Ground-Water Flow and Land Subsidence Model of Las Vegas Valley, Nevada—A Converted MODFLOW Model: M.S. thesis, University of Nevada, Reno, NV, 133 p.

JOHNSON, A.; MOSTON, R.; AND MORRIS, D., 1968, Physical and Hydrologic Properties of Water-Bearing Deposits in Subsiding Areas in Central California: U.S. Geological Survey Professional Paper 497.

JORGENSEN, D. G., 1980, Relationships between Basic Soils-Engineering Equations and Basic Ground-Water Flow Equations: U.S. Geological Survey Water-Supply Paper 2064.

KEATING, E. H.; DOHERTY, J.; VRUGT, J. A.; AND KANG, Q., 2010, Optimization and uncertainty assessment of strongly nonlinear groundwater models with high parameter dimensionality: Water Resources Research, Vol. 46, No. 10, p. W10517.

LAVENUE, M. AND MARSILY, G., 2001, Three‐dimensional interference test interpretation in a fractured aquifer using the pilot point inverse method: Water Resources Research, Vol. 37, No. 11, pp. 2659–2675.

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 173–194

193

Zhang and Burbey LIU, J., 1993, A multiresolution method for distributed parameter estimation: SIAM Journal on Scientific Computing, Vol. 14, No. 2, pp. 389–405.

LU, D.; YE, M.; HILL, M. C.; POETER, E. P.; AND CURTIS, G. P., 2014, A computer program for uncertainty analysis integrating regression and Bayesian methods: Environmental Modelling & Software, Vol. 60, pp. 45–56.

MALMBERG, G. T., 1965, Available Water Supply of the Las Vegas Ground-Water Basin, Nevada: U.S. Geological Survey Water-Supply Paper 1780.

MAXEY, G. B. AND JAMESON, C., 1948, Geology and Water Resources of Las Vegas, Pahrump, and Indian Spring Valleys, Clark and Nye Counties, Nevada: Nevada Office of the State Engineer, Carson City, NV. MCDONALD, M. G. AND HARBAUGH, A. W., 1984, A Modular ThreeDimensional Finite-Difference Ground-Water Flow Model: Scientific Publications Company Washington, Reston, VA. MORGAN, D. S. AND DETTINGER, M. D., 1994, Ground-Water Conditions in Las Vegas Valley, Clark County, Nevada: Part II. Geohydrology and Simulation of Ground-Water Flow: U.S. Geological Survey Open-File Report 90-179.

MORGAN, D. S. AND DETTINGER, M. D., 1996, Ground-Water Conditions in Las Vegas Valley, Clark County, Nevada; Part 2.

Hydrogeology and Simulation of Ground-Water Flow: U.S. Geological Survey Water-Supply Paper 2320B. NASH, J. AND SUTCLIFFE, J., 1970, River flow forecasting through conceptual models: Part I—A discussion of principles: Journal of Hydrology, Vol. 10, No. 3, pp. 282–290.

PAVELKO, M. T., 2004, Estimates of Hydraulic Properties from a One-Dimensional Numerical Model of Vertical Aquifer-System Deformation, Lorenzi Site, Las Vegas, Nevada: U.S. Geological Survey Water-Resources Investigations Report 03-4083.

PLUME, R. W., 1984, Ground-Water Conditions in Las Vegas Valley, Clark County, Nevada; Part I. Hydrogeologic Framework: U.S. Geological Survey Open-File Report 84-130.

POETER, E. P. AND HILL, M. C., 1998, Documentation of UCODE: A Computer Code for Universal Inverse Modeling: DIANE Publishing, Collingdale, PA. POETER, E. P.; HILL, M. C.; BANTA, E. R.; MEHL, S.; AND CHRISTENSEN, S., 2005, Ucode_2005 and Six Other Computer Codes for Universal Sensitivity Analysis, Calibration, and Uncertainty Evaluation: U.S. Geological Survey Techniques and Methods 6-A11.

POETER, E. P.; HILL, M. C.; LU, D.; AND MEHL, S. W., 2014, UCODE_2014, with New Capabilities to Define Parameters Unique to Predictions, Calculate Weights Using Simulated Values, Estimate Parameters with SVD, and Evaluate Uncertainty with MCMC: International Ground Water Modeling Center Report GWMI 2004-02.

POLAND, J. F. AND DAVIS, G. H., 1969, Land subsidence due to withdrawal of fluids: Reviews in Engineering Geology, Vol. 2, pp. 187–270.

194

POLAND, J. F.; LOFGREN, B.; AND RILEY, F. S., 1972, Glossary of Selected Terms Useful in Studies of the Mechanics of Aquifer Systems and Land Subsidence due to Fluid Withdrawal: U.S. Geological Survey Water Supply Paper.

PRASAD, K. L., 2001, Estimating net aquifer recharge and zonal hydraulic conductivity values for Mahi Right Bank Canal Project area, India, by genetic algorithm: Journal of Hydrology, Vol. 243, pp. 149–161.

RAO, S. V. N., 2003, Optimal groundwater management in deltaic regions using simulated annealing and neural networks: Water Resources Management, Vol. 17, No. 6, pp. 409–428.

RENARD, P., 2007, Stochastic hydrogeology: What professionals really need?: Groundwater, Vol. 45, No. 5, pp. 531–541.

RENARD, P. AND DE MARSILY, G., 1997, Calculating equivalent permeability: A review: Advances in Water Resources, Vol. 20, No. 5, pp. 253–278.

SCHOUPS, G. AND VRUGT, J. A., 2010, A formal likelihood function for parameter and predictive inference of hydrologic models with correlated, heteroscedastic, and non‐Gaussian errors: Water Resources Research, Vol. 46, No. 10, p. W10531.

SCHULZ, K.; BEVEN, K.; AND HUWE, B., 1999, Equifinality and the problem of robust calibration in nitrogen budget simulations: Soil Science Society of America Journal, Vol. 63, pp.

1934–1941.

SHI, X.; YE, M.; CURTIS, G. P.; MILLER, G. L.; MEYER, P. D.; KOHLER, M.; YABUSAKI, S.; AND WU, J., 2014, Assessment of parametric uncertainty for groundwater reactive transport modeling: Water Resources Research, Vol. 50, No. 5, pp.

4416–4439.

TERZAGHI, K., 1925, Principles of soil mechanics: IV—Settlement and consolidation of clay: Engineering News Record, Vol. 95, No. 3, pp. 874–878.

TRESCOTT, P. C.; PINDER, G. F.; AND LARSON, S., 1976, Finite-Difference Model for Aquifer Simulation in Two Dimensions with Results of Numerical Experiments: U.S. Geological Survey Techniques of Water-Resources Investigations 07-C1.

VRUGT, J. A.; TER BRAAK, C.; DIKS, C.; ROBINSON, B. A.; HYMAN, J. M.; and HIGDON, D., 2009, Accelerating Markov chain Monte Carlo simulation by differential evolution with self-adaptive randomized subspace sampling: International Journal of Nonlinear Sciences and Numerical Simulation, Vol. 10, No. 3, pp.

273–290.

YAN, T., 2007, Effects of Delayed Drainage on Subsidence Modeling and Parameter Estimation: M.S. Thesis, Geosciences Department, Virginia Tech, Blacksburg, VA. YAN, T. AND BURBEY, T. J., 2008, The value of subsidence data in ground water model calibration: Groundwater, Vol. 46, No. 4, pp. 538–550.

ZHANG, M.; BURBEY, T. J.; NUNES, V. D. S.; AND BORGGAARD, J., 2013, A new zonation algorithm with parameter estimation using hydraulic head and subsidence observations: Ground Water, Vol. 52, No. 4, pp. 514–524.

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 173–194

Tracers of Groundwater Mixing in the Hueco Bolson Aquifer, Ciudad Juárez, Mexico CHRISTOPHER J. EASTOE Department of Geosciences, University of Arizona, 1040 E 4th Street, Tucson, AZ 85721, USA

ALFREDO GRANADOS OLIVAS Department of Civil and Environmental Engineering–Geoscience Academic Group, Autonomous University of Ciudad Juárez, Avenida del Charro 610 norte, C.P. 32310 Ciudad Juárez, Chihuahua, Mexico

BARRY J. HIBBS Department of Geological Sciences, California State University–Los Angeles, 5151 State University Drive, Los Angeles, CA 90032-8203, USA

Key Terms: Groundwater Hydrology, Geochemistry, Isotopes, Groundwater Quality, Dams

ABSTRACT Stable O and H isotopes and radioisotopes show that Río Bravo water recharged prior to the construction of Elephant Butte Dam is the principal source of groundwater in the Hueco Bolson aquifer beneath Ciudad Juárez. Mixing between pre-dam and post-dam river water occurs in samples from the aquifer beneath the Río Bravo floodplain, where finite tritium is also present. Along the flanks of the Sierra de Juárez, mixing occurs between predam river water and water derived from the Sierra de Juárez. Carbon-14 content of groundwater generally decreases from northwest to southeast beneath the city; corrected residence times in city supply wells range from post-bomb to 3600 years. Sulfur isotopes in groundwater of pre-dam river water origin and anion ratios (Cl/Br and Cl/SO4) are consistent with mixing of solutes from native Hueco Bolsón groundwater with river water. Stable isotopes serve to identify mixing of 9 percent or more of local water into pre-dam river-derived water. Mixing of smaller amounts of saline Hueco Bolson groundwater into river-derived groundwater can best be identified using anion ratios. The improved understanding of groundwater origins, flow paths, and quality arising from this study is crucial information for future water-supply engineering in Ciudad Juárez and for elucidating preferred pathways for contaminant movement in groundwater.

INTRODUCTION Ciudad Juárez, Chihuahua, Mexico, has until recently depended mainly on groundwater pumped

from the Hueco Bolsón aquifer beneath the city for a potable water supply. Rapid urban growth has stretched the local groundwater supply to its limits, leading to the development of a large cone of depression under the city (Hibbs et al., 1997) and to increasing salinity in older wells as the fresher, shallow groundwater has been pumped (del Hierro-Ochoa et al., 2004). In urban study sites such as Ciudad Juárez, an understanding of current and pre-development groundwater flow paths and of the ambient water-quality signature derived from pre-development recharge prepares groundwater scientists and engineers for a better understanding of contemporary contaminants, the origin of salinity, and water budget transfers.

In order to improve the understanding of groundwater origin, flow paths, and residence times in the part of the Hueco Bolsón aquifer beneath Ciudad Juárez, we undertook a study of stable isotopes (O, H, S), radioactive isotopes (tritium, carbon-14), and major anion concentrations in groundwater from the aquifer in Texas and Chihuahua (Hibbs et al., 2003; Druhan et al., 2008; and Eastoe et al., 2008, 2010). This article reports on the chemistry, including isotope chemistry, of groundwater from municipal wells in Ciudad Juárez, Chihuahua, and a well-nest beneath the Río Bravo floodplain in adjacent El Paso, TX. The aims of the study are to identify groundwater origins and flow paths in the Hueco Bolson aquifer beneath Ciudad Juárez, to identify mixtures of waters of different origins and the implications of mixing for salinity, to compare the capabilities of isotopes and anion ratios in this regard, and to quantify residence times of groundwater in the aquifer.

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 195–207

195

Eastoe, Granados Olivas, and Hibbs

water to meet all demands (both in the United States and in Mexico) in the Hueco Bolsón.

Hydrogeology

Figure 1. Map of the region around the study area.

STUDY AREA Geography The study area includes most of the urban area of Ciudad Juárez (Figure 1). It is bounded to the north and east by the international border with the United States along the Río Bravo (known as the Rio Grande in the United States), and to the southwest by the Sierra de Juárez, a hard-rock mountain range. The city has grown in population from 567,365 in 1980 to 1,332,131 in 2010 (INEG y GEC, 2014). The elevation is 1134 m above sea level (masl) in the city center, rising to 1700 masl at the crest of the Sierra de Juárez.

The climate is semiarid, with an annual average temperature of 25uC and an average annual rainfall of 250 mm (Climatological Laboratory, Universidad Autónoma de Ciudad Juárez, 2014). Precipitation is highly variable from year to year. It is largely confined to two seasons: a winter season (November to March) of infrequent frontal storms from the west, and a summer monsoon season (late June to September) during which convective storms drop moisture originating from the southeast or southwest. On average, summer rain (June–October) constitutes 67 percent of the mean annual total of 236 mm in nearby El Paso, TX (National Weather Service, 2014, http://www.srh. noaa.gov/epz/?n5elpaso_monthly_precip). Flow in the reach of the Río Bravo adjacent to Ciudad Juárez is at present intermittent, but was usually perennial prior to development. A large dam that impounds the river at Elephant Butte, NM, was completed in 1916, and flow in the river downstream of the dam has been managed for irrigation and municipal supply since that date. At present, there is insufficient river

196

The Hueco Bolsón is an asymmetric graben of the Basin and Range Province. Beneath Ciudad Juárez, the trough is more than 1,000 m deep, and it is filled with Neogene lacustrine, fluvial, eolian, and alluvial sediment (Figure 2). The following description uses stratigraphic terminology applied locally in Texas and New Mexico. Fluvial sand and gravel of the Camp Rice Formation, up to 300 m thick, overlie lacustrine silt and clay of the Fort Hancock Formation. The transition from the Fort Hancock Formation to the Camp Rice Formation reflects the hydrologic evolution of the basin from an internal drainage with playa lakes fed by the ancestral upper Río Bravo to a basin in which the ancestral Río Bravo was throughflowing about 1.8 Ma (Stuart and Willingham, 1984; Connell et al., 2005). Until about 0.67 Ma, the river entered the Hueco Bolsón through Fillmore Gap in New Mexico (41 km north of the present point of entry) and flowed south along the eastern flanks of the Franklin Mountains and the Sierra de Juárez.

The Camp Rice Formation is thickest near the western boundary fault system of the Hueco Bolsón graben, where the paleo-river was localized by rapid subsidence; in this area, it comprises braided fluvial channel deposits (Stuart and Willingham, 1984). At present, the river flows south through the Mesilla Valley, west of the Franklin Mountains, and enters the Hueco Bolsón through the Narrows, a rock-cut gorge between the Franklin Mountains and the Sierra de Juárez.

Figure 2. Schematic cross section of study area, modified from a cross section prepared by J. Hawley for the regional study of which this research forms part. Location of cross section is shown in Figure 1.

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 195–207

Groundwater Mixing in Ciudad Juárez

The Hueco Bolsón aquifer is hosted by the Camp Rice and Fort Hancock Formations. Potable water is limited to the Camp Rice Formation. In Texas, the Fort Hancock Formation hosts saline water (Hibbs et al., 1997; Druhan et al., 2008). The Hueco Bolsón aquifer, with pre-development water levels .30 m below the surface, is distinguished from the Río Bravo alluvial aquifer (known as the Rio Grande alluvial aquifer) hosted by recent fluvial deposits of the active floodplain. Water in the alluvial aquifer is shallow, as close as 2 m below the surface, and the aquifer extends no more than 30 m below the surface.

Potable water supplied by the Junta Municipal de Agua y Sanamiento (JMAS) at the time of sampling for this study in 2003–2005 came entirely from 143 deep wells that ranged in depth to 251 m, drilled within the Camp Rice Formation, in the upper part of the Hueco Bolsón aquifer (IMIP, 2003). Continued pumping over many years has led to drawdown of water levels. Figure 3.10 of Hibbs et al. (1997) showed maximum drawdown of about 18 m between 1987 and 1993 at the center of a cone of depression 8 to 11 km across, beneath the southern half of the urban area, and extending from the Sierra de Juárez foothills to the Río Bravo floodplain. In 2014, the JMAS was drawing water from up to 173 wells in the Hueco Bolsón (M. C. Ezequiel Rascón, JMAS, pers. comm.). In the northwestern part of the urban area, groundwater levels rebounded a few meters between 1987 and 1993. Since 2010, pumping of groundwater from the Conejos-Médanos aquifer, west of the Sierra de Juárez, has allowed the JMAS to draw less water from the Hueco Bolsón aquifer, and this has led to a small recovery of static water levels in the center of the cone of depression (M. C. M. Herrera, JMAS, pers. comm.).

Two major geological units make up the aquifer system beneath Ciudad Juárez. A lower unit of Neogene age consists of clay, siltstone, sandstone, and conglomerate, includes fluvial facies, and is approximately equivalent to the Camp Rice Formation. An upper unit, of Quaternary age, contains alluvial fan and piedmont slope deposits; fluvial and eolian units are also present. In detailed view, many lenses or thin beds of clay and silt occur where the supply wells are located (Leos Rodriguez, 2004). Together, the units form a single, semi-confined aquifer system of moderate permeability, with a depth of more than 500 m. Transmissivity has been measured at 1.8 6 10−3 m2/s.

Groundwater quality varies around 600 mg/L total dissolved solids (TDS) in the urban areas (INEG y GEC, 1999).

PREVIOUS WORK Isotope data (O and H stable isotopes and tritium) for groundwater reported to be from irrigation wells in the Juárez Valley, without specific location data, were published by Payne (1976). Payne noted that the data plotted on an evaporation trend, and that salinity and tritium were higher for more evaporated samples. He interpreted the data as indicating concurrent evaporation, salinity increase, and exposure to atmospheric tritium in irrigated fields. Streeter (1990) analyzed heavy metals in groundwater samples from central Ciudad Juárez in a study focusing on public heath, but also potentially provided tracers of groundwater movement since industrial development in the area. Elevated antimony concentrations were detected in well water from beneath the Río Bravo floodplain.

Eastoe et al. (2008) showed that groundwater from most Ciudad Juárez supply wells had originated as surface water from the Río Bravo, and that most had infiltrated prior to the completion of the Elephant Butte Dam. Eastoe et al. (2010) presented stable O and H isotope data and MODFLOW modeling in a 300 m vertical section of the Hueco Bolsón aquifer beneath the Río Bravo between the cities of El Paso and Juárez. Within 5 km of the point of entry of the river, pre-dam river water could be identified as far as 300 m below the surface. Further downstream, the stable isotopes and modeling indicated a zone in which native saline Hueco Bolsón groundwater was being drawn southward beneath the river and toward the cone of depression beneath Ciudad Juárez.

SAMPLE COLLECTION AND METHODS Groundwater samples were collected in 2003 and 2004 from public supply wells of the JMAS of Ciudad Juárez, from springs in the Sierra de Juárez, and from a U.S. Geological Survey (USGS) well nest and from Río Bravo surface water in 2003 and 2004. The JMAS well samples represent groundwater from the entire screened interval of each well, while the wellnest samples represent discrete screen intervals listed in Table 1. Precipitation samples were collected during 2004 and 2005 in a rain gauge at an elevation of 1120 masl in central Ciudad Juárez and combined to make seasonal aggregates weighted for amount. Isotope measurements (except for carbon-14) were performed at the Environmental Isotope Laboratory, University of Arizona, Tucson, AZ. Stable O, H, and C isotopes were measured on a Finnigan Delta SH dual-inlet mass spectrometer equipped with an automated CO2 equilibrator (for O) and an automated Cr-reduction furnace (for H). Stable S isotopes were measured on a Thermo Electron Delta Plus XLH continuous-flow

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 195–207

197

198

Date

Screen (m)

Surface Water RG1 7/11/2002 RG2 8/23/2002 RG3 11/13/2002 RG4 2/27/2003 RG5 1/6/2003 RG6 9/16/03 RG7 3/20/2004 RG8 5/6/2004 U.S. Geological Survey Well Nest, El Paso JL-49-21-324 7/8/2002 8.5–11.5 JL-49-21-320 7/9/2002 34.5–37.6 JL-49-21-319 7/9/2002 54.8–57.9 JL-49-21-318 7/9/2002 105.5–108.5 JL-49-21-323 7/11/2002 171.5–174.5 JL-49-21-322 7/11/2002 199.7–202.7 Municipal Wells, Ciudad Juárez JMAS 1R 3/18/2004 JMAS 3Z 2003 JMAS 4Z 2003 JMAS 5 2003 JMAS 9R 3/18/2004 JMAS 13RR 2003 JMAS 15 2003 JMAS 17R 3/18/2004 JMAS 17R 2003 JMAS 19R 3/18/2004 JMAS 38 2003 JMAS 42R 2003 JMAS 45 2003 JMAS 48 3/18/2004 JMAS 50R 2003 JMAS 53R 2003 JMAS 56R 2003 JMAS 62 3/18/2004 JMAS 62R 2003 JMAS 76 2003 JMAS 82R 2003 JMAS 84 2003 JMAS 87 2003 JMAS 92 2003 JMAS 99R 2003 JMAS 112 2003 JMAS 115 3/18/2004 JMAS 120 2003

Sample

210 345 425 254 137 134 367 200 402 161 121 409 105 234 162 116 109 93 61 73 250 122 105

133 289 342 118 130 127 189 175 222 117 65 285 139 279 313 80 81 95 50 54 155 121 98

−106.423 −106.340 −106.364 −106.456 −106.489 −106.488 −106.459 −106.470 −106.470 −106.415 −106.428 −106.427 −106.450 −106.402 −106.437 −106.495 −106.369 −106.486 −106.486 −106.401 −106.467 −106.466 −106.460 −106.451 −106.444 −106.415 −106.394 −106.401 31.595 31.686 31.659 31.610 31.600 31.625 31.733 31.731 31.731 31.647 31.608 31.630 31.635 31.703 31.660 31.606 31.662 31.600 31.600 31.731 31.668 31.652 31.664 31.678 31.691 31.665 31.672 31.652

351 240 274 398 130 60

231

211 168 291 1,196 1,718 317

−106.327 −106.327 −106.327 −106.327 −106.327 −106.327

31.311 31.311 31.311 31.311 31.311 31.311

506 1,028 241 688

200 204 396

SO4 (mg/L)

162

386 768 202 615

93 109 273

Cl (mg/L)

−106.327 −106.327 −106.327 −106.327 −106.327 −106.327 −106.327 −106.327

Long. (uW)

31.311 31.311 31.311 31.311 31.311 31.311 31.311 31.311

Lat. (uN)

0.2 0.1 0.2 0.1 0.1 0.1 0.2 0.2 0.1

0.1 0.6 0.1 0.2

0.2 0.4 0.8 0.2 0.1 0.1 0.2 0.2 0.1 0.1

0.2

1.08 0.26 0.79

0.20 0.18 0.33 0.38

Br (mg/L)

−84 −77 −88 −86 −72 −83 −81 −87 −85 −83 −85 −85 −85 −82 −89 −73 −88 −82 −85 −89 −92 −87 −89 −91 −86 −88 −90

−11.0 −11.7 −11.9 −11.4 −11.6 −11.8 −11.0 −11.4 −11.7

−66 −70 −71 −76 −77 −76

−64 −61 −66 −68 −67 −43 −56 −65

δD (‰)

−10.8 −9.4 −11.3 −10.9 −8.6 −10.9 −10.4 −10.9 −10.9 −10.2 −11.0 −10.8 −10.9 −10.1 −11.7 −9.0 −11.6 −10.2

−7.2 −8 −8.3 −9.8 −10.3 −10.7

−6.0 −6.2 −7.3 −8.5 −7.6 −1.7 −5.6 −6.9

δ18O (‰)

Table 1. Geochemical data, surface water, and groundwater.

1.4 1.4 ,0.4 5.5 ,0.9

−9.2 −7.3 −9.6 −7.4

−4.1

−8.9 −7.5 −9.1 −8.7 −7.8

4.9 7 6.1

7.1

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 195–207 3.3

0.8

,0.9

3.0 ,0.6

,0.5 2.3

−8.8 −9.6 5.8

3.3 3.5 7.2

1.2

−9.3

−2.3

7.3

7.5 5.9 7.9 7.2 1.2 ,0.5

7.4 7.0 7.8

Tritium (TU)

−8.9

−8.3 −6.9 −8.5 −8.5 −9.3 −9.4

−7.4 −8.0 −8.7 −9.8

δ13C (‰)

3.2

1.1 2.9 3.3 7.3 8.8 7.4

4.3 2.9 5.8 2.7

2.6 2.7

δ34S (‰)

70.2 67.3

57.9 73.4 80.2

64 82.1

81.6

92.3

110.2 101.4 96.7 71.7 14.7 7.6

C (pMC)

14

80.2 107.7

98.3 73.4 132.1

74.7 82.1

86.2

105.5

143.5 195.8 120.3 89.2 nc nc

Corr.

C (pMC)

14

1,770 post-bomb

141 2,484 post-bomb

2,347 1,584

1,191

post-bomb

post-bomb post-bomb post-bomb 916

Age (yr B.P.)

Eastoe, Granados Olivas, and Hibbs

Date

nc 5 not corrected (see text).

JMAS 130 2003 JMAS 134 2003 JMAS 136R 2003 JMAS 141 3/18/2004 JMAS 142 2003 JMAS 151 2003 JMAS 161 3/18/2004 JMAS 165 2003 JMAS 176 2003 JMAS 180 2003 JMAS 183 2003 JMAS 186 2003 JMAS 193 3/18/2004 JMAS 194 2003 JMAS 221 2003 Springs, Sierra de Juárez Anapra Spring 3/18/2004 Spring R062001A 6/19/2003 Spring R062002A 6/19/2003 Precipitation Spring 2004 Precipitation Summer 2004 Precipitation Autumn 2004 Precipitation Winter 2004–2005

Sample

Screen (m)

−8.9 −9.9 −8.4 −3.6 −7.1 −6.9

265

−106.488 −106.489

31.691 31.663

141

0.2

−9.5

−11.2 −11.0 −11.5 −10.1 −11.3 −11.8 −11.1 −11.9 −10.7 −11.7 −11.8 −11.8 −11.3 −11.3 −10.5

−106.596

0.1 0.2 0.1 0.2 0.2 0.1 0.1 0.1 0.1 0.0 0.1 0.1 0.1 0.1 0.4

31.733

93 116 60 436 126 88 110 97 143 100 68 87 99 87 242

163 327 48 233 208 75 179 80 143 85 51 62 74 62 255

δ18O (‰)

−106.381 −106.467 −106.431 −106.434 −106.469 −106.372 −106.456 −106.402 −106.448 −106.343 −106.425 −106.410 −106.378 −106.382 −106.464

Br (mg/L)

31.662 31.622 31.697 31.701 31.689 31.707 31.735 31.675 31.739 31.731 31.721 31.852 31.891 31.833 31.730

SO4 (mg/L)

Cl (mg/L)

Long. (uW)

Lat. (uN)

Table 1. Continued.

−74 −55 −16 −54 −49

−63

−65

−87 −82 −88 −82 −85 −90 −86 −92 −81 −90 −91 −90 −88 −86 −78

δD (‰)

2.4

−6.2

−7.6

0.6 9.3 8.7 5.1

0.9

33.3 14.2

,0.9 ,0.4 −5.4

7.1

3.1

50.1 39.5

,0.9 ,0.9

−6.3 −8.4 −7.9

7.9 6 6.3

37.9

60.1

68.1 25.6

,0.4

−8.5

6.7

41.3 51

34.5

C (pMC)

14

,0.5 ,0.7

1.3

Tritium (TU)

−6.1 −8.1

0 −8.8

δ13C (‰)

6.1 2.3

7.3

δ34S (‰)

nc nc

63.8 nc

84.7 nc

nc 69.7

nc

Corr.

C (pMC)

14

3,615

1,329

2,904

Age (yr B.P.)

Groundwater Mixing in Ciudad Juárez

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 195–207

199

Eastoe, Granados Olivas, and Hibbs

A&M University System, and at the Department of Soil, Water and Environmental Sciences at the University of Arizona. The data are listed in Table 1, and sample sites are shown in Figure 3A.

ISOTOPES IN PRECIPITATION AND RECHARGE SEASONALITY The δ18O and δD values of the seasonal aggregate rain samples plot close to the global meteoric water line (GMWL) of Craig (1963), and they are compared in Figure 4 with data for Tucson, AZ (400 km west of the study area), adjusted for altitude. Tucson, with an average ratio of summer (May–October) to winter precipitation of 1.7 (National Weather Service, 2015), is a better match for the study area (ratio 3.4, http://www. srh.noaa.gov/epz/?n5elpaso_monthly_precip) than the nearest International Atomic Energy Agency (IAEA) database site at Ciudad Chihuahua (ratio 7.9, http:// www.chihuahua.climatemps.com/). In semiarid places, data for a single year are not good estimates of longterm mean δ18O and δD. Therefore it is instructive to compare the Ciudad Juárez data with the long-term data set from Tucson. Rain in Ciudad Juárez appears to show a similar strong seasonal variation in isotope composition. Groundwater discharging from springs in the Sierra de Juárez provides a longer-term estimate of the isotope composition of rainwater infiltrating the mountain block. The values of δ18O and δD are lower than those in 2004–2005 rain and plot close to winter means for Tucson, adjusted to altitudes appropriate for the present study area. Therefore winter recharge appears to predominate over summer in the Sierra de Juárez, the small amount of winter rain notwithstanding. Rainwater from Ciudad Juárez contained 5.1 to Figure 3. Maps of Ciudad Juárez showing: (A) sample locations; (B) distribution of δ18O values; (C) distribution of δ34S values; (D) distribution of tritium; (E) distribution of 14C; and (F) distribution of the ratio Cl/SO4.

mass spectrometer equipped with a CostechH elemental analyzer for preparation of SO2. Tritium was measured in a Quantulus 1220H Spectrometer by liquid scintillation counting after electrolytic enrichment. Carbon-14 was measured by accelerator mass spectrometry at the National Science Foundation–Arizona Accelerator Facility, University of Arizona. Analytical precisions (1σ) are 0.08‰ (O), 0.9‰ (H), 0.15‰ (C), and 0.15‰ (S). Detection limits are 0.6 tritium units (TU; 3 H) and 0.2 percent modern carbon (pMC, 14C). Major anions were analyzed by ion chromatography at the Texas A&M Research and Extension Center at El Paso, Texas A&M AgriLife Research, The Texas

200

Figure 4. Plot of δD vs. δ18O for available seasonal rain data from Ciudad Juárez, in relation to seasonal averages from Tucson Basin, corrected for altitude (data from Eastoe et al., 2004) and isotope data for springs in the Sierra de Juárez.

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 195–207

Groundwater Mixing in Ciudad Juárez

Figure 5. Plot of δD vs. δ18O for groundwater from Ciudad Juárez (CJ). GMWL 5 global meteoric water line (Craig, 1963); SJ 5 Sierra de Juárez; RBEL 5 Río Bravo evaporation line; ellipses labeled A, B, and C refer to data groups for Hueco Bolsón groundwater (Eastoe et al., 2008). Numbers refer to municipal wells.

9.3 TU in 2004–2005 (Table 1), i.e., higher than the post-bomb mean of 5.2 TU for Tucson Basin since 1992 (Eastoe et al., 2010).

RESULTS Hydrogen and Oxygen Isotopes Figure 5 shows O and H isotope data for the JMAS wells and the springs. Well locations are classified into three geographic zones: near the Sierra de Juárez, near the Río Bravo, and “central” for all other JMAS wells as indicated in Figure 3A. The data are compared with data clusters proposed in Eastoe et al. (2008) for the Hueco Bolsón regional aquifer, and they are plotted relative to the GMWL and the Río Bravo evaporation line (based on data from Phillips et al., 2003). Cluster A, mixtures of pre-dam and post-dam river water, is found in the Río Bravo floodplain (Figure 3B). Cluster B corresponds to pre-dam river water. It is found in all three geographic zones of Ciudad Juárez. Cluster C, resembling local winter precipitation recharged along the western boundary of the Hueco Bolsón (Eastoe et al., 2008), is present only in the springs of the Sierra de Juárez.

Samples from near the river are mixtures of pre-dam and post-dam river water, plotting close to the presentday evaporation trend of the river. Samples from near the Sierra de Juárez appear to form a linear array in Figure 5, indicating mixing between pre-dam river water having δ18O , −11.5‰ and mountain-derived water. Three other samples, from well 134 (near the river), from well 176 (towards the mountains), and from a Sierra de Juárez spring, also plot on the mixing

trend. In the case of well 134, the cluster C water involved is more likely to be from the Hueco Bolsón north of the Río Bravo than from the Sierra de Juárez.

The spring sample cannot be a natural mixture containing river-derived water because recharge in this case is at elevations higher than the river. An alternative explanation could be human-caused addition of water from the city supply.

Samples classified as central are almost all cluster B water with δ18O ranging from −11.9‰ to −10.2‰; these plot along an evaporation trend slightly below the present river trend. The spatial distribution of O isotopes is illustrated in Figure 3B. Among central samples, those in the southeastern part of the city have the lowest δ18O values.

Sulfur Isotopes The data are presented as plots of inverse SO4 concentration versus δ34S (Figure 6A and B) and as a map (Figure 3C). Values of δ34S in Río Bravo surface water (see Table 1) are negative upstream of, and positive downstream of, the Elephant Butte reservoir.

Data for the central group of samples form a horizontal array with δ34S values from +2‰ to +8‰, with the exception of two outlying samples with negative δ34S values. Near-river samples form an array extending between present-day river water and the high-δ34S end of the central group. Samples from near the Sierra de Juárez have a δ34S range of +5.8‰ to +7.1‰, and those from springs in the mountains have δ34S values of +2.4‰ and +3.1‰. Deeper samples at the USGS well nest (with δ34S . +7‰) resemble the central group or groundwater from the Hueco Bolsón north of the Río Bravo (data of Druhan et al., 2008; shown in Figure 6B). Shallower samples resemble river water.

Wells 3Z and 120 have δ34S values like those of present-day river water.

The river data, linked sequentially, show the effect of sulfate reduction in Elephant Butte reservoir, resulting in increased δ34S in the reservoir and downstream.

Pre-dam river water entering the Hueco Bolsón, particularly during spring floods, would have resembled river water now found upstream of the Elephant Butte reservoir. Pre-dam river water entering the basin at times of low flow may have resembled the presentday river samples shown in Figure 6A, which were taken at times of very low flow and were dominated by sulfate added at the terminus of Mesilla Valley.

Figure 6B shows mixing trends between both types of water and native Hueco Bolsón groundwater (data of Druhan et al., 2008). Most central samples plot along the pre-dam flood mixing trend, with dominant sulfate from native Hueco Bolsón groundwater.

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 195–207

201

Eastoe, Granados Olivas, and Hibbs

Figure 7. Plots of (A) tritium and (B) 14C vs. δ18O for groundwater from Ciudad Juárez and the U.S. Geological Survey well nest (Figure 3A). Point labels refer to municipal well numbers.

Tritium

Figure 6. Plots of inverse sulfate concentration (100/SO4) vs. δ34S for groundwater from Ciudad Juárez (CJ) and the Sierra de Juárez (SJ), and surface water from the Río Bravo (RB). NM 5 New Mexico, HB 5 Hueco Bolsón. (A) Data from this study. (B) Interpretation in terms of possible mixing trends between surface water and groundwater from the Hueco Bolsón north of the Río Bravo (Druhan et al., 2008).

The two central group samples with negative δ34S values were taken beneath the oldest part of Ciudad Juárez, and may represent contamination. Alternatively, they could represent sulfate from the river far upstream, concentrated by evaporation, or oxidation of sulfide in floodplain sediment. No pyritic sediment has been observed in the Hueco Bolsón. However, in Albuquerque Basin, Plummer et al. (2004) presented evidence of sulfide oxidation induced by pumping of oxygenated groundwater into previously anoxic, pyritic sediment.

202

Figure 3D shows that finite tritium values . 0.6 TU are limited to the Río Bravo floodplain, with the exception of well 53R, near the Narrows. Tritium levels comparable to those of contemporaneous river water (7 to 8 TU) were found only at the USGS well nest and at JMAS well 3Z (Table 1); elsewhere in the floodplain, tritium levels were no higher than 3 TU. Southwest of the floodplain, a few groundwater samples have finite tritium at levels , 0.6 TU (Table 1), indicating infiltration of a small amount of surface water, as occurs elsewhere in aquifers in Basin and Range alluvium, e.g., Tucson Basin (Eastoe et al., 2004). Figure 7A shows that water bearing no tritium has a δ18O range from −11.9‰ to −10.5‰. In water containing finite tritium, δ18O correlates generally with tritium content, substantiating the mixing relationship between post-dam and pre-dam river water, also indicated by O and H isotopes. River water from 2003–2004 had lower tritium than predicted by the groundwater mixing trend, consistent with higher tritium in Río Bravo surface water at the time of the bomb tritium pulse.

These data suggest that part of the groundwater present beneath the Río Bravo floodplain has infiltrated in the last few decades, but the samples for this study are mixtures of such water with pre-bomb

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 195–207

Groundwater Mixing in Ciudad Juárez

water. Pre-bomb recharge supplied most of the groundwater southwest of the floodplain.

Carbon-14 Figure 3E is a map of 14C distribution in Ciudad Juárez groundwater. The highest values (70–92 pMC) occur beneath the floodplain, with an exception in well 134. Outside the floodplain, values range from 14 to 70 pMC. Figure 7B shows that 14C and δ18O generally decrease together for values of pMC . 50. Below that value, δ18O increases with decreasing pMC. Correction of 14C Data The procedure for correction of 14C data was adapted from that in Clark and Fritz (1997, p. 210).

In recharge water, 14C is commonly derived from soil gas, which contains CO2 derived from decomposing plant matter and has a 14C concentration similar to that of the contemporaneous atmosphere. This 14C is diluted when the infiltrating water containing dissolved CO2 dissolves rock calcite, as in Eq. 1: CaCO3 þ CO2 þ H2 O ¼ Ca2þ þ 2HCO 3:

(1)

If the rock calcite has a consistent δ13C value (δ13Ccarb) and contains no 14C, the added bicarbonate will have δ13C near δ13Ccarb, and a 14C content of 0 pMC. A dilution factor, q, for the 14C may be defined as in Eq. 2: q ¼ a14 CDIC =a14 Crech ;

0vqv1;

(2)

where the a14C terms are activities of 14C expressed, for instance, as pMC; rech is recharge; and DIC is dissolved inorganic carbon. Using an isotope balance based on δ13C data, q may also be expressed as in Eq. 3: q¼

d13 CDIC d13 Ccarb : d13 Crech d13 Ccarb

(3)

The value δ13Crech is commonly considered to derive from that of soil CO2, which in turn originates from decaying plant matter. However, if recharge occurs from the saturated bed of a large perennial stream (as the Río Bravo appears to have been prior to development), an alternative approach to estimating δ13Crech arises. Saturated alluvium contains little soil gas. In such a case, δ13Crech is δ13C of the river water, and a14Crech is the typical 14C content of the river water. This approach is used here, invoking the following assumptions.

Figure 8. Plot of δ13C vs. 14C in dissolved inorganic carbon of groundwater from Ciudad Juárez and the U.S. Geological Survey well nest (Figure 3A).

1. The rock carbonate, in this case, calcite in the basin-fill sediments, has average δ13Ccarb values near –4‰ (Monger et al., 1998).

2. The value of δ13C of DIC in river water can be estimated by examining the range of δ13C in DIC of groundwater, as a function of pMC (Figure 8).

The range −9.6‰ to −6.4‰ appears largely to be independent of pMC, and to indicate that dissolution of rock carbonate results in increases in δ13C in DIC of the infiltrating water. By this reasoning, the average δ13C of river water would be −9.6‰ or less.

Acceptable results cannot exceed the peak pMC of seasonally averaged atmospheric CO2 during the bomb spike of the 1960s, approximately 190 pMC (Burchuladze et al., 1989). The corrected pMC values in Table 1 include only one unreasonable value (196 pMC for a sample from the USGS well nest) when calculated using a value of −9.6‰ for δ13Crech. The use of lower values of δ13Crech generates higher values of corrected pMC. The calculation of bulk groundwater residence ages from the pMC values requires an estimate of a14Crech. No direct measurement was made on river water, but sample JL-49-21-324, from the top 11.5 m of alluvium next to the river, contained 110 pMC in 2003. C4 plant material in rural Arizona in 2002– 2003 contained 108 pMC (Eastoe, unpublished data), for which the corresponding atmospheric pMC is 109.2 (Saliège and Fontes, 1984), i.e., close to the 110 pMC of recharge just before the sampling date. Therefore, observation suggests that the 14C level in river water is close to that in the atmosphere and was close to 100 pMC before the bomb 14C spike. Ages shown in Table 1 were calculated for a14Crech 5 100 pMC and range from post-bomb near the river to 3600 years at the southern end of the study area.

The relationship between 14C and δ18O in samples with pMC , 50 can be explained by mixing between type B water and very old type C water with δ18O

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 195–207

203

Eastoe, Granados Olivas, and Hibbs

functions of Cl concentration. The ratios are commonly higher than those of 2003–2004 river water.

Cl/Br ratios are .1,000 in many cases, consistent with halite as the ultimate source of the halide anions.

These findings suggest that mixing of river water and native Hueco Bolsón groundwater can be discussed in terms of three end members:

Figure 9. Plot of anion ratios (Cl/Br and 1000 Cl/SO4) in Río Bravo surface water vs. location in km from the source. Data at times of high flow (“high”) are from Phillips et al. (2003), and data at times of low flow (“low”) are from this study.

values of −10‰ or greater. This set of samples includes wells 142, 176, and 221, all of which are close to the Sierra de Juárez, and all of which show independent evidence of mixing of type B water with type C water from the mountain range (Figure 5). In the case of well 134, the mixing is most likely with type C water from north of the Rio Bravo; this is discussed further below. The source of ancient water for well 194 is unclear. Correction of 14C data is not possible for such mixed waters using the method and assumptions outlined here; therefore no corrected values are given for them in Table 1 and Figure 3E. Anions Only SO4, Cl, and Br are considered here, because of their generally conservative behavior. Figure 9 shows anion ratio behavior in the present-day river for average flow conditions in winter 2001. Both Cl/ SO4 and Cl/Br increase downstream as a result of progressive evaporation, addition of deep basin groundwater, and leaching of salt from irrigated land (Phillips et al., 2003). Elephant Butte reservoir has a strong effect, particularly on Cl/SO4, which increases sharply at the reservoir and remains about constant between the dam and Ciudad Juárez. Under low-flow conditions, represented in a sample set (RG1 to RG8 in Table 1) taken in El Paso during this study, salinity and the anion ratios increase as a result of evaporation and the effect of additions of saline water at the southern terminus of the Mesilla Valley.

Concentrations of all three anions and their ratios vary widely in Ciudad Juárez groundwater (Table 1) and follow no clear spatial patterns (e.g., Figure 3F).

Figure 10A and B show anion ratio variations as

204

1. Pre-dam river water under flood conditions. As indicated by O, H, and S isotopes, this would have resembled present-day river water from near Albuquerque, NM. The composition used here (Cl 9.2, SO4 58, Br 0.037 mg/L) is typical of reaches just north of Albuquerque (Phillips et al., 2003; Plummer et al., 2004), with the location chosen in order to avoid perturbations due to the city.

2. Pre-dam or post-dam river water under low-flow conditions, strongly influenced by salty water emerging at the southern terminus of Mesilla Valley. Mixing between present-day river sample RG6, the sample with highest observed total dissolved solids (Cl 768, SO4 1,028, Br 1.08 mg/L), and Albuquerque river water yields curves labeled M1 in Figure 10A and B. Other present-day river samples (including river water from Elephant Butte reservoir) fall close to these curves.

3. Native Hueco Bolsón groundwater represented by the sample from JMAS well 134 (Cl 327, SO4 116, Br 0.19 mg/L). Mixing curves between river water sampled near Albuquerque and well JMAS 134 are labeled M2.

The anion data are consistent with derivation of solutes from flood-stage pre-dam river water mixed with low-flow stage river water and saline native Hueco Bolsón water. Several samples have elevated salinities (100–350 mg/L Cl) and anion ratios that plot near the M1 trends of Figure 10A and B, indicating that recharge under conditions of low flow in the river is important in addition to recharge of water when the river is in flood. Near-river samples with Cl . 265 mg/L are mixtures of low-flow river water and native Hueco Bolsón groundwater. Samples of lower salinity, both near the river and central, are mixtures of all three end members.

Some inconsistencies are observed between Figure 10A and B. Two samples (JMAS wells 130 and 180) that plot above trend M2 of Figure 10A might be explained by dissolution of halite, but the effect of surplus Cl is not evident in the Cl/SO4 ratios. Furthermore, large-scale dissolution of evaporite salts from sediment seems unlikely from the fluvial deposits beneath Ciudad Juárez. Samples JMAS 42R and 141 also plot in inconsistent fashion; in Figure 10A, they appear to be mixtures of river water and native

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 195–207

Groundwater Mixing in Ciudad Juárez

Figure 10. (A) Plot of Cl/Br vs. Cl for Río Bravo surface water and Ciudad Juárez groundwater. (B) Plot of Cl/SO4 vs. Cl for Río Bravo surface water and Ciudad Juárez groundwater.

groundwater, but in Figure 10B, they resemble lowflow river water.

DISCUSSION Complementary Uses of Isotope and Ion Analyses Isotope evidence indicates the Río Bravo as the principal source of recharge to the Hueco Bolsón aquifer beneath Ciudad Juárez, with pre-dam river water predominant, except in a few wells near the Río Bravo.

Clusters C (native Hueco Bolsón groundwater) and B differ in δ18O and δD by maximum amounts of 2.5‰ and 28‰, respectively (Figure 5). For pairs of samples that differ in δ18O or δD by d‰, and for which the isotope analytical precision is σ‰ (one standard deviation), the difference has a relative error RE, calculated as rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi h i 100 2 ð2rÞ2 RE% ¼ : (4) d For σ 5 0.08‰ (the analytical uncertainty for O isotopes), RE 5 9 percent for d 5 2.5‰ and 22% for d 5 1‰. Similar RE values apply in the case of H isotopes. In the data set from Ciudad Juárez, O and H isotopes may therefore be incapable of detecting

mixtures of less than 20 percent of type C to type B water. The anion concentration data for Ciudad Juárez groundwater traces smaller additions given the strong salinity contrast between the two water types.

As tracers, therefore, water isotopes and major anions perform complementary functions. Stable O and H isotopes make the clearest argument for the importance of pre-dam Río Bravo water in the potable water beneath Ciudad Juárez, for the presence of postdam Río Bravo water in limited areas near the river, and for large-scale mixing with native Hueco Bolsón groundwater. The anion data provide the clearest evidence for the mixing of smaller volumes of highsalinity water, both native groundwater and river water, into the samples taken for this study. In the northeastern part of Ciudad Juárez, such mixing results from the southward encroachment of salty groundwater from the Hueco Bolsón regional aquifer on the north side of the river, as shown in Eastoe et al.

(2010). In other cases, the mixing reflects the presence of native groundwater of high salinity beneath the potable groundwater throughout the city. Solutes from the native groundwater appear in the sample set where pumping of potable water has brought about upwelling of salty water, and where well screens extend to the original base of the freshwater zone.

Practical Results of this Research That almost all of the potable groundwater underlying Ciudad Juárez is recharged from the Río Bravo means that previous interpretations of groundwater origin and flow are incorrect. There are technical, engineering, and policy issues that make these findings relevant to management and administration of the water resources of the Hueco Bolsón. The relatively young potable groundwater beneath most of Ciudad Juárez (. 45 pMC) compared to counterparts north of the river, where locally recharged potable groundwater typically contains less than 20 pMC, implies a more dynamic flow system and shorter residence times in the aquifer beneath Ciudad Juárez. Undoubtedly, the relatively dynamic recharge flux beneath Ciudad Juárez was due to flows from the Río Bravo that were intermittently very high during spring runoff from the highlands in Colorado and northern New Mexico. The annual cycle of scour and fill in the Río Bravo channel combined with stream avulsion probably created highly permeable streambed materials that united with high stream hydraulic head to allow ample volumes of recharge to the Hueco Bolsón aquifer beneath Ciudad Juárez in pre-development times.

More recently, the channelization of the Río Bravo streambed, the grout lining of segments of the channel

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 195–207

205

Eastoe, Granados Olivas, and Hibbs

along the urban corridor, and the control of spring runoff by reservoirs in New Mexico have probably substantially reduced this important natural source of groundwater recharge. The over-pumping of the Hueco Bolsón, especially beneath Ciudad Juárez, has been exacerbated by human-caused changes to the natural recharge processes.

Natural recharge along the Río Bravo channel near the El Paso Narrows is now eclipsed by pumpinginduced infiltration from unlined areas of the Río Bravo and from Río Bravo alluvium in areas below the Narrows. The natural recharge still operates to some extent to affect natural groundwater flow paths and fluxes. These flow paths converge on areas known to contain industrial and agricultural contaminants.

Natural recharge, to the extent that it is allowed near the Narrows, may therefore act as a diluent of contaminated areas in the aquifer, including contemporary and legacy groundwater contaminant plumes.

Finally, our findings resolve a disagreement over the source of groundwater beneath Ciudad Juárez.

According to previous interpretations, the groundwater originated as pre-development recharge of internal runoff along the mountain fronts north of the international border. By this account, the groundwater moved south across the border and flowed beneath Ciudad Juàrez before turning southeast and eventually upwelling and discharging along the Río Bravo.

Furthermore, by this account the post-development cones of depression beneath El Paso in the United States capture water that would have flowed to Mexico under natural circumstances. Our study shows that such a hypothesis can no longer be defended.

CONCLUSIONS 1. Stable O and H isotopes show that potable groundwater in the Hueco Bolsón aquifer beneath Ciudad Juárez originated mostly as pre-dam (pre-1916) surface water from the Río Bravo. This water mixes with post-dam river near the Río Bravo, with water originating from the Sierra de Juárez locally around the base of the range, and with native Hueco Bolsón groundwater in the northeastern corner of the city.

2. Measurable tritium is largely limited to the Río Bravo floodplain, indicating residence times on the order of decades in the floodplain.

3. Southwest of the floodplain, 14C data indicate bulk groundwater residence times of up to 3600 years in samples consisting mainly of pre-dam river recharge. Corrections are not possible in samples for which the 14C and δ18O data together indicate 206

4.

5.

6.

7.

8.

mixing of pre-dam river water with native Hueco Bolsón groundwater.

Sulfur isotopes are consistent with widespread mixing of sulfate derived from native Hueco Bolsón groundwater into pre-dam and post-dam river recharge.

Anion rations (Cl/Br and Cl/SO4) indicate mixing between three end-member water types: pre-dam river water of low salinity recharged under highflow conditions, pre-dam or post-dam river water of high salinity recharged under low-flow conditions, and high-salinity native Hueco Bolsón groundwater.

The anion concentration data complement the stable O and H isotope data and allow the identification of mixing of high-salinity water into lowsalinity river recharge in amounts too small to detect with the isotopes.

In the sample set for this study, such mixing is human-caused and results either from the way wells are constructed, or from southward encroachment of salty water into the cone of depression beneath Ciudad Juárez.

The findings described in this paper have implications for the long-term management of the groundwater in the Hueco Bolsón aquifer beneath Ciudad Juárez.

ACKNOWLEDGMENTS

This work was funded by Sustainability of SemiArid Hydrology and Riparian Areas (SAHRA) under the Science and Technology Centers (STC) Program of the National Science Foundation, Agreement No. EAR9876800, and by Center for Environmental Analysis, Centers for Research Excellence in Science and Technology (CEA-CREST) under National Science Foundation Cooperative Agreement No. HRD0317772. The authors acknowledge the kind support of the Junta Municipal de Agua y Sanamiento de Ciudad Juárez, who provided access to their wells.

REFERENCES BURCHULADZE, A. A.; CHUDÝ, M.; ERISTAVI, I. V.; PAGAVA, S. V.; POVINEC, P.; SIVO, A.; AND TOGONIDZE, G. I., 1989, Anthropogenic 14C variations in atmospheric CO2 and wines: Radiocarbon, Vol. 31, pp. 771–776.

CLARK, I. D. AND FRITZ, P., 1997, Environmental Isotopes in Hydrogeology, CRC Press, Lewis Publishers, Boca Raton, Florida, 328 p.

CLIMATOLOGICAL LABORATORY, UNIVERSIDAD AUTÓNOMA DE CIUDAD JUÁREZ, 2014, Climatological Laboratory, Universidad Autónoma de Ciudad Juárez: Electronic document, available at http://www.uacj.mx/Clima CONNELL, S. D.; HAWLEY, J. W.; AND LOVE, D. W., 2005, Late Cenozoic drainage developments in the southeastern Basin and

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 195–207

Groundwater Mixing in Ciudad Juárez Range of New Mexico, southeasternmost Arizona, and western Texas. In Lucas, S. G.; Morgan, G.; and Zeigler, K. E. (Editors), New Mexico’s Ice Ages: New Mexico Museum of Natural History and Science Bulletin 28, pp. 125–150.

CRAIG, H., 1963, Isotopic variation in meteoric waters: Science, Vol. 133, pp. 1702–1703.

DEL HIERRO-OCHOA, J.; GRANADOS-OLIVAS, A.; DOMINGUEZACOSTA, M.; GARCIA, T.; HIBBS, B.; AND EASTOE, C., 2004, Temporal variation of groundwater quality in the transboundary Paso del Norte Hueco Bolson: Evaluation of 45 years of data. In Proceedings of the Second International Symposium on Transboundary Water Management: Sustainability of Semi-Arid Hydrology and Riparian Areas, Instituto Mexicano de Technología del Agua, University of Tucson at Tucson, AZ, Nov. 2004.

DRUHAN, J. L.; HOGAN, J. F.; EASTOE, C. J.; HIBBS, B.; AND HUTCHISON, W. R., 2008, Hydrogeologic controls on groundwater recharge and salinization: A geochemical analysis of the northern Hueco Bolson Aquifer, El Paso, Texas: Hydrogeology Journal, Vol. 16, pp. 281–296.

EASTOE, C. J.; GU, A.; AND LONG, A., 2004, The origins, ages and flow paths of groundwater in Tucson Basin: Results of a study of multiple isotope systems. In Hogan, J. F.; Phillips, F. M.; and Scanlon, B. R. (Editors), Groundwater Recharge in a Desert Environment: The Southwestern United States: Water Science and Applications Series, Vol. 9, American Geophysical Union, Washington, D.C., pp. 217–234.

EASTOE, C. J.; HIBBS, B. J.; GRANADOS-OLIVAS, A.; HOGAN, J.; HAWLEY, J.; AND HUTCHISON, W. R., 2008, Isotopes in the Hueco Bolson aquifer, Texas (USA) and Chihuahua (Mexico): Local and general implications for recharge sources in alluvial basins: Hydrogeology Journal, Vol. 16, pp. 737–747.

EASTOE, C. J.; HUTCHISON, W. R.; HIBBS, B. J.; HAWLEY, J.; AND HOGAN, J. F., 2010, Interaction of a river with an alluvial basin aquifer: Stable isotopes, salinity and water budgets: Journal of Hydrology, Vol. 395, pp. 67–78.

HIBBS, B. J.; BOGHICI, R. N.; HAYES, M. E.; ASHWORTH, J. B.; HANSON, A. N.; SAMANI, Z. A.; KENNEDY, J. F.; AND CREEL, R. J., 1997, Transboundary Aquifers of the El Paso/Ciudad Juárez/Las Cruces Region. Contract Report, Texas Water Development Board: Texas Water Development Board & New Mexico Water Resources Research Institute, Las Cruces, NM, 148 p.

HIBBS, B.; PHILLIPS, F.; HOGAN, J.; EASTOE, C.; HAWLEY, J.; GRANADOS, A.; AND HUTCHISON, B., 2003, Hydrogeologic and isotopic study of the groundwater resources of the Hueco Bolson aquifer El Paso/Juárez area: Hydrological Science and Technology, Vol. 19, No. 1–4, pp. 109–122.

INSTITUTO MUNICIPAL DE INVESTIGACION Y PLANEACION (IMIP), 2003, Plan de Desarrollo Urbano de Ciudad Juárez 2003:

Ayuntamiento de Juárez, Chihuahua. Instituto de Investigación y Planeacion (IMIP) Ciudad Juárez, Chihuahua, México, 490 p.

INSTITUTO NACIONAL DE ESTADISTICA Y GEOGRAFIA Y GOBIERNO DEL ESTADO DE CHIHUAHUA (INEG y GEC), 1999, Estudio Hidrológico del Estado de Chihuahua: Instituto Nacional de Estadística, Geografía e Informática, Ciudad Chihuahua, 222 p.

LEOS RODRIGUEZ, A., 2004, Modelo Digital Conceptual Geohidrologico del Bolsón del Hueco en Ciudad Juárez, Chihuahua: Unpublished Thesis, Master’s Degree Program on Environmental Engineering and Ecosystems. Department of Civil and Environmental Engineering, University of Ciudad Juárez, Juárez, Mexico, 86 p.

MONGER, H. C.; COLE, D. R.; GISH, J. W.; AND GIORDANO, T. H., 1998, Stable carbon and oxygen isotopes in Quaternary soil carbonates as indicators of ecogeomorphic changes in the northern Chihuahuan Desert, USA: Geoderma, Vol. 82, pp.

137–172.

NATIONAL WEATHER SERVICE, 2015, Climate report, National Weather Service, Tucson, AZ. w2.weather.gov/climate/getcli mate.php?wfo5twc PAYNE, B. R., 1976, Environmental isotopes as a hydrogeological tool. In Arbeitstagung “Isotope in der Natur”: Zfl-Mitteilungen 5, Akademie der Wissenschaft der DDR, Zentralinstitut für Isotopen- und Strahlenforschung, Leipzig, Germany, pp.

177–199.

PHILLIPS, F. M.; HOGAN, J.; MILLS, S.; AND HENDRICKX, J. M. H., 2003, Environmental tracers applied to quantifying causes of salinity in arid-region rivers: Preliminary results from the Río Bravo, southwestern USA. In Alsharhan, A. S. and Wood, W. W. (Editors) Water Resources Perspectives: Eval‐ uation, Management and Policy: Elsevier Science, Amsterdam, pp. 327–334.

PLUMMER, L. N.; BEXFIELD, L. M.; ANDERHOLM, S. K.; SANFORD, W. E.; AND BUSENBERG, E., 2004, Geochemical Characterization of Ground-Water Flow in the Santa Fe Group Aquifer System, Middle Río Grande Basin, New Mexico: U.S. Geological Survey Water-Resources Investigation Report 03-4131, 395 p.

SALIÈGE, J.-F. AND FONTES, J.-C., 1984, Essai de détermination expérimentale du fractionnement des isotopes 13C et 14C du carbone au cours de processus naturels: International Journal of Applied Radiation and Isotopes, Vol. 35, pp. 55–62.

STREETER, L. A., 1990, Analysis of Metals in Well Water, Tap Water and Surface Water from Ciudad Juárez, Mexico: Unpublished M.S. Thesis, University of Texas, Houston, TX, 117 p.

STUART, C. J. AND WILLINGHAM, D. L., 1984, Late Tertiary and Quaternary fluvial deposits in the Mesilla and Hueco Bolsons, El Paso area, Texas: Sedimentary Geology Vol. 38, pp. 1–20.

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 195–207

207

Landscape Change as Recorded by the Ocean Shore Railroad MATTHEW A. THOMAS1 KEITH LOAGUE Department of Geological Sciences, Stanford University, 450 Serra Mall, Building 320, Stanford, CA 94305-2115

Key Terms: Ocean Shore Railroad, California, Coastal Landscape Change, Erosion, Geographic Information Systems

ABSTRACT This study focuses on the eroding bench of the Ocean Shore Railroad (1907–1920) along the central California coast. Investigation of the remnant feature demonstrates the impacts of landscape change on the human timescale.

Topographic maps and georeferenced aerial photographs aid in the first digital registration of the coastal rail sections. Sixty locations at the northern end of the study area provide site-specific rates of erosion. The geographic information system–based effort reveals a 66.7 km route divided among steep hillslope- and terrace-dominated sections, where natural and anthro‐ pogenic processes have shifted land cover from railway to roadway, open, agricultural, and developed spaces.

For the 60 erosion assessment locations, only 25 percent of the 1928 rail bench width remains after 82 years, and, in areas repurposed for the Pacific Coast Highway (1936–1957), only 38 percent of the 1956 road bench width remains after 54 years. The highest estimates of erosion are greater than 0.6 m yr−1. These erosion values do not reflect episodic mass wasting, highlighting the limited utility of steady erosion rates in land-use decisions.

INTRODUCTION Seaside population centers in the United States host densities three times the national average (Crossett et al., 2004). These areas often rely on infrastructure that intersects hydrogeomorphic processes capable of inflicting severe damage (Griggs et al., 2005). Today, 1 Corresponding author present address: Sandia National Laboratories, 4100 National Parks Highway, Building A, Carlsbad, NM 88220; email: tmatthew@alumni.stanford.edu; phone: (650) 678-0375.

assessments for existing infrastructure are often conducted, and repairs implemented, amid a frenzied atmosphere of emergency relief (Scott, 2010). Most coastal-zone regulations use long-term erosion rates.

However, the steady erosion paradigm (Hampton et al., 2004) does not reflect episodic mass wasting.

Coastal landscape change can have impacts at the human timescale (Young, 2015).

The City of Pacifica in California (see Figure 1) gained national attention during the 1997–1998 winter season when 10 ocean-front homes along the 500 block of Esplanade Avenue were lost due to rapid bluff retreat (Pettit et al., 2014). Presently, apartment buildings at 320 and 330 Esplanade Avenue, the footings of which are undercut by the receding bluff, have been tagged and evacuated (Berton and Lee, 2009). South of Pacifica, landslides at Devil’s Slide over many de‐ cades have resulted in several closures of the Pacific Coast Highway (PCH). Slope instability at Devil’s Slide led to the permanent abandonment of 2.5 km of the PCH in 2013. The new inland double-bore tunnel cost $500 million (Thomas and Loague, 2014a).

The Esplanade Avenue and Devil’s Slide examples illustrate the rapidly changing landscape along the coast. Foreshadowing these problems was the Ocean Shore Railroad (OSR). The position and fate of the OSR (1907–1920) underline the problem of rapid landscape change.

The two objectives of this study were: (1) to document the former position of the OSR within a digital geographic information system (GIS) environment and (2) to detect and quantify changes in the bench for selected locations in the northern part of the study area. Numerous studies have been carried out to better understand landscape change in the area (e.g., Tinsley, 1972; Lajoie and Mathieson, 1998; Snell et al., 2000; Hapke and Reid, 2007; and Collins and Sitar, 2008).

The major contribution of this study is to quantify landscape change as recorded by the Ocean Shore Railroad. As recognized by Cosmos (1974), the OSR bench provides a marker that can be used to conduct an investigation in coastal erosion.

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 209–223

209

Thomas and Loague

historical bench width was compared against the modern bench width to calculate average erosion rates.

Ground surveys were conducted to establish a new datum for data gaps identified by this study, enabling ongoing quantitative observations.

STUDY AREA

Figure 1. Approximate location of the completed and uncompleted sections of the Ocean Shore Railroad (OSR). The route, which was originally intended to connect the communities of San Francisco (SF) and Santa Cruz via rail, relied on an approximately 42-km-long inland automobile shuttle service to ferry passengers between Tunitas Creek and Swanton, CA.

With respect to objective one, available topographic maps and aerial photographs were georeferenced in the ArcGIS environment to locate the former OSR. The approximate centerline path of the route was registered as a vector. The work reported here examines the first (known to the authors) continuous, digital registration of the OSR. Utilizing modern orthoimagery (U.S. Geological Survey, 2014a) and aerial light detecting and ranging (LiDAR) data sets (Hines, 2011; U.S. Geological Survey, 2014b), along-route land-cover and topographic characteristics were examined. With respect to objective two, 60 locations in the northern part of the study area, dominated by steep hillslopes, were selected to assess site-specific rates of erosion as recorded by the OSR or the PCH. The

210

The study area, formerly traversed by the Ocean Shore Railroad, consists of approximately 70 km of coastline intersecting San Francisco, San Mateo, and Santa Cruz Counties in central California (see Figure 1). A Mediterranean climate, averaging 700 mm of rainfall per year (National Oceanic and Atmospheric Administration, 2015), supports predominantly herbaceous and shrub/ scrub vegetation (Multi-Resolution Land Characteristics Consortium, 2014). As noted by Griggs et al. (2005), the study area, part of the Coast Ranges geomorphic province, owes much of its rugged topography to active tectonism and mass-wasting events facilitated by rainfall and wave attack. As shown in Figure 2, the topography, with which the OSR contended, can be broadly characterized as hillslope- or terrace-dominated areas. For this study, the eastern margin for the hillslope- and terracedominated areas was delineated by the ridgeline above the rail bench and the rail bench, respectively. The western margin of both areas was defined by the coastline. The surficial geology, illustrated in Figure 2, consists of: (1) sandstone/mudstone, (2) greenstone/basalt, (3) mélange, (4) alluvium, and (5) granodiorite/quartz monzonite (U.S. Geological Survey, 2015). The characteristics of these materials vary widely. For example, the Quaternary sedimentary deposits of the Merced and Colma Formations (i.e., in categories 1 and 4) are poorly consolidated, while the Cretaceous metasedimentary/metavolcanic rocks of the Franciscan Complex and granodiorite/quartz monzonite of the Montara Granodiorite (i.e., in categories 2, 3, and 5) are generally well indurated.

HISTORICAL CONTEXT Organized solicitation for the construction of a coastal railway linking the communities of San Francisco and Santa Cruz started in the early 1880s, more than two decades before grading ultimately began (Bridges, 1881). The San Francisco and Ocean Shore Railway Company, in their 1881 report, estimated that the transport of people, produce, and timber along the 130 km of track would net annual profits in excess of $700,000. Although the San Francisco and Ocean Shore Railway Company never material‐ ized as a formal builder, much of the proposed route was adopted by the Ocean Shore Railway Company,

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 209–223

Landscape Change Recorded by Ocean Shore Railroad

Figure 2. Elevation and surficial geologic profile corresponding to the eastern margin of the study area for the (a) Northern and (b) Southern Division of the Ocean Shore Railroad. Geologic characteristics were abstracted from U.S. Geological Survey (2015).

which first broke ground in September 1905 (Brandt, 1952). The ambitious effort, which commenced simultaneously from San Francisco and Santa Cruz, included plans for at least 44 stops, 34 trestles, and three tunnels (Wagner, 1974).

Early railway construction efforts were fraught with setbacks. Chief among them were the damages incurred by the 1906 San Francisco earthquake. For example, an estimated 1.2 km section of track was lost due to landsliding. Amid higher-than-anticipated construction costs, the company began passenger services in 1907 (Wagner, 1974). Figure 3 is a photograph of a south-bound train amid the rough coastal terrain.

By 1908, the Ocean Shore Railway extended south from downtown San Francisco to Tunitas Glen and north from downtown Santa Cruz to Swanton (see Figure 1). A Stanley Steamer bus route (42 km) connected the Northern and Southern Divisions. Weekend operations routinely hosted upwards of 6,000 visitors (Wagner, 1974; Hamman, 1980; and Hunter, 2004).

Financial strain spurred the 1911 transition from the Ocean Shore Railway Company to the Ocean Shore Railroad (Wagner, 1974). Annual stockholder reports, although toned in optimism, reveal tenuous circumstances. Despite increasing ridership and hauled freight, operating expenses were 83, 97, and 101 percent of the operating revenue in 1912, 1913, and 1914, respectively (Williams, 1913; Crosby, 1914; and Sutton, 1915). A significant factor that kept the budget unbalanced was landslide activity, which required non-routine maintenance. For example, a January 1916 landslide destroyed 1.6 km of track south of Pacifica (Brandt, 1952). The $300,000 repair, with the road bed unusable for 4 weeks (Brandt, 1952), constituted nearly half of the annual profits originally estimated in 1881 (Bridges, 1881).

Recurring maintenance issues limited profits and inhibited expansion. There is evidence of some forward grading and tunneling along the planned central section of the route, but the link between the Northern and Southern Divisions was never completed (Wagner,

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 209–223

211

Thomas and Loague

Figure 3. Photograph (ca. 1907 to 1920) of south-bound train of the Ocean Shore Railroad at Devil’s Slide (see Figure 4a). The approximate elevation of the grade is 75 m above mean sea level (image source: LaJoie, 1995).

1974). Formal abandonment of the OSR was approved in October 1920 (Hamman, 1980). Subsequently, the State of California began taking steps to acquire right-of-way from the OSR to support construction for the Ocean Shore Highway, later known as the Pacific Coast Highway.

METHODS Along-Route Digital Inventory Five steps were employed to digitally register and examine the coastal trace of the former OSR from Thornton Beach to Tunitas Creek and Santa Cruz to Scott Creek. In step one, seven maps and 56 aerial photographs were collected. Several of the topographic and right-of-way maps document planned and built sections of the route (Goode et al., 1902; Thompson et al., 1915; and Baker, 1916). The maps provide a continuous profile of the former railway, although their spatial resolution is not easily amenable to meter-scale georeferencing. The earliest aerial photographs of the study area are from 1928, approximately 8 years after the OSR was abandoned. In step two, aerial

212

photographs of the coastline dating from 1928, 1931, 1946, and 1948 (University of California–Santa Cruz, 2014) were georeferenced in the ArcGIS environment utilizing prominent anthropogenic and natural landscape features. On average, 10 control points were used for each photograph. In step three, the discernible remnants of the route were traced from the georeferenced aerial photographs to digitize the approximate centerline path of the former OSR. The resulting profiles, which show good agreement with available maps, are shown in Figures 4a and 5a. In step four, the vector lines for the Northern and Southern Divisions were replaced as points with 50 m spacing.

Along-route characteristics, including average slope, height above mean sea level, and nearest distance to shoreline, were calculated with elevation data from the 2010 American Recovery and Reinvestment Act Golden Gate LiDAR Project (Hines, 2011) and the 1999 U.S. Geological Survey National Elevation Dataset (U.S. Geological Survey, 2014b) for the Northern and Southern Divisions, respectively. In step five, April 2011 and June 2006 ortho-imagery (U.S. Geological Survey, 2014a) was employed to investigate alongroute shifts in land cover for the Northern and

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 209–223

Landscape Change Recorded by Ocean Shore Railroad

Figure 4. (a) Northern Division of the former Ocean Shore Railroad extending from Thornton Beach south to Tunitas Creek. (b and c) Summaries of modern land cover and topographic characteristics for, respectively, the hillslope- and terrace-dominated sections of the Northern Division.

Southern Divisions, respectively. The modern land cover was categorized as (1) roadway, (2) agricultural, (3) open, or (4) developed. Roadways were considered sections of the route that accommodate vehicular traffic without restriction. Open, reclaimed, and vacant zones were identified as unmanaged sections. Agricultural areas intersected or were directly adjacent to cultivated fields. Developed zones consisted primarily of residential and commercial structures. Modern snapshots of along-route topography and land cover are shown in Figures 4b and c and 5b.

Site-Specific Assessments of Erosion The quantitative assessment of erosion along the OSR bench focused on selected locations within the

Northern Division. Three criteria were used to assess the candidacy of locations for calculating rates of erosion. First, the bench section under scrutiny needed to be largely unmodified by anthropogenic influences since the timing of vehicular abandonment. Second, the width of the alignment had to be visible in imagery and measurable within a GIS framework. Third, if not completely destroyed, the ocean-bound edge and inland cut slope (if applicable) of the former route had to be identifiable in modern LiDAR and orthoimagery. Three sites within the steep, hillslope-dominated zones of the Northern Division were selected for this study: (1) Thornton Beach, (2) Rockaway Point, and (3) Point San Pedro (see Figure 4a).

The specific locations considered for the erosion estimates at Thornton Beach, Rockaway Point, and Point

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 209–223

213

Thomas and Loague

Figure 5. (a) Southern Division of the former Ocean Shore Railroad extending from Scott Creek south to Santa Cruz. (b) Summary of modern land cover and topographic characteristics for the terrace-dominated sections of the Southern Division.

San Pedro are shown in Figures 6, 7, and 8, respectively.

Seven steps were employed to estimate rates of erosion. In step one, aerial photographs taken just before, or soon after, all vehicular activity was abandoned were georeferenced in the same manner as the alongroute inventory. For the Rockaway Point and Point San Pedro sites, vertical aerial imagery taken in the summer of 1928 (University of California–Santa Cruz, 2014) was used to measure the historical OSR bench width (Wagner, 1974; Hamman, 1980). As mentioned previously, some sections of the OSR were regraded to accommodate the Pacific Coast Highway.

Vertical aerial imagery from September 1956 (U.S. Geological Survey, 2014a) was used to measure the highway bench that followed the OSR in the vicinity of Thornton Beach (Petersen, 1958). In step two, the along-route vector lines for the Thornton Beach, Rockaway Point, and Point San Pedro sites were replaced as points with 50 m spacing. At each point, a cross section spanning the historical bench was extracted from the summer 2010 LiDAR (Hines, 2011). In step three, the surviving ocean-bound edge of the OSR bench for each cross section was located, and the corresponding elevation was noted. In step four, the adjacent cut slope was identified and projected with a least-squares approach to reduce measurement errors associated with deposition on the

214

bench. In step five, the intersection of the horizontal line corresponding to the ocean-bound edge and the cut slope was calculated. In step six, the distance between the intersection point and the ocean-bound edge was recorded as the modern bench width. In step seven, erosion rates, r [LT−1] were calculated as: r¼

bi bm d

ð1Þ

where bi is the historical bench width [L], bm is the modern bench width [L], and d is the duration (in years) between the historical and modern bench width [T]. Table 1 provides a summary of the erosion rates for Thornton Beach, Rockaway Point, and Point San Pedro, which are reported in Figures 9, 10, and 11, respectively.

RESULTS Along-Route Digital Inventory The rail corridor is 66.7 km in length. As shown in Figures 4a and 5a, the route is split between the 45.1 km Northern Division and 21.6 km Southern Division. Within a century, natural and anthropogenic processes have shifted land cover from 100 percent

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 209–223

Landscape Change Recorded by Ocean Shore Railroad

Figure 6. Location map for the 34 cross sections used in the erosion assessment for the PaciďŹ c Coast Highway at Thornton Beach (see Figure 4a).

rail to 29 percent roadway, 31 percent open, 29 percent agricultural, and 11 percent developed.

Figure 4a shows that the Northern Division includes hillslope- and terrace-dominated zones. The hillslopedominated section, which constitutes 10.1 km of the coastal route (i.e., 15 percent of the total), is surrounded by rugged terrain with slopes averaging 26u.

The average elevation above mean sea level and nearest distance to shoreline are 46 m and 123 m, respectively.

As shown in Figure 4b, land cover has shifted from 100 percent railway to 16 percent roadway, 83 percent open, and 1 percent developed in 91 years. The terrace-dominated section, which constitutes 35 km of the coastal route (i.e., 53 percent of the total), is typified by gentle slopes averaging 5u. The average elevation above mean sea level and nearest distance to shoreline are 25 m and 356 m, respectively.

As shown in Figure 4c, land cover has shifted from 100 percent railway to 30 percent roadway, 27 percent open, 22 percent agricultural, and 21 percent developed in 91 years.

Figure 5a shows that the Southern Division is an entirely terrace-dominated zone. The 21.6 km coastal

section (i.e., 32 percent of the total) is typified by gentle slopes averaging 4u. The average elevation above mean sea level and nearest distance to shoreline are 22 m and 384 m, respectively. As shown in Figure 5b, land cover has shifted from 100 percent railway to 31 percent roadway, 13 percent open, 55 percent agricultural, and 1 percent developed in 86 years.

Site-Specific Assessments of Erosion For this study, 60 of the 79 locations satisfy the criteria established to estimate average annual rates of erosion. A record of the 79 locations is provided in the electronic supplementary material (ESM), Table ESM-1. Erosion estimates were made for 34, 11, and 15 locations at the Thornton Beach, Rockaway Point, and Point San Pedro sites, respectively. As shown in Table 1, 45 percent of the bench sections (OSR or PCH) were completely destroyed by 2010. Obviously, the erosion values for these sites represent conservative, steady-state estimates. The rates of maximum erosion for Thornton Beach, Rockaway Point, and

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 209–223

215

Thomas and Loague

Figure 7. Location map for the 11 cross sections used in the erosion assessment for the Ocean Shore Railroad (OSR) at Rockaway Point (see Figure 4a). April 2015 global positioning system (GPS) measurements along the remaining OSR bench edge are shown with white circles.

Point San Pedro are $0.6, $0.1, and $0.2 m yr−1, respectively.

The 2,000 m Thornton Beach site hosts the greatest range of erosion values, from zero to greater than 0.6 m yr−1. For the 34 locations shown Figure 9, the 54 year average erosion rate is 0.2 m yr−1, and the average intact bench width is 38 percent. Numerous deep-seated slope failures, including the Northridge Bluff landslide (Collins et al., 2007), have erased many sections of the route. Where the bench is discernible, several meters of colluvium obscure the original asphalt road surface.

The 500 m Rockaway Point site has an 82 year average erosion rate of 0.1 m yr−1. For the 11 locations shown in Figure 10, the average intact bench width is 6 percent. Deep-seated landslide headscarps and debris chutes are visible in the area. Unlike the Thornton Beach site, the route appears to have been predominantly affected by the dislocation of large blocks of rock. In two locations where the route is not completely absent (see Figure 10), it is possible to stand on the original base rock surface for the railroad.

216

The 1,000 m Point San Pedro site includes erosion values ranging from zero to greater than 0.2 m yr−1.

For the 15 locations shown in Figure 11, the 82 year average erosion rate is 0.1 m yr−1, and the average intact bench width is 38 percent. Sections of the bench at Point San Pedro have been displaced 4 to 5 m by deepseated landslides or have been scoured by debris flows.

To a lesser extent than the Thornton Beach site, colluvium obscures the original base rock surface for the railroad.

Perhaps the most striking result from this study is the minimal extent of the OSR that remains after less than a century. This is especially true in the hillslopedominated sections, where, on average, the OSR was perched twice as high and twice as close to the coastline as the terrace-dominated sections (see Figures 4b and c and 5b). For example, at the Rockaway Point and Point San Pedro sites, only 25 percent of the historical rail bench width remains after 82 years. Among the site-specific locations investigated, many sections of the OSR have been completely lost. Certainly, no extended section could presently accommodate rail

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 209–223

Landscape Change Recorded by Ocean Shore Railroad

Figure 8. Location map for the 15 cross sections used in the erosion assessment for the Ocean Shore Railroad at Point San Pedro (see Figure 4a).

traffic. As shown in Table 2, the remaining bench sections for the Thornton Beach, Rockaway Point, and Point San Pedro sites will likely be gone in another hundred years.

DISCUSSION Evolving Features of California’s Coastal Landscape The work reported here is the third part in a series of investigations focused on coastal landscape change in the San Francisco Bay area. In the first part, Thomas and Loague (2014a) calculated an average erosion value of 0.1 m yr−1 at Devil’s Slide (see Figure 4a).

The 145 year (i.e., 1866–2010) estimate matches the average value for nearby Point San Pedro (see Table

1). After the OSR was abandoned, sections of the rail bench at Devil’s Slide were utilized for the PCH. Given the topographic and geologic similarities among Point San Pedro and Devil’s Slide (Thomas and Loague, 2014b; Thomas et al., 2015), it is fortuitous that a greater extent of the former OSR was not exploited for the PCH, as was originally intended (Wilcox, 2010).

In the second part, Pettit et al. (2014) estimated an erosion value of 0.5 m yr−1 (over 68 years) for the bluffs that front Esplanade Avenue (see Figure 4a).

In the current study, a Trimble R10 Global Navigation Satellite System was employed to establish a new local datum for the area. As shown in Figure 12 and recorded in Table ESM-2, 505 point measurements with 1 m spacing were collected along the bluff-top

Table 1. Summary of erosion rates (m yr−1) for the three sites considered for this study.

Bench Is Partially Visible in 2010 1

Bench Is Completely Absent by 2010

Site

Cross Sections

Minimum

Maximum

Average

Cross Sections

Minimum

Maximum

Average

Thornton Beach2 Rockaway Point3 Point San Pedro3

21 2 10

0 0.1 0

0.5 0.1 0.1

0.1 0.1 0.1

13 9 5

$0.2 $0.1 $0.1

$0.6 $0.1 $0.2

$0.3 $0.1 $0.2

1

See Figures 6, 7, and 8.

Bench analyzed corresponds to the Pacific Coast Highway.

3 Bench analyzed corresponds to the Ocean Shore Railroad.

2

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 209–223

217

Thomas and Loague

Figure 9. Annual erosion rates and estimated bench remaining for the Pacific Coast Highway in the vicinity of Thornton Beach. Image heading, tilt, and altitude are 130u, 76u, and 104 m, respectively (source: Google, 2015). Note, key corresponds to rates calculated for locations where the bench is completely absent and for locations where the bench remains.

edge. The average distance to the former OSR route from the April 2015 point measurements near the 100 block is approximately 150 m (see Figure 12). Given the Pettit et al. (2014) erosion rate, 300 more years would pass before the OSR would have been affected by bluff retreat in the area east of Esplanade Avenue.

218

However, the new point measurements record up to 12 m of bluff retreat from April 2011 (U.S. Geological Survey, 2014a) to April 2015 along the 300 and 400 blocks of Esplanade Avenue, highlighting the reality of episodic bluff retreat for the area. Taken from a risk-averse perspective (i.e., a rate of 3 m yr−1), only

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 209–223

Landscape Change Recorded by Ocean Shore Railroad

Figure 10. Annual erosion rates and estimated bench remaining for the Ocean Shore Railroad in the vicinity of Rockaway Point. Image heading, tilt, and altitude are 136u, 82u, and 100 m, respectively (source: Google, 2015). Note: Key corresponds to rates calculated for locations where the bench is completely absent and for locations where the bench remains.

50 years would pass before the OSR would have been affected by bluff retreat in the same area.

Average Erosion Rates and the Legacy Data Problem This study indicates that average erosion rates, often utilized for setback requirements (Hampton et al., 2004), can be useful for relative comparison, but they should be interpreted with caution. These results show good agreement with the work of Cosmos (1974) and are commensurate with the strong disparity in topographic relief and bedrock strength characteristics between, for example, the Thornton Beach and Rockaway Point sites (see Figure 2a). Average rates do not resolve the event-based mass-wasting processes operating at the scale of tens to hundreds of meters.

For example, as shown in Figure 9, the Northridge Bluff landslide removed more than 100,000 m3 of sediment (Collins et al., 2007), erasing an entire bench section. It is worth noting that, only 100 m south of the landslide area, approximately 80 percent of the road bench width remains.

Any assessment of coastal landscape change is dependent on the resolution and representativeness of the database utilized as well as the skill and experience of the investigator (Carter and Woodroffe, 1994). As demonstrated with the criteria established to assess the candidacy with which to examine rates of erosion in the vicinity of the former OSR, great care was taken to not overstep the resolution of legacy data. For example, the width of the OSR bench, as photographed in 1928, is not readily discernible along the southern flank of Rockaway Point and, therefore, is

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 209–223

219

Thomas and Loague

Figure 11. Annual erosion rates and estimated bench remaining for the Ocean Shore Railroad in the vicinity of Point San Pedro. Image heading, tilt, and altitude are 118u, 80u, and 98 m, respectively (source: Google, 2015). Note: Key corresponds to rates calculated for locations where the bench is completely absent and for locations where the bench remains.

not included in the analysis of site-specific erosion. To ensure future investigators can monitor the OSR bench at the Rockaway Point site, the same global positioning equipment employed at Esplanade Avenue was used to establish a new local datum for the area. As shown in Figure 7 and recorded in Table ESM-3, 285 point measurements were made for three zones along

the 2015 remaining ocean-bound bench edge of the OSR at the Rockaway Point site. Even with the new coverage at the Rockaway Point site, 20 percent of the locations intended for estimates of erosion as part of this study (see Figures 6, 7, and 8) lack a datum to enable future investigation. In the future, unmanned aerial vehicles that can employ topographic imaging

Table 2. Projections of average bench width survivability for up to one century.

Projected Average Bench Width Remaining after 2010 (m) Site1

Average Bench Width Remaining (m)

Average Erosion Rate (m yr−1)2

+25 Years

+50 Years

+100 Years

Thornton Beach Rockaway Point Point San Pedro

9.3 3.6 6.7

0.1 0.1 0.1

6.8 1.1 4.2

4.3 0 1.7

0 0 0

1

See Figures 6, 7, and 8.

See Table 1.

2

220

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 209–223

Landscape Change Recorded by Ocean Shore Railroad

Figure 12. Location map of the April 2015 bluff-top edge global position system (GPS) measurements collected along the 100, 200, 300, 400, and 500 blocks of Esplanade Avenue in Pacifica, CA (2011 base image source: U.S. Geological Survey, 2014a).

techniques (e.g., structure from motion) could be useful in accessing the terrain more frequently and at lower cost than traditional LiDAR methods (Colomina and Molina, 2014).

SUMMARY AND CONCLUSIONS Agencies that are charged to safeguard human life and infrastructure along coastal zones must commonly rely on average erosion rates for shaping development policies (Hampton et al., 2004). Unfortunately, the perception that hydrologic and geologic processes slowly and steadily march along to shape the coastal landscape can be misleading. This study, which focuses on the OSR along the central California coast, provides a case study in human-timescale landscape change. The bench associated with the OSR is used here to detect shifts in land cover and quantify rates of erosion.

Topographic maps and georeferenced aerial photographs were employed to establish the first digital registration for the coastal sections of the OSR. Along-route land cover and topographic features were examined utilizing modern ortho-imagery and LiDAR data

sets. The trace of the Northern and Southern Divisions of the OSR, characterized by steep hillslope- and terrace-dominated areas, is shown to presently accommodate road, open, agricultural, and developed spaces.

Sixty locations along the Northern Division of the OSR or, if overprinted, the PCH, were selected to estimate site-specific erosion rates. The historical bench widths for the railway and highway were measured from 1928 and 1956 aerial photographs, respectively.

The modern bench widths were estimated from a 2010 LiDAR data set. Erosion rates for the Thornton Beach, Rockaway Point, and Point San Pedro sites were divided into two categories: (1) the bench is still partially visible, and (2) the bench is completely absent. The highest estimates of erosion recorded by the bench are greater than 0.6 m yr−1.

This study suggests the OSR was short-lived because of an inability to identify or design for select hazards along the hillslope-dominated zones of the Northern Division. For locations minimally disturbed by human activities, only 25 percent of the historical bench width remains after 82 years. Sections of the failed railroad that were later exploited for the PCH have been severely impaired after only 54 years. The remaining

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 209–223

221

Thomas and Loague

hillslope-dominated bench sections could be gone within another 100 years. There is considerable variability in the disruption of the OSR route. For example, bench sections that have been completely erased commonly flank those where greater than 90 percent of the historical bench width remains. The average erosion rates, which show good agreement with those of others (e.g., Cosmos, 1974), may be useful for relative comparison, but they do not capture the reality of episodic mass-wasting processes.

The OSR legacy affords a unique opportunity to investigate coastal landscape change in the context of impacted infrastructure. At the very least, the OSR provides food for thought regarding rational coastal development. Relative to coastal hazards, the OSR is a reminder that the human timescale cannot be ignored in land-use decision. The current expression of the coastal landscape is only a temporary feature of an evolving system.

ACKNOWLEDGMENTS The authors are grateful for the deep-records expertise of Steven Gendel and Jane Ingalls at the Stanford Branner Library and Carol Peterson and Joan Levy at the San Mateo County Historical Museum in Redwood City, CA. Spirited discussions with Ocean Shore Railroad enthusiasts John and Dwight Harbaugh and Keith and Mary Ann Kvenvolden provided a rich historical context for this project. The sizable vectorization effort undertaken by Daniel Javier Roda-Stuart is greatly appreciated. David Janesko facilitated ground-truthing measures in the field. The authors are most thankful to George Cosmos for his kindness and generosity. The authors also appreciate the comments of Professor J. David Rogers, Professor John J. Clague, and Dr. Roy J. Shlemon on earlier versions of this manuscript. The work reported here was supported by the Stanford UPS Foundation Endowment Fund.

REFERENCES BAKER, A. E., 1916, Right-of-Way and Track Map: Ocean Shore Railroad Company, San Francisco, CA, 17 p.

BERTON, J. AND LEE, H. K., 2009, Eroding cliff prompts Pacifica evacuation: Electric document, available at http://www.sfgate. com/bayarea/article/Eroding-cliff-prompts-Pacifica-vacuation3206742.php BRANDT, R., 1952, Ocean Shore Railroad: Western Railroader, Vol. 15, pp. 1–32.

BRIDGES, L., 1881, Report upon the San Francisco and Ocean Shore Railway Company, California: Unpublished report for Ocean Shore and Railway Company, Special Collections and University Archives, Cecil H. Green Library, Stanford University, 16 p.

CARTER, R. W. G. AND WOODROFFE, C. D., 1994, Coastal evolution: An introduction. In Carter, R. W. G. and Woodroffe,

222

C. D. (Editors), Coastal Evolution: Cambridge University Press, Cambridge, U.K., pp. 1–31.

COLLINS, B. D.; KAYEN, R.; REISS, T.; AND SITAR, N., 2007, Terrestrial LiDAR investigation of the December 2003 and January 2007 activations of the Northridge Bluff landslide, Daly City, California: U.S. Geological Survey Open File Report 20071079, 51 p.

COLLINS, B. D. AND SITAR, N., 2008, Processes of coastal bluff erosion in weakly lithified sands, Pacifica, California, USA: Geomorphology, Vol. 97, pp. 483–501.

COLOMINA, I. AND MOLINA, P., 2014, Unmanned aerial systems for photogrammetry and remote sensing: A review: ISPRS Journal Photogrammetry Remote Sensing, Vol. 95, pp. 79–97.

COSMOS, G. J., 1974, Wave Erosion Rates along the Abandoned Ocean Shore Railroad, San Mateo County, California: Unpublished M.A. Thesis, Department of Geography, San Francisco State University, CA, 120 p.

CROSBY, J. W., 1914, Second Annual Report of Ocean Shore Railroad Company for the Fiscal Year Ended December 31, 1913: Office of Ocean Shore Railroad Company, San Francisco, CA, 24 p.

CROSSETT, K. M.; CULLITON, T. J.; WILEY, P. C.; AND GOODSPEED, T. R., 2004, Population Trends along the Coastal United States: 1980–2008: National Ocean and Atmospheric Administration Coastal Trend Report Series, 54 p.

GOODE, R. U.; BARNARD, E. C.; MARSHALL, R. B.; AND SEARLE, A. B., 1902, California Santa Cruz Quadrangle: U.S. Geological Survey, Reston, VA. GOOGLE, 2015, Google Earth: Electronic resource, available at: https://www.google.com/earth/ GRIGGS, G. B.; PATSCH, K.; AND SAVOY, L. E., 2005, Living with the Changing California Coast: University of California Press, Berkeley, CA, 540 p.

HAMMAN, K., 1980, California Central Coast Railways: Pruett Publishing Company, Boulder, CO, 309 p.

HAMPTON, M. A.; GRIGGS, G. B.; EDIL, T. B.; GUY, D. E.; KELLEY, J. T.; KOMAR, P. D.; MICKELSON, D. M.; AND SHIPMAN, H. M., 2004, Formation, Evolution, and Stability of Coastal Cliffs— Status and Trends: U.S. Geological Survey Professional Paper 1693, 123 p.

HAPKE, C. J. AND REID, D., 2007, National Assessment of Shoreline Change, Part 4: Historical Cliff Retreat along the California Coast: U.S. Geological Survey Open-File Report 2007-1133, 57 p.

HINES, E., 2011, ARRA Golden Gate LiDAR Project: Electronic resource, available at http://online.sfsu.edu/ehines/arra_golden_ gate_lidar_project.htm HUNTER, C., 2004, Images of Rail Ocean Shore Railroad: Arcadia Publishing, Charleston, SC, 127 p.

LAJOIE, K., 1995, personal communication, U.S. Geological Survey, Meulo Park, CA. LAJOIE, K. R. AND MATHIESON, S. A., 1998, 1982–83 El Niño Coastal Erosion, San Mateo County: U.S. Geological Survey Open-File Report 98-041, 61 p.

MULTI-RESOLUTION LAND CHARACTERISTICS CONSORTIUM, 2014, National Land Cover Database: Electronic resource, available at http://www.mrlc.gov/ NATIONAL OCEANIC AND ATMOSPHERIC ADMINISTRATION, 2015, Climate Data Online: Text and Map Search: Electronic resource, available at http://www.ncdc.noaa.gov/ PETERSEN, L. M., 1958, Pacifica freeway new section bypasses old coastal highway: California Highways Public Works, MayJune 1958, Vol. 37, pp. 19–21.

PETTIT, M. M.; THOMAS, M. A.; AND LOAGUE, K., 2014, Retreat of a coastal bluff in Pacifica, California: Environmental Engineering Geoscience, Vol. 20, pp. 153–162.

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 209–223

Landscape Change Recorded by Ocean Shore Railroad SCOTT, J., 2010, Pacifica declares emergency to help fund bluff repairs: Electronic document, available at http://www.mercury news.com/news/ci_14420465 SNELL, C. B.; LAJOIE, K. R.; AND MEDLEY, E. W., 2000, Sea-cliff erosion at Pacifica, California, caused by 1997/98 El Niño storms. In Griffiths, D. V.; Fenton, G. A.; and Martin, T. R. (Editors), Slope Stability 2000: American Society of Civil Engineers, Denver, CO, pp. 294–308.

SUTTON, J. G., 1915, Second Annual Report of Ocean Shore Railroad Company for the Fiscal Year Ended December 31, 1914: Office of Ocean Shore Railroad Company, San Francisco, CA, 22 p.

THOMAS, M. A. AND LOAGUE, K., 2014a, Devil’s Slide: An evolving feature of California’s coastal landscape: Engineering Environmental Geoscience, Vol. 20, pp. 45–65.

THOMAS, M. A. AND LOAGUE, K., 2014b, Hydrogeologic insights for a Devil’s Slide–like system: Water Resources Research, Vol. 50, pp. 6628–6645.

THOMAS, M. A.; LOAGUE, K.; AND VOSS, C. I., 2015, Fluid pressure responses for a Devil’s Slide–like system: Problem formulation and simulation: Hydrological Processes, Vol. 29, pp. 1450–1465.

THOMPSON, A. H.; JOHNSON, W. D.; MARSHALL, R. B.; AND MCKEE, R. H., 1915, California San Mateo Quadrangle: U.S. Geological Survey, Reston, VA.

TINSLEY, J. C., 1972, Sea Cliff Erosion as a Measure of Coastal Degradation, San Mateo County, California: Unpublished report for Department of Geology, School of Earth Sciences, Stanford University, Stanford, CA, 88 p.

UNIVERSITY OF CALIFORNIA–SANTA CRUZ, 2014, Aerial Photograph Collection, McHenry Library, 1156 High Street, Santa Cruz, CA 95064.

U.S. GEOLOGICAL SURVEY, 2014a, EarthExplorer: Electronic resource, available at http://earthexplorer.usgs.gov/ U.S. GEOLOGICAL SURVEY, 2014b, National Elevation Dataset: Electronic resource, available at http://ned.usgs.gov/ U.S. GEOLOGICAL SURVEY, 2015, California geologic map data: Electronic resource, available at http://mrdata.usgs.gov/ WAGNER, J. R., 1974, The Last Whistle [Ocean Shore Railroad]: Howell-North Books, Berkeley, CA, 135 p.

WILCOX, G., 2010, personal communication, California Department of Transportation, 111 Grand Avenue, Oakland, CA 94612.

WILLIAMS, A., 1913, First Annual Report of Ocean Shore Rail‐ road Company for the Fiscal Year Ended December 31, 1912: Office of Ocean Shore Railroad Company, San Francisco, CA, 14 p.

YOUNG, A. P., 2015, Recent deep-seated landsliding at San Onofre State Beach, California: Geomorphology, Vol. 228, pp. 200–212.

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 209–223

223

Landslide Risk Reduction in the United States—Signs of Progress JEROME V. DE GRAFF1 Department of Earth & Environmental Sciences, California State University, Fresno, CA 93740, USA

WILLIAM J. BURNS Oregon Department of Geology and Mineral Industries, 800 NE Oregon Street #28, Suite 965, Portland, OR 97232, USA

VICKI MCCONNELL Geological Society of America, 3300 Penrose Place, Boulder, CO 80301, USA

Key Terms: Landslides, Risk Reduction, State Geological Surveys, United States

ABSTRACT Scientific literature is replete with papers describing advances in applying new technologies to the study of landslides and advances in our understanding of factors affecting landslide occurrence, distribution, and move‐ ment. Far fewer papers look at how this knowledge is implemented to achieve landslide risk reduction. State and local governments exercise the greatest control on how landslide risk reduction is accomplished. Data generated from a questionnaire to state geological surveys, review of their agency Web sites, and two case studies demonstrate that progress has been made during the last 20 years in implementing actionable policy changes and programs to achieve a reduction is landslide risk in the United States. Progress is evident from the pivotal role played by state and local government in the areas of: (1) developing guidelines and training, (2) increasing public awareness and education, (3) implementing loss reduc‐ tion measures, and (4) conducting emergency preparedness, response, and recovery. The number of states requiring registration for practicing geologists has nearly doubled during the last 20 years. This increase was accompanied by a shift from indi‐ vidual state tests as part of determining competency to a uniform national test. During this period, state geological surveys established Web sites as a primary means for promoting public awareness and edu‐ cation about landslide hazards. Measures reducing the impact of landslides have included providing 1

Corresponding author.

information specific to residents and landowners, as well as supporting land-use planning efforts by local governments. Many state geological surveys are involved with emergency management of landslide hazards.

INTRODUCTION On Wednesday, October 11, 1995, the organization then named the Association of Engineering Geologists (AEG; now Association of Environmental & Engineering Geologists) held free public forums across the United States to disseminate information on landslide hazards to practicing professionals and the general public (DeGraff, 1995). National Landslide Awareness Day, as it was called, coincided with the United Nations’ International Decade for Natural Disaster Reduction (IDNDR) Day. It consciously followed the style of teach-ins popularized during the late 1960s and early 1970s, including the first Earth Day. AEG regional sections organized local events ranging from outdoor exhibitions at the Washington Park landslide in Portland, OR, to a day-long forum featuring speakers in St. Louis, MO, addressing technical topics for specialists in the morning and general topics for the public in the afternoon. Public field trips to local landslides were conducted by local geologists in the San Francisco Bay area. These events generated informative landslide feature articles in local media including in the Portland Oregonian (Hill, 1995) and the Salt Lake Tribune (Siegel, 1995). Other articles in the print media were generated by press releases issued by the U.S. Geological Survey (USGS), AEG sections, and state geological surveys from Colorado to New

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 225–243

225

De Graff, Burns, and McConnell

Hampshire. State geological surveys participated in a number of other ways. They disseminated landslide information in their monthly non-technical publications like New Mexico Bureau of Mines and Geology’s Lite Geology (Haneberg, 1995) and issuance of new general publications such as one on landslides in Ohio by their Division of Geological Survey (Hansen, 1995). The Utah Geological Survey made a display in their Salt Lake City office where the public could peruse landslide publications they had produced.

In Denver, CO, the USGS’s National Landslide Information Center held a news conference and distributed selected publications about reducing landslide hazards. They also featured a lunch-time talk about long-term research at the Slumgullion Landslide, CO. During the 20 years since National Landslide Awareness Day, there have been many examples of an improved scientific understanding of landslide processes and a greater array of technology available for investigating landslide problems and mitigating their impact. While the relevant published literature is too numerous for a comprehensive listing, it is possible to highlight some significant advances. Scientific advances in understanding factors important to landslide occurrence include refining our knowledge of how precipitation initiates and promotes landslide occurrence and movement (Iverson and Major, 1987; Guzzetti et al., 2008; Priest et al., 2011; and McKenna et al., 2012). Other insights dealt with factors initiating particular types of landslides (Cannon, 2001; Stock et al., 2012). Still others have refined our knowledge of landslide movement and deposition (Major and Iverson, 1999; Petley et al., 2002; and Santi et al., 2008). Global positioning system (GPS), light detection and ranging (LiDAR), differential interferometric synthetic aperture radar (DinSAR and SAR), and other remote-sensing techniques and platforms have advanced the ability of researchers to measure and record aspects of movement, sediment transport, and other landslide process factors that had previously either eluded detection or were ineffectively measured (Tarchi et al., 2003; Schulz, 2007; Roering et al., 2009; Jomard et al., 2010; Stock et al., 2011; and Jaboyedoff et al., 2012). Some of these technological methods have improved the ability to monitor landslides in order to provide warnings useful in emergency response (Baum and Godt, 2010; Reid et al., 2012).

In addition to monitoring to warn of possible landslide occurrence, modeling has improved prediction of the likelihood or characteristics of future landslides (Guzzetti et al., 2003; Gartner et al., 2008; Cannon et al., 2010; and De Graff et al., 2015b).

Less clear is the extent to which public policy and awareness have changed during this same period.

Put another way: Has our ability to achieve landslide

226

hazard reduction kept pace with our increased knowledge of this natural phenomenon? Just as major disastrous landslides like the 1983 Thistle landslide in Utah and the 1985 Mameyes landslide in Puerto Rico called attention to issues of landslide risk at the time of their occurrence, the 2014 Oso landslide in Washington gives renewed urgency to the question of whether the past 20 years of refining our knowledge and improving related technology have contributed to additional risk reduction. While we may not be able to highlight direct linkage, we can examine efforts to incorporate current scientific information into policy to counter landslide risk through land-use planning, decision making, regulatory frameworks, and public education.

This paper will specifically attempt to assess whether efforts to institute actionable policy for reducing landslide hazard at the state and local levels mirror the trend in our improved scientific understanding. Positive trends are needed to compensate for the greater need arising from both population growth and urban expansion. The reason for the state and local emphasis in this analysis is based on the authority for making land-use planning decisions and their regulatory enforcement throughout the United States, which rests almost entirely in the hands of state and local government (Burby et al., 2000; Berry, 2005; Hart, 2005; Mader, 2005; Preuss et al., 2005; Baum et al., 2008; and De Graff, 2012).

National direction and actions directed toward landslide hazard reduction are certainly important and necessary to support these local efforts. In 1999, Congress passed Public Law 106-113 requiring the USGS to develop a comprehensive strategy to address the widespread hazard posed by landslides in the nation (Spiker and Gori, 2003). The resulting strategy was termed “a framework for loss reduction” and received review and support from the National Re‐ search Council (Committee on Review of the National Landslide Hazards Mitigation Strategy, 2004). The elements of this strategy are identified as: (1) research, (2) hazard mapping and assessments, (3) real-time monitoring, (4) loss assessment, (5) information collection, interpretation, and dissemination, (6) guidelines and training, (7) public awareness and education, (8) implementation of loss reduction measures, and (9) emergency preparedness, response, and recovery (Spiker and Gori, 2003; Wieczorek and Leahy, 2008).

We think it is significant that the National Research Council titled its review assessment, “Partnerships for Reducing Landslide Risk.” The strategy they reviewed identified new roles and partnership opportunities for state and local government and private individuals under each of the nine elements (Table 1 in Spiker and Gori, 2003). The importance of state and local

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 225–243

Landslide Risk Reduction in the United States

governments in accomplishing landslide risk reduction is made especially evident by their roles in the elements of: (1) developing guidelines and training, (2) increasing public awareness and education, (3) implementing loss reduction measures, and (4) fostering emergency preparedness, response, and recovery.

METHODOLOGY As noted in the introduction, the metrics for assessing landslide risk reduction need to reflect actions at the state and local government levels. Ideally, one of these metrics would be changes in monetary costs due to landslide occurrence. However, this type of assessment even at a state level has proven difficult to quantify comprehensively (Highland, 2012). Lacking consistent and clear cost figures, we looked for measures at the state and local government level that emphasize the national strategy elements.

We draw upon data from Internet searches and a questionnaire together with two case studies to yield qualitative and quantitative measures in our assessment of state- and local-level actionable policy for reducing landslide hazard. The data compiled focus on measures that ensure: (1) an appropriate level of training for professional geologists involved with landslide hazard assessment and loss reduction work, (2) ongoing efforts to educate the public about general and specific landslide hazards that might affect their lives and property, (3) actions taken to facilitate loss reduction measures, and (4) development of effective emergency preparedness, response, and recovery actions (per Table 1 in Spiker and Gori, 2003).

A review of publically accessible Web sites maintained by each state geological survey was completed during May and June 2015 and reviewed for changes in September 2015. Examination of the survey Web sites focused on identifying landslide information suit‐ able for the needs of: (1) the general public, (2) residents and landowners, and (3) an employee of local government with responsibility for land-use planning or emergency response and geological professionals who may be providing support to governmental entities. This provided an objective, albeit qualitative, measure of a state survey’s ability to use Internet technology to disseminate information for public education, professional development, and local government assistance. Each state geological survey Web site homepage was accessed using a common internet browser. On each homepage, a link to “landslides,” “geologic hazards,” “hazards,” or similar keyword was sought on the assumption that these would be known to an interested member of the public, government employee, or practicing professional. This link was used to open related Web pages with subsequent links

accessed depending on the options presented. It is important to note that this examination is suitable for understanding what is being communicated about landslides by a state geological survey’s Web site but is not necessarily a complete look at their efforts to understand and reduce landslide hazard.

While homepages commonly had a link to publications, this link was not accessed as part of the examination because publications represent a different media.

There were two exceptions. One exception was to identify published maps and documents in digital form that were reached via links found on the Web site. The other exception was to determine whether guidance documents for practicing geological professionals existed. Where the publications were searchable online, the search terms, “guidelines,” “geologic hazard reports,” and “preparing reports,” were used.

If the publications were not searchable online, the main publication list was downloaded in its portable document format (PDF) format, and the “find” function was used with the search terms to accomplish the search.

Series of questions were prepared to gather information on various aspects of landslide hazard assessment and prevention and compiled into a questionnaire.

This questionnaire was sent to each of the state sur‐ veys represented in the American Association of State Geologists (AASG). Twenty-nine questionnaires were returned, representing a 58 percent return rate.

Responses to questions directly relating to the topics in the following sections were compiled. The questionnaire results together with two relevant case studies, one drawn from a co-author’s recent experience and the other from published data, are used to further illustrate the issues of loss reduction and emergency response, respectively.

ANALYSIS AND RESULTS Guidelines and Training Geologists are the primary professionals identifying and defining landslide hazards and investigating factors important to mitigation designs and regulatory requirements. States play a key role in ensuring that competent geologic professionals are available for assessing and carrying out investigations suitable for avoiding or reducing landslide risk. Ensuring the competence of geologic professionals is usually carried out by requiring certification or licensure, which often involves continuing education requirements.

A simple measure for assessing the way in which states are ensuring the availability of competent, trained geologists over the last 20 years is to look at professional registration of geologists. Figure 1 is a

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 225–243

227

De Graff, Burns, and McConnell

graph showing the number of states requiring professional registration of geologists between 1956 and 2014. The cumulative upward trend in the number of states enacting laws requiring registration visibly increased between 1991 and 2002. With the November 2014 passage of a licensure act in New York, 32 states have or are implementing this requirement for geologists practicing within their state (Figure 2). Figure 2 compares states that implemented registration of geologists to a generalized map of landslide hazard in the conterminous United States. Clearly, not just states where significant landslide hazard exists have enacted registration laws. This likely reflects the fact that requiring practicing geologists to be registered or licensed is deemed necessary for a broad range of environmental and engineering geologic work, including water resource protection, mineral extraction, and construction. While landslide hazard is certainly important in order to ensure public safety, it is only one reason for requiring competency among practicing geologists.

A related change has occurred in the way in which states determine which geologists will be granted registration or licensure. States where registration was established early, like California, generally required both academic and experience requirements before a geologist could apply for registration. Registration for those who met the requirements was then based on their performance on a state-administered test.

Similar procedures were adopted as other states began registration of geological professionals. Testing varies among states in terms of both testing procedures and the substantive content of the tests. Those differences affect questions of reciprocity and associated issues for geologists registered in one state but needing to practice in another one. Some of the State Boards of Geology charged with the responsibility for registration came together to discuss these differences in 1988. This meeting and subsequent ones culminated in the 1990 formation of the Association of State Boards of Geology (ASBOG, 2015). One significant outcome from the formation of ASBOG is a stan� dardization of written tests used to determine if a candidate is competent for registration. ASBOG also provides a means for state boards to facilitate uniformity of registration procedures, including more standardization of experience required to qualify to test and register. As of November 2015, 30 states used the ASBOG test as part of their registration process (ASBOG, 2015).

Another measure of how the states can ensure competency of geological professionals is by looking at the guidance provided for the ways in which they should carry out their work. While such guidelines may exist for specific projects, such as the state of Oregon’s

228

guidelines for proposed energy-related facilities (see http://www.oregon.gov/energy/Siting/docs/rules/div22. pdf), we will focus on those broadly covering environmental and engineering project work. Prior to 1995, California, Colorado, and Utah issued guidance documents describing what information should be included in geologic reports (Shelton and Prouty, 1977; Slosson, 1984; and Association of Engineering Geologists-Utah Section, 1986). These guidance documents ranged in size from two-page notes with a brief explanation and annotated report outline to a 67 page tome. To date, California has issued additional guidance documents (CGS, 2013a, 2013b). The Colorado Geological Survey has incorporated their current report guidance into their Web site (see http://coloradogeological survey.org/land-use-regulations/report-requirements/).

In 1999, Oregon also issued Guideline for Preparing Engineering Geologic Reports. It was later updated and reissued as a 14 page document developed and adopted by the Oregon Board of Geologist Examiners (OSBGE, 2014). Because the search for documents, which provide guidance to practicing geologists, was confined to publications issued by state geological surveys, similar technical direction issued by State Boards of Geology overseeing registration or licensure may also exist (see http://www.dol.wa.gov/business/geolo gist/docs/georptguide.pdf). While it is not as comprehensive a measure nationally as the increased number of states requiring registration or their widespread adoption of a standardization of testing, the increase in updated or new guidance documents issued since 1995 is indicative of an effort to improve competency.

Registration of geological professionals can serve as a measure of how states have acted to ensure the availability of competent, trained geologists for involvement in landslide hazard reduction activities. From 1995 to present, there has been nearly a doubling of states requiring registration of practicing geologists.

Because many of these states do host a significant potential for adverse impact due to landslides, it supports the concept that states have made progress in ensuring practicing geologists within their jurisdiction are well qualified to address landslide hazard issues.

Similarly, this period coincides with a shift from individual state-prepared examinations as part of the registration process to adoption of a shared uniform examination (ASBOG, 2015). This uniformity in testing adds to the likelihood that geologists addressing landslide issues in different states are applying the same level of competency due to their training. There has also been an increase in the number of technical guidance documents for conducting engineering geology work applicable to landslides and slope instability.

Less clear is whether the issuing of such technical guidance documents for practicing geological

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 225–243

Landslide Risk Reduction in the United States

Figure 1. Cumulative graph of states requiring registration for practicing geological professionals between1956 and 2014. The year when implementation of registration occurred was used, which in some instances differed from the year the registration law was enacted. For example, the New York registration law was enacted in 2014, but the implementation will be in 2016 (ASBOG, 2015; NYSED, 2015).

professionals has kept pace with the increased number of states requiring registration.

Public Awareness and Education Whether a geologic researcher is searching for relevant scientific articles or the public is trying to find personally useful information, Internet search engines are often the initial place to begin (Wishard, 1998; DeGraff et al., 2013). These search engines electronically survey the Internet to identify Web sites where information, linked to key words entered by the user, can be found. So, it is not surprising that all 50 state geological surveys have publically accessible Web sites for distributing information to both practicing professionals and the general public. This means such Web sites are a primary mechanism for public awareness and education. At the time of National Landslide

Awareness Day, state geological surveys shared information with various interested parties mainly through publications, displays in public locations, and professional workshops. While these methods continue to be used, widespread computer accessibility and the ubiquity of Internet search engines make the state geological survey Web sites the current medium of choice for education and outreach. The present-day dominance of the Internet for this purpose is a consequence of privatization and governance efforts that have encouraged the widespread proliferation and accessibility of publicly accessible Web sites after 1995 (Leiner et al., 1997).

A recent example of the importance of the Internet occurred during the year after the Oso landslide.

The March 22, 2014, Oso landslide (Washington) stimulated interest in landslide hazard in the Pacific Northwest. Consequently, the Statewide Landslide

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 225–243

229

De Graff, Burns, and McConnell

Figure 2. Two maps of the conterminous United States comparing landslide potential (upper map) to those states requiring registration of practicing geological professionals (lower map).The upper map was prepared from digital files (Godt, 1997). The lower map was compiled from the same sources noted for Figure 1.

Information Database for Oregon (SLIDO) experienced a 400 percent increase in views compared to 2013. SLIDO was the 25th most visited of all Oregon Department of Geology and Mineral Industries (DOGAMI) interactive Web maps in 2013. This page climbed to second most visited in 2014 and continues to have higher page views in 2015 (Alison Ryan Hansen, DOGAMI, personal communication, September 17, 2015).

The questionnaire to AASG members had posed the question: “Do you conduct education and outreach regarding landslide hazards?” This question referred

230

to all media and was not limited to use of the Internet for this purpose. Of the 29 questionnaire respondents, nineteen (66 percent) stated they did do education and outreach to the general public about landslide hazards. Similarly, our review of state geological survey Web sites found that 32 (64 percent) of the 50 state geological surveys were providing some level of readily accessible landslide hazard information on their home Web page (Table 1). These results are suggestive that most state geological surveys providing landslide information to the public are currently making credible use of the Internet.

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 225–243

Landslide Risk Reduction in the United States Table 1. Categories of landslide hazard information found during examination of state geological survey Web sites. A through D represent information categories characterizing all 50 states. E and F in the lower table are two information types that may be provided.

Landslide Information A. Only identifies landslides as a hazard or includes a very brief or generalized description B. Provides information specifically for the general public,residents/landowners, and local government/practicing professionals C. Provides information for two of the above categorical groups D. Does not specifically mention landslide hazards E. Provides access to landslide mapping and/or state-wide database F. Makes available a homeowner’s guide to landslides or a listing of observable warning signs

No. of State Surveys 8

Surveys Listed by State* MD, MO, MT, NH, NM, NV, NY, OH

12

AL, AR, CA, CO, ID, KY, ME, NE, NC, OR, UT, WA

12

AK, AZ, IN, MA, ND#, NJ, PA, SC, VT, VA, WV, WY

18

CT, DL, FL, GA, HI, IL, IA, KS, LA, MI, MN, MS, OK, RI, SD, TN, TX, WI AL, CA, CO, ID, KY, MA, ME, NE, NV, NJ, ND, NC, OR, PA, SC, UT, WA, WV AL, AZ, ID, IN, NC, OR, UT, WV, WA

18 9

*U.S. Postal Service abbreviations for states used. # North Dakota provides only landslide maps on its Web site, which are mainly useful to local government/practicing professionals.

Of the 32 state geological survey home Web pages where landslide hazard information is accessible, eight surveys only make mention of landslides as being a hazard, provide very brief or general descriptions, or describe a very specific landslide-prone situation (Table 1). For example, the home Web page link to hazards for the New Hampshire Geological Survey only mentions landslides in the following statement: “While geologic hazards encompasses a multitude of natural phenomena, including landslides and earthquakes, flood inundation and river erosion constitute New Hampshire’s #1 natural hazard risks.”

The amount and detail of landslide hazard information and the potential user groups being addressed vary considerably among the remaining 24 state geological survey home Web pages offering more extensive content (Table 1). The Web site content was examined considering three user categories: the general public, residents and landowners, and local government staff and practicing professionals. Twelve state geological survey home Web pages offer access to landslide hazard information useful to all three of these user groups (Table 1). On Figure 3, these 12 states are indicated by their U.S. Postal Service abbreviation.

Table 2 explains how to access landslide hazard information on these state geological survey Web sites and provides thumbnail descriptions of the content available for the three different user groups. There are some noteworthy features among these 12 state geological survey Web sites. California, Colorado, and Utah Web sites all include information specific to landslide activity from areas burned during wildfires. Colorado provides brief, but very effective, videos of different landslide types narrated by their staff in a field setting. Online interactive interfaces for accessing landslide maps are present on the

Colorado, Kentucky, Maine, Oregon, and Washington Web sites. The Utah Web site has a high degree of interconnectedness among subheadings and content, ensuring that relevant information is not overlooked.

The landslide hazard content of the other 12 state geological survey home Web pages is not responsive to the needs of all three defined user groups (Table 1). Of these 12 Web sites, the North Dakota Geological Survey has the only home Web page that gives access to landslide hazard information primarily useful only to one potential user group—local government staff and practicing professionals. A link to Landslide Maps enables those users to open a page showing the entire state index of 1:100,000 scale mapping. Clicking on a sheet in the index opens a page showing whether there are landslide maps available and, if so, providing an index for opening them. The North Dakota Geological Survey is one of only 18 state geological surveys that offer public access to landslide mapping or a state-wide database of landslides (Table 1).

The high proportion of state geological survey Web sites providing landslide hazard information begs the question of why the remaining 18 (36 percent) state geological survey are not doing so on their Web sites.

Figure 3 enables comparison of 17 of these 18 states (excluding Alaska and Hawaii) within the conterminous United States to their relative landslide incidence and susceptibility. While there is not a similar map for Alaska and Hawaii, it is known that these states have both physical conditions favoring landslide occurrence and have experienced disastrous landslide events (Jibson et al., 2004; Deb and El-Kadi, 2009). For some states, such as Florida, Rhode Island, and Delaware, which show no landslide susceptibility or incidence,

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 225–243

231

De Graff, Burns, and McConnell

Figure 3. Overview map showing landslide incidence and susceptibility in the conterminous United States with the abbreviations for states added indicating those state geological surveys providing the widest range of landslide hazard information to the public (modiďŹ ed from Wieczorek and Leahy, 2008).

and Iowa, Kansas, Michigan, and Minnesota, which show only small areas of moderate susceptibility, not mentioning landslides as a hazard is understandable (Table 1). It is unclear is why states like Georgia, Oklahoma, and Tennessee, where there are either significant areas of high incidence and susceptibility or high susceptibility for landslides, make no mention of this potential hazard on their home Web pages. The fact that these states do not mention landslide hazard on their home Web pages may reflect their emphasizing mineral or groundwater issues, an internal financial limitation, or their being part of a larger agency where geologic issues are only a subset of the topics addressed via their agency home Web page. For example, Hawaii’s failure to mention landslide hazards may stem from the fact that their State Water Commission serves as the only state entity addressing geologic survey issues.

Both the response to the questionnaire and the examination of state geological survey home Web pages are indicative of a significant number of state surveys recognizing a need to provide education and outreach to various public groups about landslide hazards in their state. The Web site review demonstrates progress in this aspect of landslide hazard reduction since 1995. The results described here show that much of their effort comprehensively reaches potential user groups and makes effective use of changing technology.

232

Implementation of Loss Reduction Measures While important to landslide hazard reduction, the presence of well-trained practicing professionals and effective outreach and education to the public are insufficient to fully accomplish this task. There must be a response to the information provided by both knowledgeable professionals and the public to actually achieve loss reduction (DeGraff, 2012; DeGraff et al., 2015a). The state geological surveys provide a framework for this response as it is carried out at multiple levels involving state and local government. This perspective is supported by the responses to the questionnaire. Of the 29 AASG questionnaire respondents, only nine (31 percent) conducted their landslide hazard efforts in response to statutory regulation or mandates (Table 3). The questionnaire responses indicate that state geological surveys contribute to implementation of landslide loss reduction by working with other state agencies (15 [52 percent] of respondents) and federal agencies (16 [55 percent] of respondents; Table 3).

State geological survey involvement for achieving landslide loss reduction appears most active at the local community level (cities and counties). Of the 29 AASG respondents, 22 state geological surveys (76 percent) are working with local communities (Table 3). Local community involvement is important because it is only at this governmental level that effective regulation permitting the establishment of geologic hazard districts, promulgation of zoning

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 225–243

Survey Name

Geological Survey of Alabama

Arkansas Geological Survey

California Geological Survey

Colorado Geological Survey

Idaho Geological Survey

Kentucky Geological Survey

Maine Geological Survey

State

Alabama

Arkansas

California

Colorado

Idaho

Kentucky

Maine

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 225–243 http://www.maine.gov/dacf/mgs/ index.shtml Click on “Hazards” in left panel of home page. “Landslides” is 1 of 7 hazards categories listed

http://www.uky.edu/KGS/ Hover over “Research/Programs” in panel on left to choose “Other Hazards,” where “Landslides” is listed

http://www.idahogeology.org/ Click on “Geologic Hazards” in left panel of home page.

Landslides are 1 of 3 topics under “Geologic Hazards”

http://www.geology.ar.gov/home/ index.htm Click on “Geohazards” in the header bar. Landslides are included under “Geohazards” http://www.consrv.ca.gov/cgs/ Pages/Index.aspx Click on “Geologic Hazards” in the header bar. Landslides are included under “Geologic Hazards” http://coloradogeologicalsurvey. org/ Click on “Geologic Hazards” in the header bar for dropdown menu containing: “Debris flow-fan (mudflows),” “Landslides,” “Mudslides,” and “Rockfall”

http://www.gsa.state.al.us/ Click on “Geologic” on left panel of home page to reach Mapping & Hazards

Web Home Page and Path to Landslide information

There is a link to Survey notes on emergency response and hazards from specific landslide types or events such as wildfire; information specific to residents’ needs is provided by a link to a U.S. Geological Survey site The definition page and the case histories provide insight useful to residents and landowners; the video on the “Geologic Hazards” page is also instructive

A description of past and current survey landslide programs and general descriptions of landslide types

“Landslides” includes a virtual tour of Maine’s geologic hazards; answers are provided to several frequently asked questions

“Landslides” links to page with a map showing landslide locations in Kentucky

Case histories explain past major landslides as well as landslides in coastal areas and landslide susceptibility mapping; a form is provided for the public to submit information on local landslides

In addition to the general information for Idaho landslides, there are summaries of how landslides are studied, their relation to flood events, and landslide warning signs for residents and landowner’s to observe This page also provides many links to photos and information sources on landslides, mapping, and detailed reports

Another link provided reaches U.S. Geological Survey and FEMA sites useful to residents

“Landslides” includes a description and link to a general information page

Each category (debris flowfan, landslides, mudslides, and rockfall) includes a definition and description of damages attributable to that hazard type; the “Geologic Hazards” page also contains videos of a debris flow, a landslide, and a rockfall There is a general description of landslides, their causes, and relation to geology specific to Idaho

Lists of landslide warning signs and what to do before and after a slide movement

Resident/Landowner

Interactive descriptions of different slope instability

General

Landslide Information

An interface allows access online to landslide susceptibility maps for different areas and known landslide sites in the state

The Web site home page offers “Online Maps,” which includes “Landslide Information,” where a searchable map contains details on inventory maps

Further information is provided by links to a map of landslides in Idaho and a publication containing much of the same information found on the Web page

The rockfall category includes a rockfall event map of Colorado; the Landslide category includes a map tab that access a PDF map of landslides in Colorado and an interactive map viewer for the landslide map; it is possible to download geographic information system (GIS)–useable files

Landslides also has a link to maps and survey notes on landslide susceptibility and guidelines for geologic report preparations

State-wide maps of factors affecting slope stability and general landslide susceptibility with links to additional information and listing of historical landslides by county Survey-prepared landslide case studies listed by county are accessible from this link

Local Government/Practicing Professionals

Table 2. An alphabetical listing of the 12 state geological survey Web sites where landslide information is available for the general public, residents and landowners, and local government staff and practicing geological professionals. These sites were last accessed on August 31, 2015.

Landslide Risk Reduction in the United States

233

234

Survey Name

Conservation and Survey Division (Nebraska Geological Survey)

North Carolina Geological Survey

Oregon Department of Geology and Mineral Industries

Utah Geological Survey

State

Nebraska

North Carolina

Oregon

Utah

http://geology.utah.gov/ Click on “Hazards” in header bar to reveal a page titled “Earthquake & Geologic Hazards” with seven major subheadings including: Landslides & Rock Falls, Geologic Hazard Assistance, Geologic Hazards Technical Information, and A Guide to Homebuyers & Real-Estate Agents

http://www.oregongeology.org/ Click on “Hazards” in header bar

http://portal.ncdenr.org/web/lr/ geological_home Two ways are available to reach landslide information.

One is to click on “Geologic Hazards” under quick links and then choose “Landslides”; the other is to click on “Landslide Information” listed in the left-hand panel on the opening Web page

http://snr.unl.edu/csd/ Click on “Geology” and then on “Landslides in Nebraska” under the heading “Geologic Hazards”

Web Home Page and Path to Landslide information

A database of inventoried landslides in Nebraska; entries include type of slide with photos accompanied by panels giving details on location (township, latitude and longitude, topographic quad), classification (cause, geology, etc.), status (date of last movement, activity, dimensions etc.), and damage & repair (costs and comments) Links permit downloading of PDF versions of maps for the four counties where landslide mapping is complete; links to both other parts of “Landslide Information” and to other publications give background information on historical landslide events and important triggering storm events

The “Landslides” page offers access to an interactive map of landslides throughout the state as well as access to recent reports on investigations and mapping of landslides; a fact sheet is available for Understanding Landslide Deposit Maps

Maps and investigative reports are commonly grouped and accessed by county; many publications are able to be downloaded directly from links on the Web pages; certain important items such as guidelines for hazard reports and landslide investigation technical reports are included under two or more of these subheadings to facilitate their being located

A link of particular use to homeowners is called “How to recognize landslides and how to deal with them”; indicators of instability that a resident or landowner might observe and different mitigation possibilities for a variety of landslide types are provided

Among the resources on the “Landslides” page are downloadable versions of the Landslide Hazards in Oregon Fact Sheet and the Homeowner’s Guide to Landslides (produced by Portland State University)

Clicking on the subheading “A Guide to Homebuyers & Real-Estate Agents” reveals a wealth of information specific to both homeowners and realty agents; the subheading “Landslides & Rock Fall” also accesses a useful “how to reduce your risk” section

The “Landslide Information” home page is a topical introduction with a few sentences and accompanying photos; it also states counties that have completed landslide inventory maps and links to other pages with more detailed information; it also gives links to maps and publications The “Hazards” page provides an alphabetical list of hazards; click on “Landslides” to access a range of landslide resources including links to general descriptions and safety information for landslides and debris flows Click on the subheading “Landslides & Rock Fall” for general information and description of recent events in Utah; access to maps of landslide susceptibility and a landslide database is also gained from this site

Local Government/Practicing Professionals

Definitions provide diagrams, photos, and descriptions for each of the landslides types common to Nebraska; these are accessed by tabs, which also provide a Varnes classification diagram and a photo sequence of a developing landslide

Resident/Landowner

Landslide Information

Brief introduction to landslides with several photos and headings linked to “Definitions” and “Database”

General

Table 2. Continued.

De Graff, Burns, and McConnell

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 225–243

On the “Landslides” page, the link “what we do” gives access to the interactive geologic hazards maps; the geologic hazards maps are also accessible by clicking that icon on the “Geologic Hazards” page or in the topical listing on the left The “Landslides” page offers “landslide warning signs and triggers” and “reduce your risk” links useful to residents and homeowners; the link “what we do” provides access to two PDF documents with general information on landslide mechanisms, types, and past Washington State landslides On the “Geologic Hazards” page, click on the “Landslides” icon or in the listing on the left; the page offers links to general information under “why do landslides happen,” “types of landslides,” and “some historic landslides in Washington State” Department of Natural Resources Geology & Earth Sciences Washington

http://www.dnr.wa.gov/ programs-and-services/ geology-and-earth-resources Click on “Geologic Hazards” among Geology topics on the home page

Resident/Landowner General Web Home Page and Path to Landslide information Survey Name State

Table 2. Continued.

Landslide Information

Local Government/Practicing Professionals

Landslide Risk Reduction in the United States

ordinances addressing slope stability, and building permit restrictions in landslide-prone areas can be accomplished.

A measure of how state geological survey involvement at the local community level has changed since 1995 involves looking at landslide reduction efforts focused on homeowners. Nine state geological surveys make available landslide hazard guides to homeowners and buyers, or a list of non-technical, observable changes that might indicate landslide movement (Table 1). Several guides are actual publications that may be purchased in print form or downloaded in a PDF (Harris and Pearthree, 2002; Potter et al., 2013; and WDGER, 2015). In some cases, landslides are one of several geologic hazards covered by the publication (Harris and Pearthree, 2002) and others are specific to landslide hazard (Potter et al., 2013; WDGER, 2015). Potter et al. (2013) produced “Landslides and Your Property” under a collaborative effort by the Indiana Geological Survey, Kentucky Geological Survey, and the Ohio Division of Geological Survey.

Other Internet-accessible homeowner guidance and lists include: “Landslide Hazards and You” (Alabama Geological Survey), “Reducing Landslide Loss” (Idaho Geological Survey), “How to Recognize Landslides and How to Deal with them” (North Carolina Geological Survey), “Homeowner’s Guide to Landslides” (Oregon Geological Survey), “A Guide to Homebuyers & Real-Estate Agents” (Utah Geological Survey), and “Homeowner’s Guide to Geologic Hazards” (West Virginia Geological Survey), which are viewable on their respective state geological survey Web sites. In some cases, these Web-based guides are also freely downloadable in PDF format. Many of the guides and lists available only on state geological survey Web sites, if not all of them, were produced after 1995. Similarly, the three state geological survey–published homeowner guides have publication dates ranging from 2002 to 2015. This demonstrates that all of these efforts to communicate specific actions that the residents and landowners can take for landslide risk reduction were initiated during the last 20 years.

It is more difficult to fully gauge the extent of state geological survey involvement in achieving landslide risk reduction at the local level (communities and counties). The information provided on many state geological survey Web sites includes reports of damaging landslide incidents, which are certainly instructive for loss reduction efforts at the local level.

Similarly, there are many landslide maps (location of past landslide events or features, landslide susceptibility or incidence, and landslide hazards) and some studies prepared specifically for guiding land use and development (Lund et al., 2008). As noted on the

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 225–243

235

De Graff, Burns, and McConnell Table 3. Summary of responses to AASG questionnaire inquiries related to landslide loss reduction.

Questions A. How are your efforts related to statutory regulations or mandates (state, local, federal) policies, and land use?

B. Do you work with other state agencies on landslide hazards?

C. Are you working with federal agencies (e.g., USGS, FEMA)?

D. Are you working with local communities (e.g., cities, counties, other state agencies)?

Web pages of the California Geological Survey (http:// www.consrv.ca.gov/cgs/geologic_hazards/landslides/ Pages/Index.aspx), specific studies were undertaken by their staff that produced maps and guidance documents to help local governments deal with landslide hazard where urban expansion was happening or was expected to happen in the near future. An example of the work done is Geology for Planning in Western Marin County, California, by David L. Wagner (Wagner, 1977). These publications and later ones instigated under California’s Landslide Hazard Mapping Act are no longer mandated as a result of changes in state law.

While there is value in the landslide-related maps and documents produced by state geological surveys, the extent to which such technical assistance is effectively introduced into landslide risk reduction efforts at the local level is less clear. Given the potential number of local governmental bodies in the United States, determining how many have implemented mandatory measures to prevent or reduce landslide damage is difficult to assess. There are certainly documented examples in parts of California, Colorado, and Utah (Fleming et al., 1979; McCalpin, 1985; and DeGraff, 2012). While these examples of mandated measures generally remain in effect now, it should be noted that none of them was instituted after 1995.

Lacking a readily definable metric, we are left to find examples that are illustrative of landslide risk reduction resulting from such technical assistance.

One such example is the California Geological Survey’s involvement with timber harvest plans to limit landslide-contributed sediment degrading water quality as part of the Timber Harvest Plan (THP) Review project. Since 1975, California Geological Survey geologists conduct reviews of timber harvesting plans prepared by a state-licensed forester in an effort to limit the detrimental effect of landslide occurrence in California’s watersheds. California Geological Survey geologists carry out an engineering geology review of planned timber sales to provide advisory comments on slope stability concerns to the California Department of Forestry and the Board of Forestry, the state regulatory agencies that grant approval of timber

236

Number of Respondents Answering “Yes”

Percent of Respondents Answering “Yes”

9 15 16

31 52 55

22

76

sale harvest plans. Under this program, the technical assistance from the California Geological Survey ensures that approved plans achieve the desired limits on landslide-generated sediment attributable to the harvest activities. The program’s effectiveness is demonstrated by its continuing existence after 40 years of operation.

A case study undertaken by DOGAMI and the coastal City of Astoria, OR, provides a more comprehensive illustration of effective landslide risk reduction resulting from interaction between a state geological survey and local government. The Landslide Risk Reduction Process highlights the partnership between state surveys and the U.S. Geological Survey Landslide Program (USGS-LP) to facilitate implementation of risk reduction measures at the local government level.

To fully understand this case study, we need a brief description of how the Landslide Risk Reduction Process was developed. In the fall of 2012, the USGS-LP and DOGAMI completed a 5 year Collaborative Landslide Hazard Project in Oregon. The collaboration focused on research and methods development to improve the quality and capacity to deliver landslide hazard and risk information to the local governments in Oregon. During the project, groundbreaking new light detection and ranging (LiDAR)–based methods for detailed landslide inventory and susceptibility mapping fundamentally changed the abilities and understanding of the landslide hazard in Oregon.

This 5 year project focused on development of standardized methods of procedure based on recent research at the USGS-LP, North Carolina Geologic Survey, California Geologic Survey, and others (Schulz, 2004; Harp et al., 2006). The goal was to have the standardized methods in place, so that future projects could focus on performing the mapping, modeling, and risk analysis at an accelerated rate. Some of the methods have been finalized and published, including the landslide inventory and the shallow landslide susceptibility methods (Burns and Madin, 2009; Burns et al., 2012). Other methods are in progress, including deep landslide susceptibility and landslide risk analysis (Burns et al., 2013).

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 225–243

Landslide Risk Reduction in the United States

Throughout the USGS-LP/DOGAMI collaborative 5 year project and the years following, DOGAMI partnered with several local governments (Oregon state agencies, cities, and counties) to work on landslide risk reduction projects. These community projects helped to establish the Landslide Risk Reduction Process (Figure 4). DOGAMI discovered that completing these projects with a willing local partner is a very important piece of the process, because a large portion of the landslide risk reduction takes place at the community level.

It is the partnership between state and local governments that results in the best available landslide hazard data being used to perform landslide risk reduction.

Following the Landslide Risk Reduction Process and methods, DOGAMI continues to work with local partners on projects.

The specific example mentioned previously was a project carried out by the City of Astoria and DOGAMI (Burns and Mickelson, 2013). In 2007, the city was experiencing several active landslides and contacted DOGAMI for assistance. DOGAMI and the City of Astoria government applied and received funding from the Federal Emergency Management Agency (FEMA) Hazard Mitigation Grant Program to conduct a landslide risk assessment and risk reduction project. The first step was to create a landslide inventory using LiDAR-generated base maps. In total, 120 landslides were found within the city boundary.

Eighty-three landslides in this inventory were estimated or known to have moved during historic time.

Seventeen of these had caused significant damage.

For instance, the Irvine Road landslide destroyed 23 homes in 1950. The second step was to create shallow and deep landslide susceptibility maps. It was found that 55 percent of the city was classified as highly susceptible to shallow landslides, and 37 percent was classified as highly susceptible to deep landslides.

Exposure risk analysis was performed, and it was found that 59 percent of the tax lots intersected with existing landslides (Burns and Mickelson, 2013).

FEMA’s HAZUS software was used to model a variety of earthquake scenarios and estimate regional damages such as building damage, lifeline damage, and injuries (FEMA, 2005). However, the HAZUS software does not have a landslide-only module, so an earthquake was modeled with and without the seismic-induced landslide hazard. This was done to estimate the range of potential damage and losses that can be expected from landslides during a major earthquake in the Astoria area. The loss ratio (ratio of total losses to the inventory of assets) was found to increase from 12 percent to 21 percent when the landslide hazard was included. These results indicate a significant increase in losses caused by the landslides triggered by an earthquake. This increased hazard effect

is likely to occur in other localities with high seismic and landslide hazard, like many portions of the Oregon coast.

The City of Astoria is now using this landslide hazard and risk information to inform their land-use planning and update their regulations related to grading and development proposals. For example, the city has purchased lands associated with historic landslides and is using them as parks and open space. The city has begun to implement the integration of the landslide data into a Geologic Hazard and Hillside Development Ordinance.

Emergency Preparedness, Response, and Recovery Whether effective landslide loss reduction efforts are in place locally or not, unexpected landslide events may still occur. Consequently, there is a need for state geological survey involvement in subsequent emergency response actions. The AASG questionnaire included three questions pertinent to emergency situations. Among the 29 questionnaire respondents, only Maryland and Vermont indicated an actual involvement with installation or construction of mitigation or stabilization measures when an unexpected landslide takes place (Table 4). This is not surprising given that the landslides are most likely to be occurring on private or non-state public land. It is unlikely the state geological survey would have the authority to undertake mitigation or stabilization actions or to incur the liability of taking on such a responsibility. In situations where the interest of the state may be paramount, there are likely to be other entities such as departments of transportation, parks and recreation, and forestry or local county departments with the authority and engineering capability for carrying out mitigation or stabilization.

Landslide warning systems are a means for ensuring public safety by notifying people of an impending landslide threat. Often such warnings are part of a comprehensive emergency management plan involving evacuation, temporary shelters, and other support measures. This can be particularly effective when other mitigation or stabilization measures are not practical or have yet to be implemented. There are only a few examples in the United States of such systems being used in real-time so that notification and other emergency response measures can be effectively initiated before the threatened movement occurs (Reid et al., 2012; NASA, 2013). As noted by Baum and Godt (2010), many of the landslide warning systems that are or have operated in the United States are for occurrence of shallow landslides or debris flows triggered by precipitation events. Because shallow landslide prediction depends on precipitation thresholds, it means the

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 225–243

237

De Graff, Burns, and McConnell

Figure 4. Diagram of the steps in the Landslide Risk Reduction Process as applied by the Oregon Department of Geology and Mineral Industries.

National Weather Service (NWS) of the National Oceanic and Atmospheric Administration (NOAA) is typically a partner in operating these systems (Baum and Godt, 2010; Cannon et al., 2011). Such warning systems use precipitation duration or intensity thresholds linked to the relative likelihood of movement or possible landslide occurrence (Baum and Godt, 2010; De Graff et al., 2015b). In 1986, the USGS successfully deployed a warning system based on precipitation thresholds in California (Keefer et al., 1987).

More recently, similar systems in California provided early warning for locations burned by wildfires in conjunction with emergency response plans developed by affected counties (USGS, 2005; Cannon et al., 2011).

Colorado, North Carolina, Oregon, and Washington (representing 14 percent of respondents) indicated participation in landslide warning systems as part of the AASG questionnaire response. Because technological and scientific advances since 1995 have made an increased use of landslide warning systems possible, it provides us with one measure indicating progress by state geological surveys in emergency response.

The third question related to emergency response in the AASG questionnaire asked whether the state geological surveys worked with their respective department of emergency management. Twenty-one State geological surveys (72 percent of the respondents) answered affirmatively to this question. Unfortunately,

there are no comparative data with which to assess whether this is greater involvement than existed in 1995.

The case study of the Rockville rock-fall response in 2013 is an example showing rapid, effective emergency response to a current landslide event (Lund et al., 2014). On its Web site, the Utah Geological Survey identifies providing geologic hazard assistance as one of the services it offers to local governments. More specific to emergency response, the mission of the agency as defined in the state code includes: “...determine and investigate areas of geologic and topographic hazards that could affect the safety of, or cause economic loss to, the citizens of the State” (Bowman, 2015). Any such work is done by assisting both the local government and the Utah Division of Emergency Management (UDEM).

Bowman (2015) highlights the following important questions that guide the Utah Geological Survey’s response actions: . “Is the site likely safe for first responders and others to enter and work?” . “What geologic information is needed to reduce the risk?” . “Is geologic monitoring needed to increase safety?” . “Is another event likely to occur within a short time frame?”

Table 4. Summary of responses to AASG questionnaire inquiries related to emergency response.

Questions A. Do you work on any hard mitigation (e.g., stabilization)?

B. Landslide warning system?

C. Do you work with your state Department of Emergency Management?

238

Number of Respondents Answering “Yes”

Percent of Respondents

2 4 21

7 14 72

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 225–243

Landslide Risk Reduction in the United States

. “Are other nearby areas at risk from geologic hazards?”

The town of Rockville is located on Utah State Route 9 about 6.5 km west of Springdale, UT, and the south entrance to Zion National Park. The town lies on the floodplain of the Virgin River between the river channel and the base of a cliff forming the southern edge of Rockville Mesa (Lund et al., 2014). The cliff exposes the interbedded sandstone and shale bedrock of the Moenkopi Formation capped by the Shinarump conglomerate member of the Chinle Formation. This stratigraphy is conducive to basal erosion of the sandstone and shale, resulting in undermining of the more resistant conglomerate rocks. At approximately 5:00 p.m. on December 12, 2013, a large block of Shinarump conglomerate detached and bounded down the cliff face (Lund et al., 2014). When the fast-moving rock-fall debris reached the base of the cliff, it destroyed a Rockville residence along with its garage and car. More importantly, it killed two people in the home (Lund et al., 2014). Two Utah Geological Survey geologists arrived at first light the following day to assist the Springdale chief of police, who was serving as the incident commander for the emer‐ gency response (Bowman, 2015). A reconnaissance of the rock-fall source area and the area leading to the destroyed residence was carried out to determine the safety of emergency responders (Bowman, 2015). In addition to collecting the geologic circumstances of this particular event, the geologists carried out a rapid assessment of expected rock-fall runout (Lund et al., 2014). This was an important part of the geologic information needed to determine whether risk could be reduced. From observations made of the capping conglomerate and downslope rock-fall deposits, it was concluded that hazard from rock fall remained high, but predicting the timing of future events was not possible (Lund et al., 2014). This conclusion was consistent with Rockville experiencing at least five other large rock falls over the past 35 years prior to the 2013 event. The destroyed residence and other nearby ones are within a zone of mapped rockfall hazard (Knudsen and Lund, 2013). Consultation between the local government authorities and the Utah Department of Transportation determined that standard engineered rock-fall mitigation measures were not feasible for protecting at-risk residences (Lund et al., 2014). The report outlined options for the residents remaining within any high-hazard area: (1) continue to reside there knowing the consequence of the decision could include significant property damage and may prove fatal or (2) relocate from high-hazard areas (Lund et al., 2014). A recommendation was

made that the Town of Rockville ensure that present and future residents and land owners in the highhazard areas be made fully aware of the hazard. It further noted that some local governments had purchased or retired the properties of homeowners who wished to relocate or to move their houses to safer locations (Lund et al., 2014). These recommendations represented a response to the question, “Are other nearby areas at risk from geologic hazards?” (Bowman, 2015).

DISCUSSION AND CONCLUSIONS It is evident that both our understanding of landslide processes and the technology that can be applied to furthering this knowledge have undergone significant improvement during the 20 years since National Landslide Awareness Day. Emphasizing the actions taken at the state level of government, we can see an improvement in actionable policy for reducing landslide hazard. Table 5 shows some of the metrics from our analysis for comparing pre- and post-1995 changes for accomplishing landslide risk reduction for the categorical elements of developing guidelines and training, increasing public awareness and education, implementing loss reduction measures, and achieving emergency preparedness, response, and recovery (Table 5).

Changes in the number of states requiring practicing geologists to be registered, the uniformity introduced into tests used to determine which individuals are qualified to be registered, and the guidelines available for conducting geological investigations are indicative of improvement in the categorical element of guidelines and training. Table 5 shows that states requiring registration or licensure for practicing geological professionals has nearly doubled since 1995 and is now instituted in 64 percent of the states. The testing used to determine which geologists are qualified for registration was up to individual states prior to 1995.

Now this testing employs the same procedures in nearly all states requiring registration (Table 5). Only three states provided guideline documents applicable to landslide hazard work prior to 1995. Now five states have such guidance documents, including recently updated versions for two of the three states with pre1995 guidance documents (Table 5). It is recognized that none of these measures is directed solely toward landslide risk reduction. If practicing geologic professionals are generally more skilled and knowledgeable, we assume those who specialize in landslide-related issues will be too. The end result is to have a greater workforce capacity at the state level to accomplish landslide risk reduction.

Public awareness and education are enhanced by the adoption by all state geological surveys of Internet-

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 225–243

239

De Graff, Burns, and McConnell Table 5. Change in measures between pre- and post-1995 for the four categorical elements.

Category Guidelines & training

Public awareness & education Implementation of loss reduction Emergency response

Metric

Pre-1995 (No. States)

Post-1995 (No. States)

Registration/licensure ASBOG testing Guidance documents Landslide hazard information on Web page Lists or guides directed at homeowners Warning systems

19 0 3 0 0 1

32 30 5 31 9 5

based outreach to the general public, residents, and landowners, and local government staff and practicing professionals. Table 5 indicates that 31 state geological surveys (62 percent) use this technology specifically to communicate landslide hazard information. Consequently, a post-1995 communications medium is facilitating dissemination of information on landslide hazards to these stakeholder groups.

Implementation of risk reduction due to landslide activity requires incorporating our current understanding of the landslide process into land-use planning and construction. Schwab et al. (2005) published a notable national-level handbook assisting planning professionals with facilitating implementation of landslide risk reduction. Certainly, efforts to ensure that residents and landowners are able to recognize early signs of landslide activity represent progress. Such recognition enables residents and landowners to determine if their property’s intended use might be impaired or to take early action to forestall damage. Table 5 shows that prior to 1995, there were no guides specifically directed at individual homeowners that would enable them to make such determinations. Currently, there are nine states where the state geological survey site offers lists of observable indicators of impending or ongoing landslide activity or an actual publication providing such information. The Oregon case study involving DOGAMI and the City of Astoria clearly demonstrates how an effective landslide risk reduction effort can be implemented at the local government level.

Many state geological surveys have an ongoing relationship with their state’s agency for emergency management. As such, they have an important role in providing geological information for emergency personnel and disaster response planning. Warning systems provide an indicative measure of progress by state geological surveys in conducting emergency response actions. Table 5 shows that one state, California, had effectively used an early warning system before 1995. During the last 20 years, five states have established early warning systems to some degree and with varying success. All of these efforts have been fostered to a greater or lesser extent by the USGS. The case study explaining the work

240

done by the Utah Geological Survey during the 2013 rock-fall emergency in Rockville, UT, is illustrative of how a state geological survey can contribute to emergency preparedness, response, and recovery actions.

Based on the metrics presented in this paper, we conclude there are signs of progress during the last 20 years toward implementing actionable policy changes and programs to achieve landslide risk reduction in the United States. Local communities can accelerate progress toward greater landslide risk reduction by implementing land-use requirements based on our improved understanding of landslide hazard. State geological surveys are key actors in furthering this trend because of their public education efforts and partnership with local communities to develop mandatory measures that assure risk reduction at the local level. Motivations for reducing landslide risk would be helped by more comprehensive efforts to document actual losses, particularly on a monetary scale. Data of this type would not only be meaningful to elected officials, but also the insurance and banking sectors.

Population growth and urban expansion mean that efforts toward the following will need to continue and expand: having an appropriate level of training for professional geologists involved with landslide risk reduction work, fostering efforts to educate the public about general and specific landslide hazards that might affect their lives and property, taking actions at the local government level to implement risk reduction measures, and developing effective emergency preparedness, response, and recovery actions.

ACKNOWLEDGMENTS The authors wish to express their appreciation for those AASG members who participated in the survey used as part of the data in this paper. The thoughtful and thorough reviews provided by James McCalpin and two anonymous reviewers are greatly appreciated, too.

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 225–243

Landslide Risk Reduction in the United States

REFERENCES ASSOCIATION OF ENGINEERING GEOLOGISTS-UTAH SECTION, 1986, Guidelines for Preparing Engineering Geologic Reports in Utah: Utah Geological and Mineral Survey Miscellaneous Publication M, 2 p.

BAUM, R. L. and GODT, J. W., 2010, Early warning of rainfallinduced shallow landslides and debris flows in the USA: Landslides, Vol. 7, No. 3, pp. 259–272.

BAUM, R. L.; GODT, J. W.; AND HIGHLAND, L. M. (Editors), 2008, Landslides and Engineering Geology of the Seattle, Washington, Area: Reviews in Engineering Geology 20, Geological Society of America, Boulder, CO, 181 p.

BERRY, K., 2005, The Green Mountain landslide: A case history of geologic hazards planning in Colorado. In Schwab, J. C.; Gori, P. L.; and Jeer, S. (Editors), Landslide Hazards and Planning: American Planning Association, Planning Advisory Service Report 533/534, pp. 131–141.

BOWMAN, S., 2015, Emergency response and the Utah Geological Survey: UGS Survey Notes, Vol. 47, No. 1, pp. 1–3.

BURBY, R. J.; DEYLE, R. E.; GODSCHALK, D. R.; AND OLSHANSKY, R. B., 2000, Creating hazard resilient communities through land-use planning: Natural Hazards Review, Vol. 1, No. 2, pp. 99–106.

BURNS, W. J. AND MADIN, I. P., 2009, Protocol for Inventory Mapping of Landslide Deposits from Light Detection and Ranging (LIDAR) Imagery: Special Paper 42, Oregon Department of Geology and Mineral Industries, Portland, OR, Electronic document, available at http://www.oregongeology.org/pubs/ sp/p-SP.htm BURNS, W. J.; MADIN, I. P.; AND MICKELSON, K. A., 2012, Protocol for Shallow-Landslide Susceptibility Mapping: Special Paper 45, Oregon Department of Geology and Mineral Industries, Portland, OR, Electronic document, available at http://www. oregongeology.org/pubs/sp/p-SP-45.htm BURNS, W. J. AND MICKELSON, K. A., 2013, Landslide Inventory, Susceptibility Maps, and Risk Analysis for the City of Astoria, Clatsop County, Oregon: Oregon Department of Geology and Mineral Industries Open-File Report O-13-05, 33 p. and 9 plates.

BURNS, W. J.; MICKELSON, K. A.; JONES, C. B.; PICKNER, S. G.; HUGHES, K. L.; AND SLEETER, R., 2013, Landslide Hazard and Risk Study of Northwestern Clackamas County, Oregon: Oregon Department of Geology and Mineral Industries Open-File Report O-13-08, 74 map plates.

CALIFORNIA GEOLOGICAL SURVEY (CGS), 2013a, Guidelines for Engineering Geologic Reports for Timber Harvesting Plans: California Geological Survey Note 45, 6 p.

CALIFORNIA GEOLOGICAL SURVEY (CGS), 2013b, Guidelines for Preparing Geological Reports for Regional-Scale Environmental and Resource Management Planning: California Geological Survey Note 52, 7 p.

CANNON, S. H., 2001, Debris-flow generation from recently burned watersheds: Environmental & Engineering Geoscience, Vol. 7, No. 4, pp. 321–341.

CANNON, S. H.; BOLDT, E. M.; LABER, J. L.; KEAN, J. W.; AND STALEY, D. M., 2011, Rainfall intensity-duration thresholds for post-fire debris-flow emergency-response planning: Natural Hazards, Vol. 59, pp. 209–236.

CANNON, S. H.; GARTNER, J. E.; RUPERT, M. G.; MICHAEL, J. A.; REA, A. H.; AND PARRETT, C., 2010, Predicting the probability and volume of postwildfire debris flows in the intermountain western United States: Geological Society of America Bulletin, Vol. 122, No. 1–2, pp. 127–144.

COMMITTEE ON THE REVIEW OF THE NATIONAL LANDSLIDE HAZARDS MITIGATION STRATEGY, 2004, Partnerships for Reducing

Landslide Risk—Assessment of the National Landslide Hazards Mitigation Strategy: The National Academies Press, Washington, DC, 131 p., Electronic document, available at http://www.nap.edu/catalog/10946/partnerships-for-reducinglandslide-risk-assessment-of-the-national-landslide DEB, S. K. AND EL-KADI, A. I., 2009, Susceptibility assessment of shallow landslides on Oahu, Hawaii, under extreme-rainfall events: Geomorphology, Vol. 108, No. 3, pp. 219–233.

DEGRAFF, J. V., 1995, National Landslide Awareness Day: AEG News, Vol. 38, No. 3, p. 11.

DEGRAFF, J. V., 2012, Solving the dilemma of transforming landslide hazard maps into effective policy and regulations: Natural Hazards and Earth System Science, Vol. 12, No. 1, pp. 53–60.

DE GRAFF, J. V.; ANDERSON, M. G.; AND HOLCOMBE, E., 2015a, Landslide risk reduction—Complementary routes to learning.

In Lollino, G.; Manconi, G.; Guzzetti, F.; Clushaw, M.; Bobrowsky, P.; and Luino, F. (Editors), Engineering Geology for Society and Territory, Volume 5: Springer International Publishing, Berlin, pp. 727–730.

DEGRAFF, J. V.; DEGRAFF, N.; AND ROMESBURG, H. C., 2013, Literature searches with Google Scholar: Knowing what you are and are not getting: GSA Today, Vol. 23, No. 10, pp. 44–45.

DE GRAFF, J. V.; GALLEGOS, A. J.; REID, M. E.; LAHUSEN, R. G.; AND DENLINGER, R. P., 2015b, Using monitoring and modeling to define the hazard posed by the reactivated Ferguson rock slide, Merced Canyon, California: Natural Hazards, Vol. 76, No. 2, pp. 769–789.

FEDERAL EMERGENCY MANAGEMENT AGENCY (FEMA), 2005, Hazus-MH Software, FEMA’s Tool for Estimating Potential Losses from Natural Disasters: Electronic document, available at http://www.fema.gov/hazus/ FLEMING, R. W.; VARNES, D. J.; AND SCHUSTER, R. L., 1979, Landslide hazards and their reduction: Journal American Planning Association, Vol. 45, No. 4, pp. 428–439.

GARTNER, J. E.; CANNON, S. H.; SANTI, P. M.; AND DEWOLFE, V. G., 2008, Empirical models to predict the volumes of debris flows generated by recently burned basins in the western US: Geomorphology, Vol. 96, No. 3, pp. 339–354.

GODT, J. W., 1997, Digital Compilation of Landslide Overview Map of the Conterminous United States: U.S. Geological Survey Open-File Report 97-289, Electronic document, available at http://landslides.usgs.gov/hazards/nationalmap/ GUZZETTI, F.; PERUCCACCI, S.; ROSSI, M.; AND STARK, C. P., 2008, The rainfall intensity-duration control of shallow landslides and debris flows: An update: Landslides, Vol. 5, No. 1, pp. 3–17.

GUZZETTI, F.; REICHENBACH, P.; AND WIECZOREK, G. F., 2003, Rockfall hazard and risk assessment in the Yosemite Valley, California, USA: Natural Hazards and Earth System Science, Vol. 3, No. 6, pp. 491–503.

HANEBERG, W. C., 1995, Landslide potential in New Mexico: Lite Geology, Fall, pp. 1–2.

HANSEN, M. C. (Compiler), 1995, Landslides in Ohio: GeoFacts No. 8, Ohio Department of Natural Resources, Division of the Geological Survey, Columbus, OH. HARP, E. L.; MICHAEL, J. A.; AND LAPRADE, W. T., 2006, Shallow-Landslide Hazard Map of Seattle, Washington: U.S. Geological Survey Open-File Report 2006-1136, 20 p.

HARRIS, R. C. AND PEARTHREE, P. A., 2002, A Homebuyer’s Guide to Geologic Hazards in Arizona. Down-to-Earth, Volume 13: Arizona Geological Survey, Tucson, AZ, 36 p.

HART, D. A., 2005, Policy tools for addressing landslide hazards in Pittsburgh, Pennsylvania. In Schwab, J. C.; Gori, P. L.; and

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 225–243

241

De Graff, Burns, and McConnell Jeer, S. (Editors), Landslide Hazards and Planning: American Planning Association, Planning Advisory Service Report 533/534, pp. 125–130.

HIGHLAND, L. M., 2012, Landslides in Colorado, USA—Impacts and Loss Estimation for 2010: U.S. Geological Survey Open-File Report 2012–1204,49 p.

HILL, R. L., 1995, Landslide: The Oregonian (Portland), October 19, 1995, p. B1, B7.

IVERSON, R. M. AND MAJOR, J. J., 1987, Rainfall, ground-water flow, and seasonal movement at Minor Creek landslide, northwestern California: Physical interpretation of empirical relations: Geological Society America Bulletin, Vol. 99, No. 4, pp. 579–594.

JABOYEDOFF, M.; OPPIKOFER, T.; ABELLÁN, A.; DERRON, M. H.; LOYE, A.; METZGER, R.; AND PEDRAZZINI, A., 2012, Use of LIDAR in landslide investigations: A review: Natural Hazards, Vol. 61, No. 1, pp. 5–28.

JIBSON, R. W.; HARP, E. L.; SCHULZ, W.; AND KEEFER, D. K., 2004, Landslides triggered by the 2002 Denali Fault, Alaska, earthquake and the inferred nature of the strong shaking: Earthquake Spectra, Vol. 20, No. 3, pp. 669–691.

JOMARD, H.; LEBOURG, T.; GUGLIELMI, Y.; AND TRIC, E., 2010, Electrical imaging of sliding geometry and fluids associated with a deep seated landslide (La Clapière, France): Earth Surface Processes and Landforms, Vol. 35, No. 5, pp. 588–599.

KEEFER, D. K.; WILSON, R. C.; MARK, R. K.; BRABB, E. E.; BROWN, W. M.; ELLEN, S. D.; HARP, E. L.; WIECZOREK, G. F.; ALGER, C. S.; AND ZATKIN, R. S., 1987, Real-time landslide warning during heavy rainfall: Science, Vol. 238, No. 4829, pp. 921–925.

KNUDSEN, T. R. AND LUND, W. R., 2013, Geologic Hazards of the State Route 9 Corridor, LaVerkin City to Town of Springdale, Washington County, Utah: Utah Geological Survey Special Study 148, 13 p.

LEINER, B. M.; CERF, V. G.; CLARK, D. D.; KAHN, R. E.; KLEINROCK, L.; LYNCH, D. C.; POSTEL, J.; ROBERTS, L. G.; AND WOLFF, S. S., 1997, The past and future history of the Internet: Communications Association Computing Machinery, Vol. 40, No. 2, pp. 102–108.

LUND, W. R.; KNUDSEN, T. R.; AND BOWMAN, S. D., 2014, Investigation of the December 12, 2013, Fatal Rock Fall at 368 West Main Street, Rockville, Utah: Utah Geological Survey Report of Investigation 273, 20 p.

LUND, W. R.; KNUDSEN, T. R.; VICE, G. S.; AND SHAW, L. M., 2008, Geologic Hazards and Adverse Construction Conditions, St. George–Hurricane Metropolitan Area, Washington County, Utah: Utah Geological Survey Special Study 127, 105 p.

MADER, G., 2005, A model of effective use of geology in planning, Portola Valley, California. In Schwab, J. C.; Gori, P. L.; and Jeer, S. (Editors), Landslide Hazards and Planning: American Planning Association, Planning Advisory Service Report 533/534, pp. 96–112.

MAJOR, J. J. AND IVERSON, R. M., 1999, Debris-flow deposition: Effects of pore-fluid pressure and friction concentrated at flow margins: Geological Society of America Bulletin, Vol. 111, No. 10, pp. 1424–1434.

MCCALPIN, J. P., 1985, Engineering geology at the local government level: planning, review, and enforcement: Bull Assn Eng Geologists, Vol. 22, No. 3, pp. 315–327.

MCKENNA, J. P.; SANTI, P. M.; AMBLARD, X.; AND NEGRI, J., 2012, Effects of soil-engineering properties on the failure mode of shallow landslides: Landslides, Vol. 9, No. 2, pp. 215–228.

NATIONAL AERONAUTICS AND SPACE ADMINISTRATION (NASA), 2013, Sizing up the Landslide at Bingham Canyon Mine: National Aeronautics and Space Administration (NASA) Earth Observatory, Electronic document, available at http://earthobservatory.nasa.gov/IOTD/view.php?id581364

242

NATIONAL ASSOCIATION OF STATE BOARDS OF GEOLOGY (ASBOG), 2015, National Association of State Boards of Geology: Electronic document, available at http://www.asbog.org/ NEW HAMPSHIRE DEPARTMENT OF ENVIRONMENTAL SERVICES (NHDES), 2015, Flood and Geologic Hazards: Overview: Electronic document available at http://des.nh.gov/organization/ commissioner/gsu/fegh/categories/overview.htm NEW YORK STATE EDUCATION DEPARTMENT (NYSED), 2015, Geology: Electronic document available at http://www.op. nysed.gov/prof/geo/ OREGON STATE BOARD OF GEOLOGIST EXAMINERS (OSBGE), 2014, Guide for Preparing Engineering Geologic Reports: Oregon State Board of Geologist Examiners, Salem, OR, 14 p. Electronic document, available at http://www.oregon.gov/osbge/pdfs/ Publications/EngineeringGeologicReports_5.2014.pdf PETLEY, D. N.; BULMER, M. H.; AND MURPHY, W., 2002, Patterns of movement in rotational and translational landslides: Geology, Vol. 30, No. 8, pp. 719–722.

PREUSS, J.; BUSS, K.; AND LOPEZ, L., 2005, The Kelso, Washington, Aldercrest-Banyon landslide. In Schwab, J. C.; Gori, P. L.; and Jeer, S. (Editors), Landslide Hazards and Planning: American Planning Association, Planning Advisory Service Report 533/534, pp. 113–124.

PRIEST, G. R.; SCHULZ, W. H.; ELLIS, W. L.; ALLAN, J. A.; NIEM, A. R.; AND NIEM, W. A., 2011, Landslide stability: Role of rainfall-induced, laterally propagating, pore-pressure waves: Environmental & Engineering Geoscience, Vol. 17, No. 4, pp. 315–335.

POTTER, P. E.; BOWER, M.; MAYNARD, J. B.; CRAWFORD, M. W.; WEISENFLUH, G. A.; AND AGNELLO, T., 2013, Landslides and Your Property: Indiana Geological Survey Miscellaneous Item 111.

REID, M. E.; LAHUSEN, R. G.; BAUM, R. L.; KEAN, J. W.; SCHULZ, W. H.; AND HIGHLAND, L. M., 2012, Real-Time Monitoring of Landslides: U.S. Geological Survey Fact Sheet 2012-3008.

ROERING, J. J.; STIMELY, L. L.; MACKEY, B. H.; AND SCHMIDT, D. A., 2009, Using DInSAR, airborne LiDAR, and archival air photos to quantify landsliding and sediment transport: Geophysical Research Letters, Vol. 36, No. 19, p. L19402.

SANTI, P. M.; HIGGINS, J. D.; CANNON, S. H.; AND GARTNER, J. E., 2008, Sources of debris flow material in burned areas: Geomorphology, Vol. 96, No. 3, pp. 310–321.

SCHULZ, W. H., 2004, Landslides Mapped Using LiDAR Imagery, Seattle, Washington: U.S. Geological Survey Open-File Report 2004-1396, 11 p.

SCHULZ, W. H., 2007, Landslide susceptibility revealed by LiDAR imagery and historical records, Seattle, Washington: Engineering Geology, Vol. 89, No. 1, pp. 67–87.

SCHWAB, J.; GORI, P., AND JEER, S. (Editors), 2005, Landslide Hazards and Planning: American Planning Association, Planning Advisory Service Report 533/534.

SHELTON, D. C. AND PROUTY, D., 1979, Nature’s Building Codes, Geology and Construction in Colorado: Special Publication 12, Colorado Geological Survey, Golden, CO, Appendix E, pp. 61–67.

SIEGEL, L., 1995, Growth will aggravate Utah’s landslide hazard: Salt Lake Tribune, October 12, 1995, pp. C1–C2.

SLOSSON, J. E., 1984, Genesis and evolution of guidelines for geologic reports: Bulletin Association of Engineering Geologists, Vol. 21, No. 3, pp. 295–316.

SPIKER, E. C. AND GORI, P. L., 2003, National Landslide Hazards Mitigation Strategy—A Framework for Loss Reduction: U.S. Geological Survey Circular 1244, 56 p.

STOCK, G. M.; BAWDEN, G. W.; GREEN, J. K.; HANSON, E.; DOWNING, G.; COLLINS, B. D.; BOND, S.; AND LESLAR, M., 2011, High-resolution three-dimensional imaging and analysis of rock falls in Yosemite Valley, California: Geosphere, Vol. 7, No. 2, pp.

573–581.

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 225–243

Landslide Risk Reduction in the United States STOCK, G. M.; MARTEL, S. J.; COLLINS, B. D.; AND HARP, E. L., 2012, Progressive failure of sheeted rock slopes: The 2009–2010 Rhombus Wall rock falls in Yosemite Valley, California, USA: Earth Surface Processes and Landforms, Vol. 37, No. 5, pp. 546–561.

TARCHI, D.; CASAGLI, N.; FANTI, R.; LEVA, D. D.; LUZI, G.; PASUTO, A.; AND SILVANO, S., 2003, Landslide monitoring by using ground-based SAR interferometry: An example of application to the Tessina landslide in Italy: Engineering Geology, Vol. 68, No. 1, pp. 15–30.

U.S. GEOLOGICAL SURVEY (USGS), 2005, A NOAA-USGS Demonstration Flash-Flood and Debris-Flow Early-Warning System: U.S. Geological Survey Fact Sheet 2005-3104.

WAGNER, D. L., 1977, Geology for Planning in Western Marin County, California: California Division of Mines and Geology Open-File Report OFR 77-15, 42 p.

WASHINGTON DIVISION OF GEOLOGY AND EARTH RESOURCES (WDGER), 2015, Landslides Hazards in Washington State: Washington Division of Geology and Earth Resources Fact Sheet 1.

WIECZOREK, G. F. AND LEAHY, P. P., 2008, Landslide hazard mitigation in North America: Environmental & Engineering Geoscience, Vol. 14, No. 2, pp. 133–144.

WISHARD, L., 1998, Precision among Internet search engines: An earth sciences case study: Issues in Science and Technology Librarianship, 18. doi:10.5062/F4R78C6M

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 225–243

243

Characterization of Failure Parameters and Preliminary Slope Stability Analysis of the Cedar Canyon Landslide, Iron County, Utah ASHLEY TIZZANO Menard U.S.A., Inc., 9930 Johnnycake Ridge Road, Suite 2H, Mentor, OH 44060 atizzano@menardusa.com

ABDUL SHAKOOR1 Department of Geology, Kent State University, Kent, OH 44242 ashakoor@kent.edu

WILLIAM R. LUND Utah Geological Survey, Emeritus, 436 North Main Street, Cedar City, UT 84721 billlund@utah.gov

Key Terms: Cedar Canyon Landslide, Rockfall, Straight Cliffs Formation, Dakota Formation, Tropic Shale, Stability Analysis, Remediation Measures

extra weight from previous rockfalls, and buildup of pore pressure due to a rainstorm prior to failure contributed to the landslide.

ABSTRACT The 2011 Cedar Canyon landslide, 8 mi (12.8 km) east of Cedar City, UT, severely damaged State Route 14.

Bedrock stratigraphy at the site consists of the cliffforming Straight Cliffs Formation and the underlying slope-forming Tropic Shale and Dakota Formation.

The lower part of the Straight Cliffs Formation and all of Tropic Shale and Dakota Formation are covered with a thick deposit of colluvial soil that has accumulated at the base of the steep cliff formed by the Straight Cliffs Formation. The landslide initiated in the colluvium and propagated as a translational slide along the contact between the colluvial soil and underlying bedrock units. The Utah Department of Transportation drilled and installed slope inclinometers in three borings and conducted a geophysical survey from the crest to the toe of the landslide. We collected orientation data for 186 discontinuities from the Straight Cliffs and Dakota Formations to determine the principal joint sets for use in kinematic analysis; tested samples of colluvial soil and bedrock to determine natural water content, density, and shear strength parameters of soil, bedrock, and the soil-bedrock contact; used SLIDE software to perform a preliminary stability analysis; and performed a sensitivity analysis with respect to varying positions of the water table. Results show that a combination of relatively steep colluvial slope (~306), 1

Corresponding author. email: ashakoor@kent.edu.

INTRODUCTION During the early morning of October 8, 2011, a massive landslide in Cedar Canyon, 8 mi (12.8 km) east of Cedar City, Iron County, UT (Figure 1), caused extensive damage to State Route 14 (SR 14), a secondary transportation route that connects Interstate 15 to the west and U.S. Highway 89 to the east (Lund et al., 2009). The landslide detached approximately 1.5 million yd3 (1.1 million m3) of material from the south side of the canyon (Figure 2), displaced part of SR 14 (Figure 3), and covered a 1200 ft (365 m) stretch of the road with slide debris (Figure 2). The landslide debris extended to the bottom of the canyon and impinged on Coal Creek, a perennial stream draining Cedar Canyon (Figure 2). Cedar Canyon is narrow at the landslide site, and Coal Creek has the potential to undercut both the canyon slope and the new landslide material. The landslide was approximately 1000 ft (303 m) long (from crown to toe) and a maximum of 1700 ft (515 m) wide (from side to side) (Lund et al., 2011). The main scarp formed in slope colluvium adjacent to the Straight Cliffs Formation (limey sandstone), with the largest scarp exposure measuring approximately 50 ft (15 m) high (Figure 4). Field observations showed that the landslide material consisted chiefly of colluvial soil with sandstone boulders derived from Straight Cliffs Formation rockfalls.

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 245–258

245

Tizzano, Shakoor, and Lund

Figure 1. Location of the Cedar Canyon landslide (http://www.merriam-webster.com/cgi-bin/nytmaps.pl?utah, Google Earth).

Based on landslide damage and costs incurred between 1979 and 1989, Utah is one of eight states with a landslide hazard rating of “severe,” the highest hazard class (Brabb, 1989). Conditions favorable to landslide activity exist primarily in mountain ranges and along the edges of high plateaus (Harty, 1991).

The recent history of landslides, rockfalls, and debris flows in Cedar Canyon, in the vicinity of the Cedar

City landslide, can be seen in Figure 4 (Lund et al., 2011). The earliest recorded landslide in Cedar Canyon occurred on November 24, 1915; the Parowan Times from that date mentions a large landslide stopping travel through the canyon (Parowan Times, 1915). In

Figure 2. Aerial view of the Cedar Canyon landslide (courtesy of UDOT).

Figure 3. Displaced portions of SR 14 on the east side of the landslide (courtesy of UDOT).

246

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 245–258

Cedar Canyon Utah Landslide

Figure 4. Slope hazard history of Cedar Canyon in the vicinity of the Cedar Canyon landslide (Lund et al., 2011).

March 1989, another large landslide, adjacent to the 2011 failure to the west, with a volume of approximately 2 million yd3 (1.5 million m3), also closed SR 14. This landslide initiated in the colluvial slope material below the road, carrying the road downslope (Harty, 1989). A debris flow in 2005 littered Coal Creek with trees from several miles away (Giraud and Lund, 2005). In 2009, a large rockfall (~60,000 yd3; 45,060 m3) buried SR 14 under tens of large (hundred ton plus) boulders and hundreds of smaller boulders (Lund et al., 2009).

Three abandoned, interconnected coal mines, with a total of about 4000 ft (1212 m) of underground workings, penetrate the bedrock formations at and near the landslide (Figure 4), but the extent to which these underground workings may have contributed to the 2011 landslide and previous slope instability at this location is unknown. Two of the mine entrances have been sheared off by landsliding, and the third (farthest west) discharges a substantial amount of water (i.e., dewaters the cliffs.)

Erskine et al., 2001; Hintze and Kowallis, 2009; Lund et al., 2009; Biek et al., 2012). The Tibbet Canyon Member is 600 ft (183 m) thick, consisting dominantly of fine- to coarse-grained, limey sandstone and minor amounts of interbedded gray mudstones and siltstones (Hintze and Kowallis, 2009). The Tropic Shale consists of approximately 30 ft (9.1 m) of sandy mudstone and muddy sandstone (Hintze and Kowallis, 2009; Lund et al., 2011). Topographically, the Tropic Shale forms slopes of moderate inclination on which disintegrated bedrock colluvium and rockfalls have accumulated so thickly that exposures suitable for detailed study are rare (Gregory, 1950; Biek et al., 2012). The Dakota Formation is over 600 ft (183 m) thick and consists chiefly of mudstone and sandy mudstone interbedded with thin sandstone and coal horizons up to several feet (~1 m) thick (Hintze and Kowallis, 2009; Lund et al., 2011). It is both a slopeand ledge-forming unit (Biek et al., 2012). The Tropic Shale and Dakota Formation are undivided in Cedar Canyon, where the Tropic Shale is a few feet to, at most, 30 ft (1–9.1 m) thick (Biek et al., 2012).

Stratigraphic Setting of the Area Cedar Canyon is in eastern Iron County, UT. The landslide is on the south wall of the east-west–trending canyon. The stratigraphy of Cedar Canyon, exposed in the steep canyon walls where the landslide occurred, consists of Cretaceous bedrock units that include the cliff-forming Tibbet Canyon Member of the Straight Cliffs Formation, and the underlying, slope-forming Tropic Shale and Dakota Formation (Gregory, 1950;

Study Objectives The main objectives of this study of the Cedar Canyon landslide were to: (1) determine the type of landslide, the location of the failure plane, and causes of the landslide; (2) determine the engineering properties of the colluvial soil and bedrock units involved in sliding; and (3) perform a stability analysis including kinematic analysis of discontinuities in the main scarp area.

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 245–258

247

Tizzano, Shakoor, and Lund

and Mah (2004), to map 186 discontinuities within the bedrock units at or near the landslide site (Figure 5). We used the window mapping method because the detailed line survey method alone can miss horizontal discontinuities parallel to the survey line (Park and West, 2002). The discontinuities were chiefly measured in the close vicinity of the Cedar Canyon landslide because of hazardous conditions at the landslide site itself. In addition to collecting orientation and spacing data for the bedrock discontinuities, we recorded other aspects of the discontinuities (continuity or length, surface irregularities, nature of infilling material, etc.). We used the discontinuity data to evaluate the potential modes of failure affecting the bedrock at the landslide site. Debris from these failures (chiefly rockfalls) accumulates on the colluvial slope at the base of the bedrock cliff, adding to the landslide driving force. We also collected samples of colluvial material and of bedrock for laboratory testing.

Subsurface Investigations The Utah Department of Transportation (UDOT) conducted detailed subsurface investigations, through contractors, to collect the data required for selecting appropriate repair measures for restoring traffic on SR 14. UDOT drilled three borings, two within the landslide at mile marker 8 and one outside the landslide area near mile 7.5. The borings were designated 8-1, 8-2, and 7.5-3 because of their proximity to mile markers 8 and 7.5 along SR 14 (Figure 6).

Boring depths ranged from 75 ft (23 m) at the landslide toe to 171 ft (52 m) near the head of the landslide. UDOT placed slope inclinometers in each borehole to monitor any continued landslide movement. In addition to borings, UDOT conducted a geophysical survey, using a surface seismic refrac� tion micro-tremor (ReMi) and a SmartSeis SE-24 seismograph. Contractors ran eight geophysical lines, 240 ft (73 m) long, across the width of the slide, approximately perpendicular to the sliding direction to determine the depth to bedrock and possible location of the failure plane.

Figure 5. Discontinuities within sandstone units: (a) Straight Cliffs Formation, and (b) Dakota Formation.

Laboratory Investigations RESEARCH METHODS Field Investigations We used a combination of the detailed line sur� vey method, developed by Piteau and Martin (1977), and the window mapping method, described in Wyllie

248

We tested core samples, obtained from UDOT, as well as colluvial soil and bedrock samples collected as part of our field investigation, in the labora tory to determine natural water content, grain size distribution, Atterberg limits, specific gravity, absorption, slake durability index, unconfined compressive strength, and shear strength parameters, as

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 245–258

Cedar Canyon Utah Landslide

Figure 6. Locations of UDOT boreholes (Google Earth, USGS, 2002).

appropriate. Where applicable, we conducted all tests using the standard procedures specified by the American Society for Testing and Materials (ASTM, 2011).

Since core samples were not available for the Straight Cliffs Formation, we used the Schmidt hammer test to obtain unconfined compressive strength (Deere and Miller, 1966). In most cases, we performed three tests to determine each property. We used grain size distribution and Atterberg limits results to classify the colluvial soil according to the Unified Soil Classification System (USCS) (ASTM method 2487; ASTM, 2011).

Absorption, slake durability index, and unconfined compressive strength were used for general characterization of the bedrock material. The amount of water absorbed by a rock is an indirect indicator of its void space, strength, and resistance to weathering (Shakoor and Bonelli, 1991; Shakoor and Barefield, 2009). The slake durability index test, conducted on the weaker, argillaceous bedrock units, provided a quantitative measure of resistance to weathering. We used density, determined from blow count values in the case of soil and from specific gravity in the case of rock, and shear strength parameters for stability analysis. The shear strength parameters for soil, bedrock, and the soil-bedrock contact were determined using the direct shear test.

FIELD AND LABORATORY DATA Principal Discontinuities Analysis of discontinuity orientation data from sandstone units within the Dakota and Straight Cliffs Formations, using DIPS software (Rocscience, 2013), showed the presence of three principal joint sets in addition to bedding, with the following trends: joint set 1—near vertical (88u) with an average dip direction of 149u; joint set 2 (a valley stress relief joint)—an average dip angle of 84u and an average dip direction of 30u; joint set 3—an average dip angle of 89u and an average dip direction of 276u; bedding—an average dip angle of 4u and an average dip direction of 156u.

Spacing between discontinuities ranged from 1 in.

(2.5 cm) to 8 ft (2.4 m). The length of discontinuities ranged from a few inches (5 cm) to tens of feet (.10 m). The infilling material in the open joints consisted of material eroded from the surrounding rock and was very weak (easily brushed off). Field observations showed evidence of water runoff on the slope face of the Dakota Formation, but during data collection, the discontinuities were dry. The joints in Dakota sandstone units were relatively smooth and straight, whereas those in Straight Cliffs sandstone horizons were very rough and undulating.

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 245–258

249

Tizzano, Shakoor, and Lund Table 1. Summary of dry density, absorption, slake durability index, and unconfined compressive strength data for the bedrock units involved in the Cedar Canyon landslide. These data are based on three samples tested for each property, except for the Dakota Formation density values, which were measured on eight samples because more samples were available.

Densitya (pcf)

Straight Cliffs Sandstone Tropic Shale Dakota Sandstone

Absorption (%)

Max

Min

Mean

St. Dev.

Max

Min

Mean

St. Dev.

132.3 165.6 165.4

130.4 162.2 139.8

131.5 163.4 156.2

1.0 1.9 10.7

8.9 n/a 7.5

7.7 n/a 6.7

8.4 n/a 7.2

0.6 n/a 0.4

pcf 5 1 Mg/m3.

145 psi 5 1 MPa.

a 62.4 b

Subsurface Data Borehole 8-1, near the main scarp of the landslide, was drilled to 171 ft (51.8 m). The landslide deposit at that location contained boulders of sandstone in a matrix of sand, silt, and clay, which UDOT classified as silty sand (SM) to low-plasticity clay (CL). The average blow count for the colluvial soil was 12, which we used to approximate in situ density for laboratory tests. The borehole log revealed the presence of very moist soil at 95 ft (28.8 m), soft to very soft, dark gray shale to siltstone at 147 ft (44.8 m), and light gray, silty, fine-grained sandstone with coal seams at 163 ft (49.4 m). The shale, siltstone, and sandstone represent the Tropic Shale and Dakota Formation undivided (Ktd). No shear zones were reported on borehole logs, either within the colluvial soil or the bedrock. The borehole was dry. UDOT placed a 2.75-in.-diameter (7-cm-diameter) Geokon inclinometer casing in the borehole to a depth of 170 ft (51.5 m), and a sand slurry backfill was poured from 171 ft (51.8 m) to 38 ft (11.5 m) and topped with a cement-bentonite grout backfill.

Borehole 8-2, near the toe of the landslide, was drilled to 75 ft (22.7 m). The borehole revealed that within the top 20 ft (6.1 m), the landslide consisted of boulders of sandstone in a matrix of sand, silt, and clay. Below 20 ft (6.1 m), the landslide debris consisted of moist, dark brown, silty to clayey gravel and cobbles, classified by UDOT as clayey gravel to silty gravel (GC to GM). At 30 ft (9.1 m), rounded to sub-rounded gravel and cobbles in a medium dense matrix of sand and silt were interpreted to be stream alluvium. No shear zones within the colluvial soil or stream alluvium were reported on the borehole log.

The water table was at 37 ft (11.2 m) below the surface.

UDOT placed a 2.75-in.-diameter (7-cm-diameter) Geokon inclinometer casing in the borehole to 74.5 ft (22.6 m) and poured a sand slurry backfill from 74.5 ft (22.6 m) to 35 ft (10.6 m). The remainder of the boring was topped with a cement-bentonite grout backfill.

Borehole 7.5-3 was drilled 0.5 mi (0.8 km) west of the Cedar Canyon landslide and extended to 100.2 ft

250

(30.4 m). The top 62 ft (18.8 m) consisted of gravel, cobbles, and boulders in a loose matrix of gray sand and silt. Bedrock below this depth was a light gray to light brown, poorly cemented, severely weathered sandstone with closely spaced discontinuities, and this was interpreted to be the Tropic Shale and Dakota Formation undivided (Ktd). The borehole was dry.

UDOT placed a 2.75-in.-diameter (7-cm-diameter) Geokon inclinometer casing in the borehole to 100 ft (33 m) and sealed the remainder of the borehole in a manner similar to the other two borings.

The geophysical survey showed that depth to bedrock ranged from 71 ft (21.6 m) near the toe of the slide to 121 ft (36.9 m) near the head of the landslide.

A comparison of borehole logs and geophysical survey data showed that the bedrock in the main scarp area was encountered at greater depth in the borehole than indicated by the geophysical survey.

The inclinometer data showed no evidence of movement during the period of monitoring (December 28, 2011, to March 21, 2012). Therefore, the inclinometer data could not be used to determine the location of the failure plane.

Engineering Characterization of Colluvial Soil and Bedrock Units Based on grain size distribution and using the USCS, we classified the colluvial soil that we collected from the landslide site as poorly graded sand (SP). The Atterberg limits indicated that the material passing the #200 sieve (0.074 mm) was low-plasticity clay (CL) according to the USCS. Based on the combined results of grain size distribution and Atterberg limits, the soil is primarily classified as poorly graded sand to clayey sand (SP-SC). Boulders, cobbles, and gravel, in varying proportions, were present throughout the soil.

The natural water content of the soil ranged from 7.6 percent to 11.6 percent, with an average of 9.3 percent.

The average bulk density of the soil, as determined from the blow count value of 12, was 110 pcf (1.8 Mg/m3).

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 245–258

Cedar Canyon Utah Landslide Table 1. Extended.

Unconfined Compressive Strengthb (psi)

Slake Durability Index (Id2) Max

Min

Mean

St. Dev.

Max

Min

Mean

St. Dev.

n/a 89.8 27.0

n/a 15.6 28.0

n/a 51.5 27.5

n/a 37.1 0.5

4,350.0 n/a 5,480.4

3,045.0 n/a 3,432.6

3,465.5 n/a 4,397.4

513.6 n/a 889.2

Table 1 lists the dry density, absorption, slake durability index, and unconfined compressive strength data for the bedrock units we sampled. The variations in the data for the three samples tested for each property are due to localized lithological variations. We determined the slake durability index for both the Tropic Shale and the Dakota Formation because these two formations comprise the bedrock underlying the colluvium in which the landslide formed (Biek et al., 2012). The mean slake durability index values for the two units are 51.5 percent and 27.5 percent, respectively. According to the International Society for Rock Mechanics (ISRM) classification (ISRM, 1979), the Tropic Shale has low to medium durability, and the Dakota Formation has low durability. The low slake durability index values for the Tropic Shale and Dakota Formation explain the thick accumulation of colluvial soil formed on the two formations.

Frequent rockfalls from the overlying Straight Cliffs Formation contribute to the variable size of the boulders in the colluvium.

The mean unconfined compressive strength values for the Straight Cliffs and Dakota Formations are 3,465.5 psi (24 MPa) and 4,397.4 psi (30.3 MPa), respectively. According to the Deere and Miller (1966) classification, the Straight Cliffs Formation belongs to class E (very low strength), and the Dakota Formation belongs to class D (low strength). The strength values determined for the Straight Cliffs Formation, a cliff-forming unit, are lower than expected, which may be due to the use of the Schmidt hammer testing method (an empirical method). However, these results are corroborated by the relatively low density and high absorption values of the sandstones, to some extent, as both density and absorption are linearly related to unconfined compressive strength (Shakoor and Bonelli, 1991). Again, the low strength values of the Straight Cliffs and Dakota Formations, the former exposed in the main scarp of the landslide and the latter comprising the bedrock underlying the colluvial soil mass, combined with the low slake durability index values for the Dakota Formation and the Tropic Shale, explain the unusually thick accumulation of colluvial soil at the cliff base.

Table 2 presents the shear strength parameters for the soil and soil-bedrock contact, as determined from direct shear tests. The results show that dry soil has a higher resistance to shearing through the soil and along the soil-bedrock contact than wet soil.

STABILITY ANALYSIS Failure Plane Location and Landslide Type Used for Stability Analysis During subsurface investigations, UDOT did not report the presence of shear zones within the soil or bedrock at the landslide site. However, in borehole 81, the contact between the soil and bedrock was very soft to soft shale and siltstone of the undivided Tropic Shale and Dakota Formation, and possibly represents the shear zone. Therefore, it can be inferred that the failure occurred along the soil-bedrock contact, or within weathered bedrock slightly below the contact.

The Cedar Canyon landslide combined rotational and translational movement, with the majority of the movement being translational along the contact between the colluvial soil and the underlying Tropic Shale and Dakota Formation. The landslide started as a rotational failure in colluvium that exposed the Straight Cliffs Formation in the main scarp and transitioned into a translational failure as it moved downslope along the soil-bedrock contact. The observation that the failure occurred along the soil-bedrock contact is further supported by the lower shear strength along the contact than within either the soil or bedrock (Table 2).

Input Parameters for Stability Analysis Table 3 shows input parameters for three potential failure surfaces, i.e., soil–Tropic Shale, soil–Dakota Formation, and soil–Tropic Shale/Dakota Formation combined. The SLIDE software (Rocscience, 2013) used for this stability analysis requires that bedrock parameters be input in accordance with the generalized Hoek and Brown failure criterion (Hoek et al., 2002), which uses mean unit weight (density), mean

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 245–258

251

Tizzano, Shakoor, and Lund Table 2. Summary of the shear strength parameters for colluvial soil and soil-bedrock contact.

Peak Friction Angle (u)

Dry soil Soil at natural water content Dry soil–Dakota Formation contact Soil at natural water content–Dakota Formation contact Dry soil–Tropic Shale contact

Residual Friction Angle (u)

Max

Min

Mean

St. Dev.

Max

Min

Mean

St. Dev.

46 39 47 37 33

36 36 36 33 25

40 37 42 35 30

5.5 1.5 5.5 1.8 4.4

44 n/a 36 35 25

34 n/a 33 33 22

38 n/a 34 34 24

5.3 n/a 1.8 1 2.1

145 psi 5 1 MPa. The residual cohesion value should not exceed the peak cohesion value. In this case, the residual value being higher than the peak value can be attributed to experimental error.

a

b

with varying positions of the water table (sensitivity analysis). The SLIDE software uses two methods of analysis: the Bishop simplified method and the Janbu simplified method. Both methods yielded similar results for the three groundwater conditions analyzed in this study.

For completely dry condition, the factor of safety was 0.8 for failure along the colluvial soil–Tropic Shale contact, 1.2 for the colluvial soil–Dakota Formation contact, and 1.02 for the average soil-bedrock parameters. Since a factor of safety of 0.8 is not possible prior to failure unless the slope is already moving, we believe that either the average strength parameters along the soil-bedrock contact influenced the stability, indicating a marginally stable slope, or our strength parameters along the colluvial soil–Tropic Shale contact are incorrect. The factor of safety for the completely saturated condition was 0.3 for the colluvial soil– Tropic Shale contact, and 0.4 for both the colluvial soil–Dakota Formation and the average soil-bedrock parameters. To investigate the effect on factor of safety of varying positions of the water table, we conducted a sensitivity analysis. As the first step in the analysis, we used varying soil density values between a dry density of 102.5 pcf (1.6 Mg/m3) and a saturated density of 126.7 pcf (2.0 Mg/m3) to investigate whether density variations associated with rainfall infiltration alone would reduce the factor of safety to #1.0. The results showed that density variation had little effect on factor of safety. Therefore, as the next step in the analysis, we placed the water table above the soil-bedrock contact, creating partially saturated soil

unconfined compressive strength, and an assigned value of the geologic strength index (GSI). The bedrock parameters were 159.1 pcf (2.6 Mg/m3) and 156.2 pcf (2.5 Mg/m3) for unit weight, 3,625 psi (25 MPa) and 2,231.5 psi (16 MPa) for unconfined compressive strength, and 53 and 20 for GSI values for the Straight Cliffs Formation and the undivided Tropic Shale and Dakota Formation, respectively. We used the chart in Marinos and Hoek (2000) to select GSI values for the bedrock units.

Stability Analysis Using SLIDE Software Figure 7 shows the cross section of the Cedar Canyon landslide used for stability analysis. It was created using a 2002 topographic map (USGS, 2002), a geologic map of the area, field observations, UDOT borehole logs, and results of the UDOT geophysical survey. The contact between the Straight Cliffs Formation and the Tropic Shale/Dakota Formation is assumed to be horizontal and smooth in the cross section due to the very gentle dip of the bedding (4u toward the southeast), although the natural contact may be irregular and undulating. Using the shear strength parameters of soil, bedrock, and the soil-bedrock contact, as determined in the laboratory, we used the SLIDE software program (Rocscience, 2013) to determine the factor of safety against sliding. The program performed translational analysis (failure along soil-bedrock contact) with respect to three groundwater conditions: (1) completely dry slope, (2) completely saturated slope, and (3) intermediate situations

Table 3. Input parameters used for stability analysis for three potential failure scenarios.

Soil–Tropic Shale Soil–Dakota Formation Soil–bedrock average

Friction Angle (u)

Cohesiona (psi)

Dry Densityb (pcf)

Saturated Densityb (pcf)

24 34 29

1.7 0 0.68

102.5 102.5 102.5

126.7 126.7 126.7

145 psi 5 1 MPa. 62.4 pcf 5 1 Mg/m3.

a

b

252

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 245–258

Cedar Canyon Utah Landslide Table 2. Extended.

Peak Cohesiona (psi)

Residual Cohesiona (psi)

Max

Min

Mean

St. Dev.

Max

Min

Mean

St. Dev.

2.3 1.8 1.7 1.5 2.3

0 0.5 0 0.5 0.6

1.4 1 1.1 1.2 1.4

1.3 0.7 0.9 0.6 0.8

1 n/a 0 0.7 2.15

0 n/a 0 0 1.25

0.6 n/a 0 0.5 1.7b

0.5 n/a 0 0.4 0.6

conditions, and we determined the friction angle needed for a safety factor of one, while keeping cohesion the same (0.68 psi or 4.7 kPa; Table 3). Figure 8 shows the friction angle required for limiting conditions (factor of safety. 5 1) for varying positions of the water table. When we used average strength parameters (Table 3), the factor of safety decreased from 1.02 to 1.00 (failure conditions) with the water table at 3.3 ft (1 m). However, the water table height needed to cause failure may be less than 3.3 ft (1 m), depending on the bedrock geology controlling the failure.

For example, the factor of safety for a dry soil–Tropic Shale condition was 0.8, so no water would be needed to induce failure. Thus, the water table rise required to induce failure would have depended on the proportion of Tropic Shale versus Dakota Formation along the failure surface. Considering that colluvial soil at the landslide site is classified primarily as poorly graded

sand with relatively high permeability, a small rise in the water table above the bedrock is very likely after heavy rainstorms.

Potential Modes of Failure Affecting Bedrock In addition to the stability analysis of the Cedar Canyon landslide, we performed a kinematic analysis using RockPack III software (Watts et al., 2012) on the discontinuities measured in the Straight Cliffs Formation in the landslide vicinity. The purpose of this analysis was to investigate the potential for different types of slope movement in Cedar Canyon and the hazard they pose to SR 14. Results of the kinematic analysis (Figure 9) clearly show that in addition to landslides involving colluvial soil, such as the Cedar Canyon landslide, there is great potential for a variety of other slope movements, such as wedge failures,

Figure 7. Slope cross section used for stability analysis, using the SLIDE software (Rocscience, 2013). Slope faces northeast.

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 245–258

253

Tizzano, Shakoor, and Lund

(Figure 10). The road was damaged at several places, and concrete barriers were damaged and moved, with most of the debris ending up in the drainage ditch. Figure 10 shows examples of rockfall hazards. It should be noted that because of the high cliffs formed by the Straight Cliffs Formation, even plane, wedge, or toppling failures descend the slope as rockfalls.

CAUSES OF LANDSLIDE

Figure 8. Variation of back-calculated friction angle for varying heights of water table and for a safety factor 5 1. Note: 3.3 ft 5 1 m.

plane failures, and toppling failures (Cruden and Varnes, 1996), along discontinuities in the bedrock comprising the walls of Cedar Canyon. The intersection of joint set 1 and the valley stress relief joints shows potential for wedge failures. The intersection of joint sets 1 and 2 shows potential for toppling failures. If the dip of the relief joints is gentler than the slope face, there is potential for plane failure. Discontinuity-controlled rockfalls (not subject to kinematic analysis) is another dominant mode of failure. For example, on December 10, 2012, following remediation of the 2011 Cedar Canyon landslide, a major rockfall detached from the cliff face above the repaired roadway and damaged SR 14 near mile marker 7.5

Cedar Canyon is in a semiarid region with few precipitation events. Colluvial soils formed in such regions often exhibit loose soil structure that is stable only as long as it remains dry (Turner, 1996). The loose nature of the colluvial soil at the Cedar Canyon landslide is corroborated by its relatively low blow count (N 5 12) and dry density (100 pcf; 1.6 Mg/ m3). The 2.13 in. (5.4 cm) of rain on October 6, 2011 (Figure 11), as well as a total precipitation of 2.68 in.

(6.8 cm) for October 2011, exceeded the mean precipitation of 1.58 in. (4.0 cm) for the month of October from 1981 to 2010 (Figure 11). The water percolated through the low-density colluvial soil and created the partially saturated conditions required for failure of a marginally stable slope. The remnants of the 2009 rockfall on the slope above SR 14 added weight to the colluvial soil. Additionally, a slope angle of ~30u was steep enough to contribute to the driving force.

Undercutting of the slope toe by Coal Creek also may have contributed to landslide initiation.

Another possibility is that the 2011 landslide is an old landslide that was reactivated by the October 2011 storm event. The 2011 landslide formed in colluvial material that had accumulated at the base of the Straight Cliffs Formation cliff and buried the Tropic Shale and Dakota Formation. Any number of landslides may have occurred in this material in the past; however, no studies have been conducted to identify their presence, and any old landslides present would have consisted chiefly of reworked colluvium.

LANDSLIDE REPAIRS Repairs implemented by UDOT to restore traffic on SR 14, are listed below:

Figure 9. Kinematic analysis of discontinuities measured within the Straight Cliffs Formation, using RockPack III software (Watts et al., 2012).

254

1. Landslide material was cleared away from the road, and the slope above the road (south side) was regraded to 2H:1V (where H 5 horizontal, V 5 vertical; Figure 12A). The slope below the road (north side) was regraded to 2H:1V where the slide occurred and 1.75H:1V west of the landslide.

2. A rip-rap–lined ditch was placed upslope of the road to collect surface runoff as well as any groundwater seeping from the toe of the slope

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 245–258

Cedar Canyon Utah Landslide

Figure 10. Examples of rockfalls, which may initiate as plane, wedge, or toppling failures higher up the slope, posing a hazard to SR 14: (a) debris from previous rockfalls in the canyon; (b) source area for the 2012 rockfall above SR 14 (Lund et al., 2012); (c) rockfall debris on SR 14 (Lund et al., 2012); and (d) damage to SR 14 from 2012 rockfall (Lund et al., 2012).

above the road (Figure 12a). The ditch is approximately 11 ft (3.3 m) wide, 2 ft (0.6 m) deep, and lined with a 1-ft-thick (0.3 m) rip-rap.

3. A modular block retaining wall (Figure 12b) was built on the east side of the landslide area, where the canyon narrows, to support the colluvial soil under SR 14.

4. Six 24-in.-diameter (61-cm-diameter) culverts, surrounded by filter fabric, were placed beneath SR 14 to drain water from the ditch to the north side of the road. At the end of each culvert, rip-rap was used on the downslope side of SR 14 to allow water from the culverts to run into Coal Creek without causing erosion.

5. Concrete barriers were placed along the road to retain small-size debris.

The repaired slope has a factor of safety of 1.1 for the average friction angle between colluvial soil and bedrock, and 1.3 for the friction angle between colluvial soil and the Dakota Formation. We obtained these factors of safety by dividing the tangent of the angle of internal friction by the tangent of the regraded slope angle. This approach ignores cohesion. The factor of safety values are marginal even without considering the effect of pore pressure due to precipitation events. By 2015, portions of the roadbed had begun

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 245–258

255

Tizzano, Shakoor, and Lund

Figure 11. Daily precipitation data from a National Oceanic and Atmospheric Administration (NOAA) station outside of Cedar City, UT (NOAA, 2012). The station is 5 mi (8 km) from Cedar City and 3 mi (1.9 km) from the landslide site. The elevation of the measuring gauge is 6,450 ft (1,954.5 m) above mean sea level (amsl). For comparison, the elevations of the three boreholes at the landslide site are: 7,183 ft (2,176.7 m) for borehole 8-1, 7,077.3 ft (2,144.6 m) for borehole 8-2, and 7,039.8 ft (2,133.3 m) for borehole 7.5-3. Note: 1 in. 5 2.54 cm.

showing the effect of renewed slope movement, particularly toward the east end of the slide. Additionally, there are still many overhangs above SR 14 in the Straight Cliffs Formation that may result in future rockfalls and damage to the road. Also, as indicated by the kinematic analysis, potential for other types of slope movement exists.

LIMITATIONS OF RESEARCH The study presented herein had the following limitations: 1. No shear zones were reported on the logs for any of the three UDOT boreholes. Therefore, the location of the failure plane was based on the presence of soft to very soft weathered bedrock along the soilbedrock contact.

2. The UDOT inclinometer data showed no evidence of movement during the 3 month monitoring

256

period. Again, the exact location of the failure could not be ascertained.

3. The colluvial soil is very heterogeneous in nature, ranging from large boulders to clay-size material.

Therefore, laboratory tests, using small samples, could not fully characterize the engineering properties of the colluvium, especially the shear strength parameters of the colluvial soil and of the soilbedrock contact.

4. The bedrock underlying the landslide consists of the Tropic Shale and Dakota Formation, which have different strength parameters. However, since the two formations are not differentiated at the study site (Biek et al., 2012), their individual influences on shear strength along the soil-bedrock contact and on stability cannot be ascertained.

5. For these reasons, our stability analysis is preliminary in nature. Since this part of the canyon has a long and complex history of slope movements, additional research is required to fully refine all the factors affecting slope stability at this site.

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 245–258

Cedar Canyon Utah Landslide

and buildup of pore pressure due to a rainstorm prior to failure.

3. The stability analysis, using the SLIDE software program, indicates a factor of safety of 0.8 to 1.2 for dry conditions and 0.3 to 0.4 for fully saturated conditions prior to the landslide. Back-calculation analysis indicates that, for an average friction angle of 29u between the colluvial soil and the undivided Tropic Shale/Dakota Formation bedrock, a 3.3 ft (1 m) rise of the water table above the soil-bedrock contact is required to cause failure; however, failure may occur with a lesser rise in the water table if the majority of the bedrock along the failure surface consists of Tropic Shale.

ACKNOWLEDGMENTS The authors would like to thank UDOT, especially Keith Brown, Dave Fadling, and Sam Grimshaw, for permission to work on this project and for sharing field photographs, subsurface and geophysical data, remediation plans, and other relevant information. The authors would also like to thank the three reviewers for their valuable comments, which greatly helped to improve the quality of this paper.

Disclaimer The authors are solely responsible for the contents of this this paper, including the accuracy of the data. The contents do not reflect the official views or policies of UDOT.

REFERENCES

Figure 12. Landslide repairs: (a) regraded slope above SR 14 and riprap-lined drainage ditch, and (b) a view of the modular block retaining wall looking to the west.

CONCLUSIONS Based on the results of this study, we draw the following conclusions: 1. The Cedar Canyon landslide exhibited a combination of rotational and translational movement, with movement chiefly along the contact between the colluvial soil and the underlying bedrock. Rotational movement was confined to the main scarp area.

2. The causes of the Cedar Canyon landslide include a relatively steep colluvial soil slope of ~30u, extra weight on the slope from previous large rockfalls,

AMERICAN SOCIETY FOR TESTING AND MATERIALS, 2011, Soil and Rock: Annual Book of ASTM Standards, Volume 4.08: ASTM International, Conshohocken, PA. BIEK, R. F.; ROWLEY, P. D.; ANDERSON, J. J.; MALDONADO, F.; MOORE, D. W.; EATON, J. G.; HEREFORD, R.; AND MATYJASIK, B., 2012, Interim Geologic Map of the Panguitch 309 6 609 Quadrangle, Garfield, Iron, and Kane Counties, Utah: Utah Geological Survey Open-File Report 599, scale 1:100,000.

BRABB, E. E., 1989, Landslides: Extent and economic significance in the United States. In Brabb, E. E. and Harrod, B. L. (Editors), Landslides: Extent and Economic Significance, Proceedings of the 28th International Geological Congress, Symposium on Landslides, Washington D.C., July 1989:A. A. Balkema, Rotterdam, Netherlands, pp. 25–50.

CRUDEN, D. M. and VARNES, D. J., 1996, Landslide types and processes. In Turner, A. K. and Schuster, R. L. (Editors), Landslides: Investigation and Mitigation: Transportation Research Board, Special Report 247, National Academy of Sciences, Washington, DC, pp. 36–71.

DEERE, D. U. and MILLER, R. P., 1966, Engineering Classification and Index Properties for Intact Rock: Technical Report No. AWFL-TR-65-116, University of Illinois, Urbana, IL, pp. 167.

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 245–258

257

Tizzano, Shakoor, and Lund ERSKINE, M. C.; FAULDS, J. E.; BARTLEY, J. M.; AND ROWLEY, P. D., 2001, The Geologic Transition, High Plateaus to the Great Basin—A Symposium and Field Guide: The Mackin Volume, Publication No. 30, Utah Geological Association, Salt Lake City, UT, 345 p.

GIRAUD, R. E. and LUND, W. R., 2005, The June 3, 2005, Black Mountain Debris Flow, Iron County, Utah: Utah Geological Survey Technical Report, 1 p.

GREGORY, H. E., 1950, Geology of Eastern Iron County, Utah: Utah Geological and Mineralogical Survey Bulletin 37, pp. 45–50.

HARTY, K., 1989, The Cedar Canyon landslide: Utah Geological Survey, Survey Notes, Vol. 23, No. 2, 14 p.

HARTY, K., 1991, Landslide Map of Utah: Utah Geological and Mineral Survey Map 133, 13 p.

HINTZE, L. F. AND KOWALLIS, B. J., 2009, Geologic History of Utah, 2nd ed.: Geology Studies Special Publication 9, Brigham Young University, Provo, UT, 225 p.

HOEK, E.; CARRANZA-TORRES, C.; AND CORKUM, B., 2002, HoekBrown failure criterion. 2002 edition. In Proceedings of the North American Rock Mechanics Symposium–Tunnelling Association of Canada Conference, Toronto, Canada, 2002, Volume 1, pp. 267–273.

INTERNATIONAL SOCIETY FOR ROCK MECHANICS (ISRM), 1979, Suggested methods for determining water content, porosity, density, absorption, and related properties and swelling and slake durability index properties: International Society for Rock Mechanics Commission on Standardization of Laboratory and Field Tests: International Journal of Rock Mechanics and Mining Sciences & Geomechanics Abstracts, Vol. 22, pp.

141–156.

LUND, W. R.; KNUDSEN, T. R.; AND BROWN, K. E., 2009, Large rock fall closes highway near Cedar City, Utah: Utah Geological Survey, Survey Notes, Vol. 41, No. 3, pp. 8–9.

LUND, W. R.; KNUDSEN, T. R.; AND FADLING, D., 2011, Another large landslide closes highway near Cedar City, Utah: AEG News, Vol. 54, No. 4, pp. 24–25.

LUND, W. R.; KNUDSEN, T. R.; AND FADLING, D., 2012, Another large landslide closes highway near Cedar City, Utah: Utah Geological Survey, Survey Notes, Vol. 44, No. 2, pp. 4–5.

258

MARINOS, P. and HOEK, E., 2000, GSI: A geologically friendly tool for rock mass strength estimation. In GEOENG 2000, Melbourne, Australia: Technomic Publishers, Lancaster, PA, pp. 1422–1446.

NATIONAL OCEANIC AND ATMOSPHERIC ADMINISTRATION (NOAA), 2012, Daily Summaries Station Details: Cedar City 5 E, UT US: Electronic document, available at http://www. ncdc.noaa.gov/cdo-web/datasets/GHCND/stations/GHCND: USC00421260/detail PARK, H. J. AND WEST, T. R., 2002, Sampling bias of discontinuity orientation caused by linear sampling technique: Engineering Geology, Vol. 66, pp. 99–110.

PAROWAN TIMES 1915, Newsy notes from Cedar City: Parowan Times, November 24 1915 pp., 2: Electronic document, available athttp://digitalnewspapers.sandbox.lib.utah.edu PITEAU, D. R. and MARTIN, D. C., 1977, Description of Detailed Line Engineering Mapping Method:Rock Slope Engineering, Part G: Reference Manual FHWA-13-97-208, Federal Highway Administration, Portland, OR, 29 p.

ROCSCIENCE, 2013, Determining Input Parameters for Rocfall Analysis: RocFall Software, University of Toronto, Ontario, Canada.

SHAKOOR, A. AND BAREFIELD, E. H., 2009, Relationship between unconfined compressive strength and degree of saturation for selected sandstones: Environmental & Engineering Geoscience, Vol. 15, No. 1, pp. 29–40.

SHAKOOR, A. AND BONELLI, R. E., 1991, Relationship between petrographic characteristics, engineering index properties and mechanical properties of selected sandstones: Bulletin of the Association of Engineering Geologists, Vol. 28, No. 1, pp. 55–71.

TURNER, A. K., 1996, Colluvium and talus. In A. K. Turner, R. L. Schuster, (Editors), Landslides: Investigation and Mitigation: Special Report 247, Transportation Research Board, National Academy of Sciences, Washington, DC, pp. 525–549.

U.S. GEOLOGICAL SURVEY (USGS), 2002, Panguitch, Utah, 7.5 Quadrangle: U.S. Geological Survey, scale 1:24,000.

WATTS, C. F.; GILLIAM, D.; HROVATIC, M.; AND HONG, H., 2012, ROCKPACK III for Windows: Rock Slope Stability Computerized Analysis Package, Part One-Stereonet Analysis: RockWare, Inc., Earth Science and GIS software.

WYLLIE, D. C. AND MAH, C. W., 2004, Rock Slope Engineering, 4th ed.: Spon Press, New York, 431 p.

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 245–258

Investigating Movement and Characteristics of a Frozen Debris Lobe, South-Central Brooks Range, Alaska JOCELYN M. SIMPSON MARGARET M. DARROW SCOTT L. HUANG University of Alaska Fairbanks, P.O. Box 755800, Fairbanks, AK 99775-5800

RONALD P. DAANEN TRENT D. HUBBARD Alaska Department of Natural Resources, Division of Geological and Geophysical Surveys, 3354 College Road, Fairbanks, AK 99709

Key Terms: Frozen Debris Lobe, Permafrost, Landslide, Alaska

ABSTRACT Frozen debris lobes (FDLs) are large masses of soil, rock, incorporated organic material, and ice moving down permafrost-affected slopes. In this paper, we focus on FDL-A in the south-central Brooks Range, Alaska, which is an impending geohazard to the Dalton Highway, located just under 40 m away from the highway embankment. We present the results of multi-faceted research, including field-based studies, laboratory testing of soil samples, slope stability analysis, and a geographic information system (GIS) analysis. Subsurface instrumentation indicates that major movement of FDL-A occurs in a shear zone 20.6 to 22.8 m below the ground surface, with temperature-dependent internal flow as a secondary movement mechanism. Surface measurements show an overall average rate of movement of 1.2 cm per day, which is an increase over historic rates. The slope stability analysis required a back analysis to determine soil strength parameters at failure, resulting in cohesion values between 43 and 53 kPa and friction angles between 106 and 166. The modeling results indicated a high sensitivity to pore-water pressure and cohesion. This is critical since the melting of massive ice and thawing of frozen soil will increase pore-water pressure and lower shear strength, resulting in the acceleration of FDL-A towards the Dalton Highway.

The GIS analysis also provided insight into the movement and instability of FDL-A and provided groundwork for a GIS protocol for examining catchment and lobe features of all FDLs along the highway corridor.

INTRODUCTION Along the Dalton Highway corridor, in the southcentral Brooks Range, Alaska, large masses of frozen debris are moving down permafrost-affected slopes.

These features, named frozen debris lobes (FDLs), are located in a region that is characterized by steep terrain, active slope processes, and continuous permafrost. FDLs consist of soil, rock, incorporated organic material, and ice, and with their recent increasing rate of movement, they have the potential to threaten the Dalton Highway and the TransAlaska Pipeline System (TAPS) within the highway corridor. We identified more than 40 FDLs along the highway corridor between milepost (MP) 194 and MP 230, selecting eight FDLs due to their proximity to infrastructure for further examination, including field investigations and remote-sensing analysis. While our focus has been on FDLs within the highway corridor, examination of available imag‐ ery has proven that many FDL-like features are present throughout the Brooks Range. One of these lobes, FDL-A, is just under 40 m from the highway and has been the main focus of FDL research since 2008 (Figure 1). Recent data indicate that FDL-A’s rate of movement has been increasing (Daanen et al., 2012), and ongoing research has helped us better understand this feature.

In this paper, we present the results of two major aspects of the research. We discuss: (1) a geotechnical investigation and analysis of FDL-A, which consisted of: (a) field-based studies, including data from drilling and instrumentation, surface measurements made using a differential global positioning system (DGPS), and observations from several years of ongoing monitoring; (b) laboratory testing of soil samples to determine strength parameters; and (c) slope stability analysis using a limit equilibrium approach;

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 259–277

259

Simpson, Darrow, Huang, Daanen, and Hubbard

Figure 1. Location of FDL study area. (a) Study area (indicated by the black rectangle) shown relative to select populated locations and major road systems in Alaska. (b) Eight FDLs selected for further analysis (shown in yellow), the Dalton Highway (shown in red), and TAPS (shown in blue). (c) Large-scale view of FDL-A. (Base map data are from Hubbard et al. [2011] and Esri [2015].)

and (2) a geographic information system (GIS) analysis of the lobe and its catchment. The purpose of the GIS analysis was to develop a tool for examining other FDLs along the Dalton Highway corridor. We used the results from both parts of the research to determine the mechanisms of movement and movement rate of FDL-A and its defining characteristics in order to understand these landslides in frozen soil and to predict their response to varying parameters such as changes in climate.

BACKGROUND Snow, glaciers, and permafrost in cold mountain areas are sensitive to changes in atmospheric conditions, and mass wasting often occurs in mountain areas with steep slopes due to changes in geomorphic processes (Haeberli and Beniston, 1998). The stability of permafrost terrain is vulnerable to thermal changes because thawing reduces the strength of frozen soil and fractured bedrock. Ice-rich soils undergo thaw consolidation and experience an increase in pore-water pressure, which then leads to increased instability (Harris, 2005). In areas with steep mountainous terrain and thawing of soil and rocks, this

260

instability can result in debris flows, landslides, deep-seated bedrock failures, rock slides, rock falls, creep-related processes, blocked drainage, and active-layer detachment failures (Harris et al., 2001; Kääb et al., 2005).

Many types of mass movement related to degrading permafrost have been studied in parts of Europe and Canada. Shallow landslides, mud and debris flows, and rock falls are commonly seen in high mountain areas of Europe (Harris and Vonder Mühll, 2001), and active-layer detachments and debris flows have been observed in Yukon Territory, Canada (Harris and Lewkowicz, 2000; Huscroft et al., 2003). Debris flows researched in Europe are frequently initiated as translational slides over the permafrost table, developing into debris flows as water contents and slope gradients increase (Harris et al., 2009). In Canada, active-layer detachment slides can result from rapid thawing at the base of the active layer, and they are usually triggered by forest fires (Lipovsky and Huscroft, 2006). Debris flows and landslides observed in the Yukon Territory occur where thawing takes place in ice-rich soils, resulting in a great reduction in volume and strength (Huscroft et al., 2003). Another common

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 259–277

Movement and Characteristics of a Frozen Debris Lobe

type of slope movement seen in these areas, as well as in Alaska, is solifluction, which is a slow downslope movement of saturated soil by frost creep, needle ice creep, or gelifluction, and it is caused by poor drainage of pore water and freezing during the winter months (Tart, 1996; Harris and Lewkowicz, 2000; Davis, 2001; and Matsuoka, 2001, 2014). As the soil thaws in the spring, it becomes saturated and slowly creeps downslope. As the upper layer thaws and drains, it becomes more stable but still moves on top of ice-rich materials.

This movement eventually stops when the ground refreezes (Tart, 1996).

Rock glaciers are a common permafrost-related movement feature in mountainous areas and are found in many major mountain systems (Davis, 2001). Rock glaciers are described as lobate or tongueshaped masses forming at the base of cliffs and talus slopes or extending from the end of small glaciers. In Alaska, Wahrhaftig and Cox (1959) investigated rock glaciers in the Alaska Range and found they consisted of a layer of coarse block rubble overlying a thicker layer of coarse blocks mixed with sand and silt material. The ice in these rock glaciers originated from compacted snow, melting snow or rain, or groundwater that rose beneath the talus and froze upon contact with cold air. Active rock glaciers were found to be almost entirely rock debris mixed with ice, accounting for approximately one half of a typical rock glacier’s volume, and their movement resulted from deformation of the interstitial ice (Wahrhaftig and Cox, 1959; Davis, 2001). The slope-dependent movement rates of rock glaciers are typically up to 1 m per year; however, in warm alpine settings, this rate could be higher (Davis, 2001; Barboux et al., 2014). Four major rock types that form blocky and boulder-like debris found in rock glaciers are granite, gneiss, sandstone, and limestone, while platy rock and micaceous schists are unfavorable to rock glacier formation (Wahrhaftig and Cox, 1959; Haeberli et al., 2006).

As part of the continuum of slope movement processes, frozen debris lobes were originally identified as flow slide deposits exhibiting slow downslope movement (Hamilton, 1978, 1979, 1981). They were later described mostly as inactive rock glaciers (Kreig and Reger, 1982; Brown and Kreig, 1983); however, ongoing research has shown they differ from rock glaciers with respect to source, composition, rate, and mechanism of movement. These features were “rediscovered” in 2008, and preliminary studies, mainly on FDL-A, identified indicators of ongoing movement such as leaning and split trees, overrun and buried trees, over-steepened slopes, frozen blocks showing slickens from sliding, and uplifting of frozen topsoil

in front of the lobe (Daanen et al., 2012). Remotesensing analysis carried out by Daanen et al. (2012) revealed that the rate of movement for FDL-A increased between 1955 and 2008, with rates increasing from 0.58 to approximately 1 cm per day. While these studies were preliminary, they provided insight into the growing hazards of slope instability and mass movement in areas where permafrost warming and degradation are occurring.

Through several years of field monitoring, we have observed evidence of movement and the geomorphic characteristics of FDL-A. Originally presented in conference papers (Darrow et al., 2013, 2015) and posters (Hubbard et al., 2013; Spangler et al., 2013), this paper summarizes and builds upon those initial analyses, including updated data sets and additional geotechnical and GIS analyses. Investigation of the catchment—the area upslope of the lobe that serves as its debris source—indicates that weak and platy phyllitic and schistose bedrock moving downslope mainly through solifluction and rock falls is the source of debris for FDL-A (Daanen et al., 2012; Spangler et al., 2013). As FDL-A continues to move downhill, evidence of its movement is ubiq‐ uitous across its surface in many forms, including longitudinal cracks (Figure 2a). Trees that lean with ongoing near-surface movement and prolific splitting of trunks are evidence of tension at the surface (Figure 2b). Trees and shrubs also are knocked down and buried at the toe of the lobe (Figure 2c). Shear at the toe is most easily observed during the winter as the frozen, rigid lobe moves over undisturbed ground leaving a void space underneath that collapses the following summer (Figure 2d).

In 2009, we observed that the main drainage on the surface of FDL-A shifted from its original channel into a new channel draining to the southwest and over the left flank (Figure 3). This resulted in significant erosion of that portion of the left flank. The creek moved back to its original channel in June 2014, but it bifurcated in August of that same year, flowing in both channels. As a result of the channel movement, there was an increase in longitudinal cracking and destabilization of the forest near the left flank of FDL-A (Darrow et al., 2015).

The summer of 2014 was the wettest on record for parts of interior Alaska, and significant rainfall caused appreciable changes in FDL-A. Perhaps the most dramatic changes occurred in two retrogressive thaw slumps (RTS), a lower RTS near the left flank and a higher RTS in the catchment area (Figure 3). The significant rainfall exposed massive ice in the headscarps of both of these areas.

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 259–277

261

Simpson, Darrow, Huang, Daanen, and Hubbard

Figure 2. Evidence of ongoing movement of FDL-A: (a) longitudinal crack (dashed red line) among leaning and falling trees; (b) split and leaning trees near the left flank; (c) trees and shrubs knocked over, buried, and incorporated into FDL-A at the toe; and (d) a void (indicated by the red arrow) under the toe of FDL-A as it shears over undisturbed ground.

METHODS Geotechnical Investigation and Analysis Field Studies Subsurface Investigation.—The focus of our research has been FDL-A, due to the imminent threat it presents to the Dalton Highway, the main transportation route between Interior Alaska and the North Slope (Figure 1). During September 2012, we worked with Alaska Department of Transportation and Public Facilities (ADOT&PF) personnel to conduct a drilling program on FDL-A to: (1) obtain soil samples for laboratory testing; (2) determine the thickness and composition of the lobe; and (3) install instruments to measure slope movement, temperature, and water pressure. In total, eight boreholes were drilled with depths ranging between 3.0 m and 30.5 m, four on and four immediately adjacent to the lobe to the south and west. All boring locations are shown in Figure 4; however, in this paper, we focus on results ob‐ tained from TH12-9004, an instrumented borehole.

262

All sampled soils were classified using the Unified Soil Classification System (USCS).

Adjacent to the lobe, soil ranged from silty sand with gravel (SM) to sandy silt with gravel (ML), overlying chloritic schist bedrock on average 3.7 m below the ground surface (bgs). In the undisturbed area to the south of FDL-A, the soil was ice-rich, with as much as 50 percent visible ice by volume. Subsequent measurements indicated the average permafrost temperature adjacent to the lobe was −2.2uC, and the permafrost table was approximately 0.6 m bgs. In contrast, the lobe consisted of frozen silty sand with gravel (SM) and silty gravel with sand (GM) and contained minor ice (i.e., less than 1 percent ice per volume, occurring as occasional veins [Vr] or as coatings on particles [Vc]; ice descriptions follow Pihlainen and Johnston, 1963). We drilled borehole TH12-9004 to 30.5 m depth, successfully penetrating FDL-A. This boring indicated the lobe was 26.4 m thick, overlying chloritic schist bedrock. Subsequent temperature measurements indicated the permafrost table was between 2 and 2.5 m bgs on FDL-A, and it averaged −1.1uC below the

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 259–277

Movement and Characteristics of a Frozen Debris Lobe

Figure 3. FDL-A surface features: (a) scarps, longitudinal cracks, and creeks, including the location of two retrogressive thaw slumps (RTS); and (b) outline of FDL-A and major longitudinal cracks. (Base map is from 2011 LiDAR image [Hubbard et al., 2011].)

depth of zero amplitude (Darrow et al., 2013). Within TH12-9004, we installed an automated slope movement measurement device called a MEMS-based inplace inclinometer (M-IPI), along with two vibrating wire (VW) piezometers and a thermistor string.

Within the first month after installation, the M-IPI data indicated a shear zone located at 20.6 to 22.8 m bgs at this location, as well as slow to moderate flow above the shear zone (Darrow et al., 2013). We placed the VW piezometers where we saw evidence of water pressure during drilling. Both piezometers recorded water pressure immediately after installation; the lower piezometer failed within 56 hours, likely due to the water pressure exceeding its range. The upper piezometer indicated a potentiometric surface about 10.8 m above the ground surface. By the end of October 2012, the casing and instrumentation within TH12-9004 sheared at 20 m. After shearing, we noted that the propylene glycol used to fill the casing had risen above the ground surface and was dripping out of the top of the casing. Additionally, the water pressure within the shear zone was high enough to cause fluid to enter into the insulation of the sheared thermistor string, and travel up the boring and out to the data logger enclosure, where we observed fluid puddled at the bottom of the otherwise dry enclosure.

These observations support the high pore-water pressure measurements. Before shearing, the M-IPI

measured approximately 79.2 cm of movement in 31 days (or approximately 2.5 cm per day), indicating a large increase in the rate of movement over the previous average rate determined from remote-sensing analysis.

Surface Movement Measurements.—In order to gather more information on the movement of FDL-A, we placed surface markers on the lobe in October 2012 for repeated measurements with a DGPS unit, having horizontal and vertical accuracies of ¥5 cm.

These markers formed a longitudinal transect and three cross-sectional transects, along with two control points on either side of the lobe (subsequent DGPS readings of the control points indicated that their positions varied by less than 5 cm over the measurement period, or within the accuracy of the device). Measurements were made twice in November 2012, and once in April, June, and August of 2013, June and August of 2014, and March, May, and August of 2015. During these trips to the study area, we also collected data from the instrumentation installed during drilling, and we made visual observations of FDL-A. Laboratory Testing of Strength Parameters We performed direct shear testing on frozen samples to determine the internal friction angle and cohesion of

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 259–277

263

Simpson, Darrow, Huang, Daanen, and Hubbard

Figure 4. Surface movement of FDL-A, measured from October 2012 to August 2015. The arrows indicate the direction of movement for each marker, and the numbers are the amount of movement in meters color-coded to correspond to average rates provided in Figure 5 (blue—upper lobe; red—middle lobe; yellow—lower lobe; green—levee). The green dashed lines indicate the location of the levees. A-A9 is the location of the cross section for volume calculations. Borehole locations are indicated by green circles, and TH12-9004 (a key boring) is labeled. (Base map is from Hubbard et al. [2011].)

the FDL-A soil. We chose this testing method based on the amount of soil sampled and relative ease of conducting tests on frozen samples. The samples were obtained during the September 2012 drilling program using a split-spoon sampler, and they were transported and stored in a freezer until testing could be carried out. All samples were taken from TH12-9004, in which all soil tested was classified as silty sand with gravel (SM). The direct shear tests, following ASTM D3080-90 (ASTM, 1990b), were carried out in a cold room to maintain the frozen state of the soil throughout testing. To ensure as little temperature fluctuation as possible, we modified the direct shear device by constructing an insulated box around the main shearing apparatus, in which temperatures were monitored while testing was in progress. Additionally, we adjusted the rate of testing to simulate the lobe’s rate of movement. Due to limitations in controlling nearfreezing temperature settings, the cold room temperature was set to the warmest sub-freezing temperature it would maintain, which was −1.49uC, and about 0.4uC colder than what we measured within the lobe.

The insulated box helped to maintain this temperature within ¡0.05uC. 264

Using the average rate of movement of 2.5 cm per day from the October 2012 M-IPI data, three series of tests were performed on samples taken from 12.3 m, 13.9 m, and 24.7 m bgs. Each sample was cut to a length of 2.54 cm to fit into the shearing box. The normal stresses applied were calculated using the overburden weight of the soils above the center of the shear zone. Next, this normal stress (i.e., 441 kPa) was divided in half and doubled in order to perform three tests within one series. This resulted in normal stresses of 220.5, 441, and 882 kPa. The number of tests performed for each series was determined by the amount of soil samples available.

Slope Stability Analysis We performed a slope stability analysis using SLOPE/ W 2007 software, which utilizes a limit equilibrium approach (GEO-SLOPE International LTD., 2008).

To produce the geometry of FDL-A for the slope stabil‐ ity analysis, a light detection and ranging (LiDAR) digital elevation model (DEM) was used to create a profile of the lobe surface. The bedrock surface underneath the lobe was estimated using the known bedrock depths from the boreholes on the lobe and in the undisturbed

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 259–277

Movement and Characteristics of a Frozen Debris Lobe

area to the south, as well as natural bedrock exposures in the catchment area and the general nature of the slope surface created by a second profile off the lobe. The pore-water pressure was estimated using the measured value within the lobe, and the bedrock was defined as impermeable without having any pore-water pressure because of the permafrost in the area. Historic earthquake information and a map of ground motion hazard values produced by the U.S. Geological Survey were used to determine the seismic load for this analysis (Wesson et al., 2007; Alaska Earthquake Center, 2013). In the last 115 years, 90 earthquakes have occurred within a 100 km radius of FDL-A. These earthquakes had an average magnitude (M) of M3.5, with M5.5 as the larg‐ est event. Based on this information, we used a horizontal acceleration of 0.15g for this analysis, because seismic slope stability evaluation assumes the failure mass is horizontally accelerated (Dismuke, 2002), with the vertical force having less effect on stability (Chen et al., 2008).

GIS Analysis of FDL-A and Catchment Landslide investigations are commonly performed using GIS analysis and high-resolution imagery, and these investigations can assist in producing hazard assessments. Lyle and Hutchinson (2006) performed slope hazard identification and landslide inventory mapping in areas of degrading permafrost in the Yukon Territory, Canada. They looked at landslide attributes and primary controls on permafrost distribution, including slope angle, slope aspect, vegetation cover, and drainage conditions. While this study took place in an area that contained discontinuous permafrost, the same principles can be applied to areas of slope instability in continuous permafrost regions.

Following a similar approach as Lyle and Hutchinson, we completed a comprehensive GIS analysis of FDL-A, examining certain characteristics of the lobe and its catchment. The purpose of this analysis was to develop a procedure to serve as a guide for future analysis of other FDLs along the Dalton Highway corridor as additional high-resolution imagery is obtained to identify similar elements that may con‐ tribute to the formation and movement of these features. We hypothesize that a statistical analysis of these attributes may yield correlations to lobe formation and movement. The components of this analysis included examining the surface area and slope angle of the catchment and the lobe, slope aspect, and rock type within the catchment, and vegetation coverage of both the catchment and lobe, based on the following rationale. The surface area of the catchment contributing material to the lobe may correlate to the lobe size.

Steeper slopes in the catchment affect how much

debris contributes to the FDL, and slope angles on the FDL could affect the rate and mechanism of movement. Slope aspect of the catchment can affect permafrost conditions and thus the stability of the material, while rock types may correlate to the formation and amount of debris contributing to FDLs. Lastly, ground cover provided by vegetation aids in maintaining cooler temperatures at the surface and thus affects permafrost conditions. It also contributes to overall soil stability, and the absence of vegetation can indicate potential areas of movement on the lobe.

In addition to the FDL attribute analysis, we mapped scarps, longitudinal cracks, and surface drain‐ age on FDL-A, and we determined volume per unit width to estimate the amount of material traveling towards the highway. All analyses were conducted with ArcMap using LiDAR imagery of the project area acquired in 2011. The majority of this study used LiDAR bare earth, or last return, data, which display the ground surface. The highest hit, or first return, data, which are associated with the first returned laser pulse, display the highest feature in the landscape and were used for the vegetation analysis.

RESULTS Geotechnical Investigation and Analysis Subsurface and Surface Movement Analysis From October 2012 to August 2015, the average total movement on FDL-A was approximately 12.8 m, with an average rate of 4.5 m per year (or 1.2 cm per day). The upper portion of FDL-A demonstrated the most movement, between 17.8 and 19.3 m, while the rest of the lobe moved between 7.8 and 15.4 m during this time frame (Figure 4). We also observed from these surface measurements that FDL-A has formed levees that move at a slower rate and constrain its lateral flow (Darrow et al., 2015); the levee surface marker movements are not included in the average rates.

FDL-A’s rate of movement not only varies spatially, but it also changes throughout the year (Figure 5).

The highest rates we measured occurred during the fall of 2012, and they decreased with the onset of winter. There are two notable downticks in the data, occurring in June 2013 and May 2015, which are most significant for the upper lobe and may be due to a reduction of water supply after the spring snowmelt and before summer rains. In general, there is more volatility in the movement rate for the upper lobe than for the lower lobe or the levees. Over our measurement period, the upper lobe moved up to 1 cm per day faster than the lower lobe. Unfortunately, as the surface measurements are dependent on the

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 259–277

265

Simpson, Darrow, Huang, Daanen, and Hubbard

Figure 5. Rate of movement trends for FDL-A.

limited number of trips made to the field, they lack temporal resolution. For example, we suspect that an increase in movement occurs during the late fall, which is not present in our data after 2012; however, the trends in the data indicate that a relationship among temperature, precipitation, and movement exists, with FDL-A moving faster when it reaches both maximum thawing and internal temperature, and moving slower when minimum temperatures are reached, and there is a general lack of water.

Despite the shearing of the M-IPI in 2012, the upper portion of the device continues to record slope movement and temperature data. Since the data now lack the fixed readings below the shear zone, we determined the surface location of the uppermost measurement using the DGPS readings. Figure 6a is a summary of the corrected data from October 27, 2012, until August 21, 2015. Since their installation in September 2012, the top of the casing and the M-IPI device have moved 15.7 m downslope.

The data presented in Figure 6b are from the portion of the M-IPI that is still functioning. This does not include any movement within or below the shear zone. The data indicate an ongoing leaning of the instrument downhill, with a greater amount of leaning near the surface. In addition to the leaning, two anomalous zones are present in the casing at 7.6 m and 16.3 m bgs. A diagram of the TH12-9004 boring log in Figure 6c is presented for comparison. The disturbance at 7.6 m bgs (indicated by the upper arrow) corresponds to where we lost fluid during drilling and grout during back-filling the boring. The lower disturbance (indicated by the lower arrow) corresponds to the location where evidence of water pressure was

266

observed during drilling. These anomalous areas most likely represent sagging or buckling of the casing within voids located at these depths.

Monthly rates of movement and corresponding temperatures collected from the M-IPI from March 2013 to August 2014 are presented in Figure 7. As with the surface marker measurements, these data also indicate that the internal movement below the active layer (i.e., greater than 2.5 m bgs) steadily increases with time until September (Figure 7a and d), at which point the internal flow rate steadily decreases until March (Figure 7b). Then the cycle repeats the following spring (Figure 7c and e). We observed similar trends in the 2014–2015 data. The rate of movement within the active layer becomes significantly higher during the summer as the active layer thaws. During late summer and early fall, the highest rate of movement occurs in a zone between the surface and the permafrost table (Darrow et al., 2015). In other words, the highest rate of movement becomes sandwiched between the cooling surface and the colder permafrost below.

The M-IPI device also contains temperature sensors that we have relied on for data since the failure of the thermistor string during shearing. Data collected from these sensors are presented in Figure 7f through h.

As the temperature sensors are located every 0.3 to 0.38 m for the length of the M-IPI, the symbols in Figure 7 only indicate the data series, not the positions of the sensors. Snowfall was below normal during the early winter of 2012–2013, which contributed to colder ground temperatures as compared to the warmer temperatures during the winter of 2013–2014. Additionally, these data indicate that the rate of movement

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 259–277

Movement and Characteristics of a Frozen Debris Lobe

Figure 6. Cumulative displacement measurements. The shear zone is approximately located in (a) with dashed lines. The data in (a) were adjusted to include all post-shearing movement based on the surface measurements. The data in (b) show only movement measured above the shear zone at 20.6 m bgs. The boring log in (c) indicates the shear zone with a red rectangle. Wood fragments are shown as brown rectangles, and the bedrock surface is at 26.4 m bgs. Frozen ground is represented by the black bar along the left side of the log with the white portions representing questionably frozen areas. The red arrows indicate the locations of anomalous zones present in the data in (b).

within FDL-A is highly dependent on temperature, as a direct correspondence exists between higher rates of movement and higher temperatures (Darrow et al., 2015).

Laboratory Testing of Strength Parameters: Results and Analysis Each direct shear test took approximately 12 hours to complete using the slow shearing rate of 2.54 cm per day. The samples were examined (Figure 8) at the end of each test and oven-dried to obtain their moisture content according to ASTM D2216-90 (ASTM, 1990a). One of the three series of tests failed; upon examination of the samples, we observed that each sample contained large rock fragments (Figure 8b and c).

This resulted in the pulverization of the upper portion of the sample as the test progressed. This series of tests was not used for further analysis. Results from the direct shear testing are presented in Figure 9 and

Table 1. While all the soil tested from this borehole was classified as silty sand with gravel (SM), visual observations of these samples indicated some differences. Sample 12-2515, from a depth of 12.3 m bgs, was fine-grained with some small rock fragments, and no significant amount of ice was visible. Sample 122521, which was from 24.7 m bgs and close to the shear zone, did not contain as many rock fragments and had Table 1. Direct shear results for FDL-A frozen soil samples; volumetric moisture content values show the range for three samples in each test series.

Sample

Volumetric Moisture Content (%)

12-2515

18.2–22.6

12-2521

31.8–49.6

Cohesion (kPa)

Internal Friction Angle (u)

106 (peak) 101 (residual) 136 (peak) 67 (residual)

11 (peak) 10 (residual) 16 (peak) 16 (residual)

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 259–277

267

Simpson, Darrow, Huang, Daanen, and Hubbard

Figure 7. Daily rate of movement (a–e) and subsurface temperatures (f–h) of FDL-A, as measured by the M-IPI device. Data show the average daily rate by month, and periods are broken into spring through fall of 2013 (a, d, f), winter of 2013–2014 (b, g), and spring through summer of 2014 (c, e, h). Graphs (d) and (e) are close-up views of the upper 5 m; the blue horizontal line indicates the active layer depth, and the green dashed horizontal line is the depth at which movement rate increases (Darrow et al., 2015).

268

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 259–277

Movement and Characteristics of a Frozen Debris Lobe

Figure 8. Frozen soil samples after direct shear testing: (a) successfully sheared sample; (b) large rock fragments pulverized the upper portion of the sample in a failed test; (c) most fragments of rock in the failed samples measured at least 3 cm in length.

a noticeable amount of ice, which is indicated by higher moisture contents. The differences in the peak and re� sidual internal friction angles of the two samples were not significant; however, sample 12-2521 demonstrated higher peak strength results and a larger difference in the peak and residual cohesion.

Slope Stability Analysis Results For this analysis, the unit weight, cohesion, and internal friction angle for the bedrock were estimated using published values (Gonzalez de Vallejo and Ferrer, 2011). The properties of the frozen silty sand with gravel found within the lobe were determined directly from laboratory testing and averages of short-term frozen direct shear testing results (Table 2); however, these values did not take into account the time-dependency of frozen soil. The strength of frozen soil varies with loading rate, as faster loading rates result in increased strength. Due to this time-dependent component, long-term strength is generally much smaller than short-term strength. The period of

time in which laboratory testing is usually performed is seldom long enough to measure what would be representative of long-term strength (Scher, 1996).

Depending on the type of soil, cohesion is significantly decreased from 1/15 to 1/2 of the short-term strength, while in frozen sand, the friction angle is typically independent of the time of load action (Tsytovich, 1975; Vyalov, 1986). The short-term friction angle and cohesion for the lobe soil (Table 2) initially used in this analysis resulted in a factor of safety greater than 1.00.

Table 2. Material properties used in the slope stability modeling.

Material

Unit Weight (kN/m3)

Cohesion (kPa)

Internal Friction Angle (u)

Silty sand with gravel (SM) Chloritic schist*

21.2 26.5

120 8,500

13 25

*Indicates material properties taken from Gonzalez de Vallejo and Ferrer (2011).

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 259–277

269

Simpson, Darrow, Huang, Daanen, and Hubbard

Figure 9. Results from direct shear testing of two frozen samples taken from FDL-A: (a, c) shear stress and horizontal displacement for samples 12-2515 and 12-2521, respectively, and (b, d) shear stress and normal stress for samples 12-2515 and 12-2521, respectively. Resulting friction angles and cohesion values, and corresponding volumetric moisture contents, are listed in Table 1.

In landslide studies, the long-term soil strength properties are often determined through a back analysis due to difficulties encountered with laboratory tests. In such an analysis, adjustments are made to the strength properties, and when the factor of safety is equal to 1.00, the conditions represent a reasonable model of the failure (Duncan, 1996). To conduct this type of evaluation on FDL-A, several sensitivity anal� yses were performed. We varied the friction angle over the range of values determined from the direct shear testing (10u to 16u) and assumed that it did not vary more than this due to the sandy nature of the soil.

Using this range of friction angles, we determined that the cohesion values resulting in a factor of safety of 1.00 ranged from 43 to 53 kPa, which are between 1/2 and 1/3 of the short-term strength and lie within the ranges reported by Tsytovich (1975) and Vyalov (1986) for long-term strength. Even with the higher friction angle values, any decrease in cohesion significantly decreased the factor of safety.

270

The SLOPE/W software includes several methods to establish the location of the shear zone. For this analysis, a fully specified shear zone was projected in the model and then optimized by the program. The optimized shear zone demonstrated a failure surface ranging from 0.1 to 1.0 m above the bedrock surface in the upper part of the lobe. In the region of the lobe where the shear zone is known, the optimized results indicated a failure surface traveling through this known point. Figure 10 is an overview of the geometry used for the analysis, produced from screen shots of the modeling output. The insets detail the optimized shear zone location at the toe, at the location of our boring, and in the upper portion of the lobe. Along the profile from the head to the toe, the optimized shear zone was located deeper in the lobe where the lobe itself was thin, and it was located farther away from the bedrock surface where the lobe was thick. The shear zone exited at the toe

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 259–277

Movement and Characteristics of a Frozen Debris Lobe

Figure 10. ProďŹ le of FDL-A used in slope stability analysis. The bottom-right inset shows where the optimized failure surface travels through the known shear zone location. The upper-left inset shows the failure surface exiting at the ground surface from the toe. The upper-right inset shows the failure surface in a thinner area of the lobe. In each inset, the green area indicates how the lobe material is divided into slices for the limit equilibrium analysis, and the bottom of these slices indicates the location of the critical slip surface. The blue region in each inset represents the water pressure used in the analysis.

more than 3 m above the bedrock (see upper left inset in Figure 10).

To analyze the effects of pore-water pressure on the factor of safety, we performed a sensitivity analysis on the pore-water pressure ratio, Ru. This value is the ratio of the pore-water pressure to the unit weight of the soil multiplied by the height of the soil column.

This method of defining pore-water pressure is not recommended for slope stability analysis unless it is for a very simple, isolated case (GEO-SLOPE International Ltd., 2008). While the complicated nature of the lobe and unknown pore-water pressure throughout made the use of the Ru difficult for rigorous analysis, we assumed an average Ru value for the whole lobe to investigate how changes in the pore-water pressure affected the factor of safety. This analysis was performed in three parts, holding two parameters constant at their average values and spanning the third over its full range. This was done using the Ru, friction angle, and cohesion values at ranges of 0 to 1, 10u to 16u, and 33 kPa to 63 kPa, respectively. While a Ru value of 0.5 typically indicates fully saturated conditions, we chose the upper value of 1 for this analysis due to the high pore-water pressure found within

FDL-A. Figure 11 is a graph of the sensitivity analysis results comparing the friction angle, cohesion, and Ru. The factor of safety is less sensitive to the friction angle than to cohesion, and it is affected most by variations in the pore-water pressure. These higher sensitivities to cohesion and pore-water pressure are significant factors if the movement of FDL-A is temperaturedependent. Considering the areas of massive ice observed during 2014, increases in temperature that cause melting of the ice will increase pore-water pressure and potentially decrease the cohesive strength of the soil, both of which will lower the factor of safety significantly.

GIS Analysis of FDL-A and Catchment We first delineated the boundaries of FDL-A and its catchment using the LiDAR imagery and observations made in the field (Hubbard et al., 2013). We developed a triangulated irregular network (TIN) using the LiDAR bare earth DEM. This was utilized to find the surface area of the catchment and FDL-A, which were 926,017 m2 and 306,687 m2, respectively. For the catchment of FDL-A, we investigated the aspect,

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 259–277

271

Simpson, Darrow, Huang, Daanen, and Hubbard

Figure 11. Sensitivity analysis results for friction angle, cohesion, and Ru coefficient.

slope, and geology. Aspect and slope rasters were developed from the bare earth DEM, while feature classes based on field observations were used to delin‐ eate geology.

Tables 3 and 4 contain summaries of the calculated percentages of the slope angle and aspect, and coverage of geological units within the catchment, respectively. The majority of the catchment has slope angles between 20u and 40u, with most of the catchment having a southwest to northwest aspect. Based on the geologic map analysis, the most common rock types found in the catchment are chloritic schist and phyllite, which correspond with the type of rock debris that we observed on the lobe itself. The unit Єgms has the highest slope angles, but it makes up a very small percentage of the catchment. The majority of the area underlain by the two remaining geological units has slope angles between 10u and 40u.

272

Vegetation coverage in the catchment was determined by subtracting the bare earth from the highest hit elevations, resulting in the vegetation height. In the field, we observed that any significant vegetation on FDL-A and in its catchment was taller than 1 m; thus, brush shorter than 1 m was eliminated in this analysis (Figure 12). The vegetation observed in the catchment greater than 1 m is mainly spruce, with some alder and willow. The upper lobe is vegetated mainly by alder and willow between 2 and 4 m high, with scattered spruce. Towards the toe, the spruce trees become more abundant and reach heights up to 20 m. The percentage of vegetation coverage in the FDL-A catchment was found using the Block Statistics tool in ArcGIS. Most of FDL-A’s catchment has little or no vegetation, as seen in Figure 12a. The areas with the most vegetation are on west-, south-, and southwest-facing slopes that are between 10u and

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 259–277

Movement and Characteristics of a Frozen Debris Lobe Table 3. Percentages of slope angle (for lobe and catchment) and aspect (for catchment only) for FDL-A. Lobe Slope (u)

Catchment

Percentage

Slope (u)

Percentage

Aspect

Percentage

13.4 53.2 27.6 5.4 0.4 — — —

0–10 10–20 20–30 30–40 40–50 50–60 60–70 70–80

0.8 16.1 50.1 26.7 3.7 1.9 0.7 0.0

North Northeast East Southeast South Southwest West Northwest

8.2 0.2 0.1 0.3 13.1 32.3 17.6 28.2

0–10 10–20 20–30 30–40 40–50 — — —

30u. The steeper-sloping areas and those facing east and north support little vegetation, if any. This analysis indicates that the very steep area on the south part of the catchment supports vegetation; however, this is most likely an anomaly in the LiDAR data due to the very high slope angles, as field observations indicate that this area is sparsely vegetated. A vegetation analysis was done for the lobe itself using the same methods. Many of the areas with low vegetation coverage on the lobe can be associated with zones of observed movement and instability. Examples include the two RTS areas (see the yellow arrows in Figure 12b), and the large scarp near the toe on the left flank of the lobe (see the black arrow in Figure 12b).

A slope analysis was performed on the surface of FDL-A, and the results are presented in Figure 13 and Table 3. The majority of the lobe surface has a slope between 10u and 20u, with the steepest slopes (i.e., between 30u and 60u) occurring along the right and left flanks, around the toe, and within some of the longitudinal cracks. A careful visual study of Figure 13 suggests that steeper areas on the lobe surface Table 4. Geological unit percentages for FDL-A’s catchment; descriptions are from Dillon et al. (1988), Mull and Adams (1989), and Spangler and Hubbard (in review).

Geological Unit Єgms Dcsp

SЄpm

Description

Percentage

Interbedded, platy to massive, graybrown to green graywacke, metasandstone and meta-siltstone Predominately interbedded, platy, purple and green-blue chloritic schist and phyllite; subordinate black to gray phyllite Predominately interbedded, platy, brown to gray-black phyllite, metasiltstone, meta-sandstone, and graywacke; subordinate shale

5.2

75.9

18.9

form risers of lobate-shaped “steps,” which we also observed in the field. These smaller surface lobes may be the result of the differential rate of near-surface movement observed in the M-IPI device data.

We also mapped several discrete surface features, including scarps, longitudinal cracks, and creeks on the lobe surface. Several techniques, such as surface roughness calculations, and analysis of multiple aspect and hillshade images at varying azimuths, were used in an attempt to define the cracks located along the length of the lobe. None of these techniques was successful, however, due to the highly variable surface of the lobe. Instead, we developed cross sections of the lobe at 10 m spacing, from which we identified the locations of substantial cracks (i.e., greater than 2 m wide). These crack locations then were plotted and connected on the LiDAR image. Figure 3a displays the center lines of the cracks, while outlines of the cracks are presented in Figure 3b to illustrate the amount of separation. A similar approach was used to map scarps by creating profiles along the length of the lobe. We verified the locations of several of these cracks and scarps in the field during the summer of 2013, and we also mapped additional scarps.

As mentioned previously, the creek draining the surface of FDL-A changed its channel location in recent years. A flow accumulation was performed using the LiDAR DEM to map the original channel location, which drained to the west. In 2013, we took handheld GPS coordinates along the new channel draining off the left flank of FDL-A to the southwest (Figure 3a).

To estimate the amount of material that could contact the highway, we performed volume per unit length calculations by creating a cross section along the widest part of the lobe (see A-A9 in Figure 4), thus yielding the most conservative volume estimate. The estimated volume in this 1-m-deep cross section is 4,454 m3 of material. Using the average measured moist unit weight of 21.2 kN/m3, this is equivalent to 94,417 kN or approximately 10,613 U.S. short tons.

With the average rate of movement at 1.2 cm per day, an estimated 46,872 tons of material per year will come into contact with the highway based on this cross-section location.

DISCUSSION The U.S. National Climate Assessment (Chapin et al., 2014) reported that over the last 60 years, Alaska has warmed at more than twice the rate of the rest of the United States. Considering the temperature dependence of permafrost, impacts of this change have been evident throughout the region. There is a measured increase in permafrost temperatures, which are projected to continue to rise (Smith et al., 2010),

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 259–277

273

Simpson, Darrow, Huang, Daanen, and Hubbard

Figure 12. FDL-A catchment and lobe vegetation analysis results: (a) percentage of vegetation coverage in the catchment, and (b) percentage of vegetation coverage of the lobe. The yellow arrows indicate the retrogressive thaw slump locations, and the black arrow indicates the large scarp and actively eroding area on the left flank. (Base map is from Hubbard et al. [2011] and GINA [2001].)

and models forecast that permafrost in Alaska will continue to thaw (Jafarov et al., 2012).

Our ongoing observations and analyses have shown that FDL-A is a dynamic feature. Both surface and subsurface measurements indicate that FDL-A’s movement is temperature dependent, and with current changes in Alaska’s climate, this has significant implications for the nearby infrastructure. The slope stabil‐ ity analysis indicated a sensitivity of the lobe’s factor of safety on cohesion and pore-water pressure. As with any slope stability issue, water is an essential factor that cannot be ignored. This is especially true for FDL-A’s current situation. The melting of massive ice (such as that exposed in RTSs) and thawing of frozen soil will increase pore-water pressure and lower the soil’s shear strength, resulting in the acceleration of FDL-A towards the Dalton Highway. While the results were instructive, our modeling efforts were constrained by lack of subsurface data. The traditional slope stability modeling also does not account for the soil strength’s temperature dependency or phase change.

Although costly, additional drilling and sampling will answer important questions, such as the location of the bedrock surface beneath the lobe, lobe thickness and depth to the shear zone, and the distribution of

274

water pressure in other locations. We received a wealth of information from the M-IPI device, even after the device sheared. Thus, we recommend the investment in another M-IPI installation, along with VW piezometers, to record subsurface movement and porewater pressure, respectively. Such a boring should be back filled with a specialized grout containing an antifreeze component so that the piezometer readings are not affected by sub-freezing temperatures. Drilling should be conducted jointly with geophysical surveys.

Geophysical methods cover greater cross-sectional area than individual borings. The recommended meth‐ od is induced polarization tomography. This method has the best potential to produce cross sections of FDL conditions deep enough to delineate the FDLA shear zone and to identify zones of high pore-water pressure within FDL-A. While the main purpose of the GIS analysis was to develop a tool for examining other FDLs along the Dalton Highway corridor, it also provided some insight into movement and areas of instability on FDL-A. Through the slope analysis, we identified differential movement processes indicated by lobate steplike features on the lobe surface, recognizing one of the several movement processes that occur on FDL-A. Areas of low vegetation cover correlated well with

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 259–277

Movement and Characteristics of a Frozen Debris Lobe

Figure 13. Results of FDL-A lobe slope analysis. Inset shows detail of lobate-shaped “steps” on FDL-A’s surface. (Base map is from Hubbard et al. [2011].)

field observations indicating highly active and unstable portions of the lobe. When additional highresolution imagery becomes available, this will be a valuable tool in recognizing surges of movement and active areas on other FDLs. A temporal analysis also can be performed to identify major changes in vegetation cover and actively moving areas on FDLs. A statistical analysis of the catchment characteristics using GIS will be useful to determine similarities among FDLs. These characteristics include the catchment size, slope angle, aspect, rock type, and vegetation coverage. We also anticipate that such analysis will reveal common conditions contributing to lobe formation, as well as significant differences that explain the absence of lobes in adjacent catchments, resulting in a comprehensive hazard analysis for this area.

CONCLUSION FDL-A is an active frozen debris lobe in the southcentral Brooks Range, Alaska. Its current location, just under 40 m upslope from the Dalton Highway, and increasing rate of movement pose a risk to infrastructure as debris may obstruct the highway and access to the TAPS. FDL-A is mainly composed of silty sand and gravel, supplied to the lobe through solifluction and rock fall, as well as massive ice present in cracks open at the surface. Measurements indicate that FDL-A exhib‐ its different mechanisms of movement, with the major‐ ity of downslope movement occurring in a shear zone located 20.6 to 22.8 m bgs where measured. Another mechanism of movement is internal flow that is highly dependent on temperature. Surface measurements taken

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 259–277

275

Simpson, Darrow, Huang, Daanen, and Hubbard

from October 2012 to August 2015 indicate an average rate of movement of 1.2 cm per day, which is an increase over historic rates and has the potential to increase further with climate projections pointing towards increasing temperatures in Alaska.

The slope stability analysis performed on FDL-A required a back analysis to determine the soil strength parameters at failure, which resulted in significantly lower cohesion values, ranging from 43 to 53 kPa while maintaining friction angle values between 10u and 16u.

When using these lower strength values, the analysis indicated a failure surface in a similar location as to what we observed from our subsurface investigation. This analysis also indicated that the stability of FDL-A is highly sensitive to cohesion and pore-water pressure.

As these features are found throughout the Dalton Highway corridor between MP 194 and MP 230, we developed a GIS protocol for the future analysis and comparison of multiple FDLs for an overall hazards assessment. The protocol includes examining catchment and lobe features such as: rock type, slope, aspect, and vegetation coverage.

This multi-faceted research has both increased our understanding of FDL-A and identified shortcomings in the available data sets and analysis techniques. A more in-depth slope stability analysis requires the development of a new modeling approach that incorporates the soil strength’s temperature dependency and phase change. Additional drilling and sampling and/or employing geophysical methods are essential for future modeling efforts. These additional field studies will help to define the pore-water pressure distribution, locations of massive ice bodies, and multiple shear zones present within the lobe. Additional GIS analysis with high-resolution data sets will help to determine historic rates of other FDLs identified for further study, thus providing a more comprehensive hazards analysis. Finally, we recommend continued surface movement measurements and field observations to document the dynamics of these mass movement features on permafrost-affected slopes.

ACKNOWLEDGMENTS This research was funded by grants from the U.S. Department of Transportation (DTRT06-G-0011), the Alaska Department of Transportation and Public Facilities (T2-12-17), and through generous support from the Alaska Division of Geological & Geophysical Surveys’ Capital Improvements Project, and the Alyeska Pipeline Service Company. We wish to thank N. Gyswyt, T. Howe, and Drs. Stuefer, Kanevskiy, and Shur, for their help and advice. We also thank our three reviewers for their constructive and helpful comments.

276

REFERENCES ALASKA EARTHQUAKE CENTER, 2013, AEIC Earthquake Database: Electronic data, available at http://www.aeic.alaska.edu/ html_docs/db2catalog.html ASTM, 1990a, D2216-90 Standard Test Methods for Laboratory Determination of Water (Moisture) Content of Soil and Rock by Mass: ASTM International, West Conshohocken, PA, 7 p.

ASTM, 1990b, D3080-90 Standard Test Method for Direct Shear Test of Soils under Consolidated Drained Conditions: ASTM International, West Conshohocken, PA, 9 p.

BARBOUX, C.; DELALOYE, R.; AND LAMBIEL, C., 2014, Inventorying slope movements in an alpine environment using DInSAR: Earth Surface Processes and Landforms, Vol. 39, pp.

2057–2099.

BROWN, J. AND KREIG, R. A., 1983, Guidebook to Permafrost and Related Features along the Elliott and Dalton Highways, Fox to Prudhoe Bay, Alaska: Alaska Division of Geological and Geophysical Surveys Guidebook 4, 230 p.

CHAPIN, F. S., III; TRAINOR, S. F.; COCHRAN, P.; HUNTINGTON, H.; MARKON, C.; MCCAMMON, M.; MCGUIRE, A. D.; AND SERREZE, M., 2014, Alaska. Climate change impacts in the United States. In Melillo, J. M.; Richmond, T. C.; and Yohe, G. W. (Editors), The Third National Climate Assessment: U.S. Global Change Research Program, Washington, D.C., pp.

514–536.

CHEN, Z.; ZHANG, J.; HO, K.; WU, F.; AND ZHONG, K. (Editors) 2008, Landslides and Engineered Slopes, From the Past to the Future: Proceedings of the 10th International Symposium on Landslides and Engineered Slopes, 30 June–4 July 2008, Xian, China, 1850 p.

DAANEN, R. P.; GROSSE, G.; DARROW, M. M.; HAMILTON, T. D.; AND JONES, B. M., 2012, Rapid movement of frozen debrislobes: Implications for permafrost degradation and slope instability in the south-central Brooks Range, Alaska: Natural Hazards and Earth System Science: Vol. 12, pp. 1521–1537.

DARROW, M. M.; DAANEN, R. P.; AND SIMPSON, J. M., 2013, Analysis of a frozen debris lobe: A first look inside an impending geohazard. In Zufelt, J. E. (Editor), ISCORD 2013: 10th International Symposium on Cold Regions Development: pp.

139–148.

DARROW, M. M.; SIMPSON, J. M.; DAANEN, R. P.; AND HUBBARD, T., 2015, Characterizing a frozen debris lobe, Dalton Highway, Alaska. In ASCE Cold Regions Engineering Conference (in press).

DAVIS, N., 2001, Permafrost: A Guide to Frozen Ground in Transition: University of Alaska Press, Fairbanks, AK, 351 p.

DILLON, J. T.; HARRIS, A. G.; DUTRO, J. T., JR.; SOLIE, D. N.; BLUM, J. D.; JONES, D. L.; AND HOWELL, D. G., 1988, Prelim‐ inary Geologic Map and Section of the Chandalar D-6 and Parts of the Chandalar C-6 and Wiseman C-1 and D-1 Quadrangles, Alaska: Alaska Division of Geological and Geophysical Surveys Report of Investigation 88-5, scale 1:63,360.

DISMUKE, J., 2002, Seismic Slope Stability and Analysis of the Upper San Fernando Dam: The University of California at Davis ECI 284, 14 p.

DUNCAN, J. M., 1996, Soil slope stability analysis. In Turner, A. K. and Schuster, R. L. (Editors), Landslides: Investigation and Mitigation, Transportation Research Record 247, National Research Council, Washington D.C., pp. 337–371.

ESRI, 2015, ArcGIS World Imagery: Electronic data, available at http://goto.arcgisonline.com/maps/world_imagery GEOGRAPHIC INFORMATION NETWORK OF ALASKA (GINA), 2001, Interferometric Synthetic Aperture Radar (IfSAR): Electronic data, available at http://gina.alaska.edu/data/ifsar

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 259–277

Movement and Characteristics of a Frozen Debris Lobe GEO-SLOPE INTERNATIONAL LTD., 2008, Stability Modeling with SLOPE/W 2007; An Engineering Methodology, 4th ed.: GEO-SLOPE International Ltd., Alberta, Canada, 355 p.

GONZALEZ DE VALLEJO, L. I. AND FERRER, M., 2011, Geological Engineering: CRC Press, New York, NY, 678 p.

HAEBERLI, W. AND BENISTON, M., 1998, Climate change and its impacts on glaciers and permafrost in the Alps: Ambio, Vol. 27, No. 4, pp. 258–265.

HAEBERLI, W.; HALLET, B.; ARENSON, L.; ELCONIN, R.; HUMLUM, O.; KÄÄB, A.; KAUFMAN, V.; BRANKO, L.; MATSUOKA, N.; SPRINGMAN, S.; AND VONDER MÜHLL, D., 2006, Permafrost creep and rock glacier dynamics: Permafrost and Periglacial Processes, Vol. 17, pp. 189–214.

HAMILTON, T. D., 1978, Surficial Geologic Map of the Chandalar Quadrangle, Alaska: U.S. Geological Survey Miscellaneous Field Studies Map MF-878A, scale 1:250,000.

HAMILTON, T. D., 1979, Surficial Geologic Map of the Wiseman Quadrangle, Alaska: U.S. Geological Survey Miscellaneous Field Studies Map MF-1122, scale 1:250,000.

HAMILTON, T. D., 1981, Surficial Geologic Map of the Survey Pass Quadrangle, Alaska: U.S. Geological Survey Miscellaneous Field Studies Map MF-1320, scale 1:250,000.

HARRIS, C., 2005, Climate change, mountain permafrost degradation and geotechnical hazard. In Huber, U. M.; Bugmann, H. K. M.; and Reasoner, M. A. (Editors), Global Change and Mountain Regions: Springer, pp. 215–224.

HARRIS, C.; AARONSON, L. U.; CHRISTIANSEN, H. H.; ETZEMÜLLER, B.; FRAUENFELDER, R.; GRUBER, S.; HAEBERLI, W.; HAUCK, C.; HÖLZLE, M.; HUMLUM, O.; ISAKSEN, K.; KÄÄB, A.; KERNLÜTSCHG, M. A.; LEHNING, M.; MATSUOKA, N.; MURTON, J. B.; NÖTZLI, J.; PHILLIPS, M.; ROSS, N.; SEPPÄLÄ, M.; SPRINGMAN, S. M.; AND VONDER MÜHLL, D., 2009, Permafrost and climate in Europe: Monitoring and modeling thermal, geomorphological and geotechnical responses: Earth-Science Reviews, Vol. 92, pp. 117–171.

HARRIS, C.; DAVIES, M. C.; AND ETZELMÜLLER, B., 2001, The assessment of potential geotechnical hazards associated with mountain permafrost in a warming global climate: Permafrost and Periglacial Processes, Vol. 12, pp. 145–156.

HARRIS, C. AND LEWKOWICZ, A. G., 2000, An analysis of the stabil‐ ity of thawing slopes, Ellesmere Island, Nunavut, Canada: Canadian Geotechnical Journal, Vol. 37, pp. 449–462.

HARRIS, C. AND VONDER MÜHLL, D., 2001, Permafrost and climate in Europe. Climate change, mountain permafrost degradation and geotechnical hazard. In Visconti, G.; Beniston, M.; Iannorelli, E. D.; and Barba, D. (Editors), Global Change and Protected Areas: Springer, pp. 71–82.

HUBBARD, T. D.; KOEHLER, R. D.; AND COMBELLICK, R. A., 2011, High-Resolution LiDAR Data for Alaska Infrastructure Corridors: Lidar Datasets of Alaska, Alaska Division of Geological and Geophysical Surveys Raw Data File 2011-3, 291 p.

HUBBARD, T. D.; SPANGLER, E. R.; DAANEN, R. P.; DARROW, M. M.; SIMPSON, J. M.; AND SOUTHERLAND, L. E., 2013, Frozen debris lobes—Characterizing a potential geologic hazard along the Dalton Highway, southern Brooks Range, Alaska (poster): Geological Society of America Abstracts with Programs, Vol. 45, No. 7, p. 152, Paper No. 15.

HUSCROFT, C. A.; LIPOVSKY, P.; AND BOND, J. D., 2003, Permafrost and landslide activity: Case studies from southwestern Yukon Territory. In Emond, D. S. and Lewis, L. L. (Editors), Yukon Exploration and Geology: Yukon Geological Survey, Whitehorse, Canada, pp. 107–119.

JAFAROV, E. E.; MARCHENKO, S. S.; AND ROMANOVSKY, V. E., 2012, Numerical modeling of permafrost dynamics in Alaska using a high spatial resolution dataset: The Cryosphere, Vol. 6, pp.

613–624.

KÄÄB, A.; REYNOLDS, J. M.; AND HAEBERLI, W., 2005, Glacier and permafrost hazards in high mountains. In Huber, U. M.; Bugmann, H. K. M.; and Reasoner, M. A. (Editors), Global Change and Mountain Regions: Springer, pp. 225–234.

KREIG, R. A. AND REGER, R. D., 1982, Air-Photo Analysis and Summary of Landform Soil Properties along the Route of the Trans-Alaska Pipeline System: Alaska Division of Geological and Geophysical Surveys, Anchorage, AK. LIPOVSKY, P. AND HUSCROFT, C. A., 2006, A reconnaissance inventory of permafrost-related landslides in the Pelly River watershed, central Yukon. In Emond, D. S.; Lewis, L. L.; and Weston, L. H. (Editors), Yukon Exploration and Geology: Yukon Geological Survey, Whitehorse, Canada, pp. 181–195.

LYLE, R. AND HUTCHINSON, D. J., 2006, Influence of degrading permafrost on landsliding processes: Little Salmon Lake, Yukon Territory, Canada. In Geohazards; Engineering Conferences International, 11 p.

MATSUOKA, N., 2001, Solifluction rates, processes and landforms: A global review: Earth-Science Reviews, Vol. 55, pp. 107–134.

MATSUOKA, N., 2014, Combining time-lapse photography and multisensor monitoring to understand frost creep dynamics in the Japanese Alps: Permafrost and Periglacial Processes, Vol. 25, pp. 94–106.

MULL, C. G. AND ADAMS, K. E. (Editors), 1989, Bedrock Geology of the Eastern Koyukuk Basin, Central Brooks Range, and EastCentral Arctic Slope along the Dalton Highway, Yukon River to Prudhoe Bay, Alaska: Alaska Division of Geological and Geophysical Surveys Guidebook 7, Vol. 1, 309 p.

PIHLAINEN, J. A. AND JOHNSTON, G. H., 1963, Guide to a Field Description of Permafrost for Engineering Purposes: National Research Council of Canada Technical Memorandum 79, 21 p.

SCHER, R. L., 1996, Geotechnical considerations. In Smith, D. W. (Editor), Cold Regions Utilities Monograph, 3rd ed.: American Society of Civil Engineers, pp. 3.1–3.115.

SMITH, S. L.; ROMANOVSKY, V. E.; LEWKOWICZ, A. G.; BURN, C. R.; ALLARD, M.; CLOW, G. D.; YOSHIKAWA, K.; AND THROOP, J., 2010, Thermal state of permafrost in North America: A contribution to the International Polar Year: Permafrost and Periglacial Processes, Vol. 21, pp. 117–135.

SPANGLER, E. R. AND HUBBARD, T. D., in review, Geologic and Geomorphic Characterization of Frozen Debris Lobe Source Basins along the Dalton Highway, Southern Brooks Range, Alaska: Alaska Division of Geological and Geophysical Surveys.

SPANGLER, E.; HUBBARD, T. D.; DAANEN, R. P.; DARROW, M. M.; SIMPSON, J. M.; AND SOUTHERLAND, L., 2013, Geological and geomorphic characterization of frozen debris lobe catchments along the Dalton Highway, southern Brooks Range, Alaska (poster): Geological Society of America Abstracts with Programs, Vol. 45, No. 7, p. 152, Paper No. 16.

TART, R. G., JR., 1996, Permafrost. In Turner, A. K. and Schuster, R. L. (Editors), Landslides: Investigation and Mitigation: Transportation Research Record 247, National Research Council, Washington, D.C., pp. 620–643.

TSYTOVICH, N. A., 1975, The Mechanics of Frozen Ground: Scripta Book Company, Washington D.C., 426 p.

VYALOV, S. S., 1986, Rheological Fundamentals of Soil Mechanics: Development in Geotechnical Engineering, Vol. 36: Elsevier Science Publishing Company, Inc., New York, NY, 564 p.

WAHRHAFTIG, C. AND COX, A., 1959, Rock glaciers in the Alaska Range: Bulletin of the Geological Society of America, Vol. 70, No. 4, pp. 383–436.

WESSON, R. L.; BOYD, O. S.; MUELLER, C. S.; BUFE, C. G.; FRANKEL, A. D.; AND PETERSEN, M. D., 2007, Revision of Time-Independent Probabilistic Seismic Hazard Maps for Alaska: U.S. Geological Survey Open-File Report 20071043, 33 p.

Environmental & Engineering Geoscience, Vol. XXII, No. 3, August 2016, pp. 259–277

277