May 2020, Volume 26, Number 2 by Association of Environmental & Engineering Geologists (AEG)

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, 201 East Main St., Suite 1405, Lexington, KY 40507. 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, 201 East Main St., Suite 1405, Lexington, KY 40507. Phone: 844-331-7867. 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 inarticles. 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.

EDITORIAL BOARD Watts, Chester “Skip” F. Radford University, Chair Hasan, Syed University of Missouri, Kansas City Nandi, Arpita East Tennessee State University

Oommen, Thomas Michigan Technological University Sasowsky, Ira D. University of Akron

ASSOCIATE EDITORS Brankman, Charles, Consultant Boston MA Bruckno, Brian Virginia Department of Transportation Clague, John J. Simon Fraser University, Canada De Graff, Jerome V. California State University, Fresno Fryar, Alan University of Kentucky Hauser, Ernest Wright State University Hutchinson, Jean Queens University, Canada Keaton, Jeff AMEC Americas Marinos, Vassillis Aristotle University of Thessaloniki, Greece

SUBSCRIPTIONS:

McBride, John Brigham Young University Mwakanyamale, Kisa Illinois State Geological Survey Santi, Paul Colorado School of Mines Dee, Seth University of Nevada, Reno Shlemon, Roy R.J. Shlemon & Associates, Inc. Stephenson, William U.S. Geological Survey Stock, Greg National Park Service Sukop, Michael Florida International University Ulusay, Resat Hacettepe University, Turkey Watts, Chester F. “Skip,” Radford University West, Terry Purdue University

Environmental & Engineering Geoscience May 2020 VOLUME XXVI, NUMBER 2

SUBMISSION OF MANUSCRIPTS

Member subscriptions: AEG members automatically receive digital access to the journal as part of their AEG membership dues. Members may order print subscriptions for $75 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.

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.

Nonmember subscriptions are $310 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.

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.

Single copies are $75.00 each. Requests for single copies should be sent to AEG, 201 East Main St., Suite 1405, Lexington, KY 40507.

To submit a manuscript go to https://www.editorialmanager.com/EEG/ default.aspx. 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”.

THIS PUBLICATION IS PRINTED ON ACID-FREE PAPER EDITORS

Brian G. Katz Environmental Consultant Tallahassee, FL 32309 eegeditorbkatz@gmail.com

Authors do not pay any charge for color figures that are essential to the manuscript. Manuscripts of fewer than 10 pages may be published as Technical Notes. For further information, you may contact Dr. Abdul Shakoor at the editorial office.

Cover photo View of the natural bridge in Natural Bridge State Park, Rockbridge County, Virginia. Photo courtesy of Brian Bruckno. See article on page 141.

Volume XXVI, Number 2, May 2020

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.

Abdul Shakoor Department of Geology Kent State University Kent, OH 44242 330-672-2968 ashakoor@kent.edu

ENVIRONMENTAL & ENGINEERING GEOSCIENCE

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, 201 East Main St., Suite 1405, Lexington, KY 40507 and additional mailing offices.

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 Volume 26, Number 2, May 2020 Table of Contents 141

Natural Bridge, Virginia: Complementary Geotechnical Investigation and Analysis Methods for Mobility Planning Brian S. Bruckno, Chester F. Watts, George Stephenson, and Christopher Mau

149

Evidence for Large Holocene Earthquakes along the Denali Fault in Southwest Yukon, Canada AndreĚ e Blais-Stevens, John J. Clague, Janice Brahney, Panya Lipovsky, Peter J. Haeussler, and Brian Menounos

167

Landslide Susceptibility and Soil Loss Estimates for Drift Creek Watershed, Lincoln County, Oregon David M. Korte and Abdul Shakoor

185

Landslide Susceptibility Mapping in the Commune of Oudka, Taounate Province, North Morocco: A Comparative Analysis of Logistic Regression, Multivariate Adaptive Regression Spline, and Artificial Neural Network Models Said Benchelha, Hasnaa Chennaoui Aoudjehane, Mustapha Hakdaoui, Rachid El Hamdouni, Hamou Mansouri, Taoufik Benchelha, Mohammed Layelmam, and Mustapha Alaoui

201

Characterization and Dynamic Analysis of the Devils Castle Rock Avalanche, Alta, Utah Patricia Pedersen, Jeffrey R. Moore, Brendon J. Quirk, Richard E. Giraud, and Greg N. McDonald

217

Microgravity Mapping of an Inception Doline Shaft System Peter J. Hutchinson, Alexander Balog, and Shad E. Hoover

227

Sisal Fiber-Polymerâ&#x20AC;&#x201C;Treated Sand Mechanical Properties in Triaxial Test Lilin Wu, Wei Qian, Jin Liu, Zezhuo Song, Debi Prasanna Kanungo, Yuxia Bai, and Fan Bu

243

Full-Scale Test and Numerical Simulation of Guided Flexible Protection System under a Blasting Load Xin Qi, Hu Xu, Zhixiang Yu, Keqin Sun, and Shichun Zhao

257

Characteristics, Controlling Factors, and Formation of Shallow Buried Karst in Eastern China: A Case Study in the Wuxi Metro Areas, Jiangsu Province Shulan Guo, Changhong Yan, Liangchen Yu, Yang Liu, Yinkang Zhou, and Xiaozhong Shi

Natural Bridge, Virginia: Complementary Geotechnical Investigation and Analysis Methods for Mobility Planning BRIAN S. BRUCKNO* Virginia Department of Transportation, 811 Commerce Road, Staunton, VA 24401

CHESTER F. WATTS GEORGE STEPHENSON CHRISTOPHER MAU Radford University, 801 East Main Street, Radford, VA 24142

Key Terms: Engineering Geology, Geophysics, Geotechnical, Instrumentation, Remote Sensing, Rock Mechanics ABSTRACT Natural Bridge, in Rockbridge County, Virginia, is a geological arch carrying U.S. Route 11 over Cedar Creek. The area has significant historical and cultural importance; it is listed on the National Register of Historic Places and is a Virginia Historic Landmark. Until 2015, the arch and area below were privately owned and operated, with only the pavement structure of U.S. Route 11 held by the Virginia Department of Transportation. Since then, the arch and area below have been leased to the Virginia Department of Conservation and Recreation, potentially transferring liability to the Commonwealth. As part of the Commonwealth’s due diligence and to help ensure that the arch is preserved for future generations, the Department of Transportation, in partnership with Radford University, completed a comprehensive, non-invasive geological and geotechnical investigation in 2017 and 2018. A complementary variety of geophysical, laser, optical, seismic, and traditional geological methods of study were used to allow for integrated data analysis. The investigation revealed potential risks to the integrity of the arch, which may eventually reduce its suitability for use as a transportation corridor. The investigation methodology allowed planning for protection of the environment, cultural resources, and local economies while avoiding any potential damage to the arch. As of the date of this article, plans are under way to relocate U.S. Route 11 onto an alternate alignment entirely, thereby helping to preserve this valuable cultural, historical, and geological asset. *Corresponding author email: brian.bruckno@vdot.virginia.gov

INTRODUCTION Natural Bridge is the common name of a natural geological arch located in Rockbridge County, Virginia. The arch carries U.S. Route 11, a primary road, over Cedar Creek. Although there are other similar features that convey transportation corridors, such as Natural Tunnel in Scott County, Virginia (Dietrich, 1970), Natural Bridge is the largest naturally occurring geological arch that carries a transportation corridor in the United States. Natural Bridge is of great historical and cultural significance. It is on the National Register of Historical Places and is a Virginia Historic Landmark (Virginia Department of Historic Resources, 2018; National Park Service, 2020). The first written accounts of Natural Bridge date to 1742. The bridge is of cultural significance to the Monacan Indian Tribe, which maintains an interpretive history and replica village at the site. The park area continues to be the location of a weekly “Sight and Sound” performance, including amplified music (although the performance itself is not affiliated with the Commonwealth.) Natural Bridge was once owned by Thomas Jefferson (Jefferson, 1787), and oral history relates that it was once surveyed by George Washington. Lack of hard evidence for the Washington claim, however, suggests that the story is apocryphal. Figures 1 and 2 represent an early portrait of the bridge and a contemporary photograph (National Gallery of Art, 2018). The view from the top of the bridge is currently blocked by a fence, but in the words of Thomas Jefferson, “Though the sides of this bridge are provided in some parts with a parapet of fixed rocks, yet few men have resolution to walk to them and look over into the abyss . . . it is impossible for the emotions arising from the sublime, to be felt what beyond what they are here, so beautiful an arch, so elevated, so light” (Jefferson, 1787).

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 141–148

141

Bruckno, Watts, Stephenson, and Mau

Figure 1. Portrait of Natural Bridge painted by William James Bennett in 1835, now hanging in the National Gallery of Art (National Park Service, 2020).

GEOLOGICAL SETTING Natural Bridge is located within the Valley and Ridge physiographic province, close to the boundary between the Blue Ridge and Valley and Ridge physiographic provinces. The region is typiﬁed by a dendritic drainage pattern in the Blue Ridge province rocks (the Precambrian Virginia Blue Ridge Complex), changing at the boundary to a trellis pattern controlled by linear ridges, in which Lower Ordovician rock units are exposed (Spencer, 1968). These rocks are nearly horizontal as the result of being in the axial part of an open, expansive syncline.

Figure 3. Regional geological map (Spencer, 1968) referenced to a Google Earth place mark.

The arch itself is believed to be the remnant of a natural tunnel formed along a fracture when the upper portion of Poague Run was captured by Cascade Creek, forming Cedar Creek. This natural tunnel followed the southeasterly regional dip of the strata. As erosion continued, the roof of the passage collapsed, leaving only the feature now known as Natural Bridge, in a section that was thicker and more resistant to erosion. Figure 3 illustrates the local geological conditions (Spencer, 1968) referenced to a Google Earth basemap (Google Earth, 2020). LOCAL SETTING

Figure 2. Contemporary photograph of Natural Bridge (Virginia Department of Historic Resources, 2018).

142

U.S. Route 11, a primary highway, runs over Natural Bridge at an oblique angle. Route 11 is one of the two major southwest-to-northeast transportation corridors in the region, the other being Interstate 81. Although Interstate 81 carries the majority of heavy commercial traﬃc in the region, accidents or other disruptions to Interstate 81—not infrequent occurrences—require traﬃc to be detoured over Natural Bridge. Between 2006 and 2018, there were 217 recorded crashes along this segment of I-81, although not all required diversion (Commonwealth of Virginia, 2018). Figure 4 shows the relationship of Natural Bridge and U.S. Route 11 in an image taken from an unmanned aerial vehicle (UAV). The arch itself is approximately 25 m long and roughly 57 m high above Cedar Creek. The thickness of the rock forming the arch varies from about 13 to 45 m, with 15 m being a common thickness

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 141–148

Natural Bridge, Virginia

Figure 4. UAV-derived aerial view of Natural Bridge and U.S. Route 11 (Watts and Stephenson, 2018).

(Spencer, 1968). The arch is shaped by widely spaced joints, which are locally visible. Northeasterly trending joints form the sides of the arch, whereas the sides of the gorge are formed by north-northwesterly trending joint sets that intersect additional joint sets trending to the east-southeast (Watts and Stephenson, 2018). U.S. Route 11 serves as a corridor for local residential and commercial traffic, serves the privately owned Natural Bridge Hotel, and also serves as a junction for Route 130, connecting the town of Glasgow and the city of Lynchburg to the greater Shenandoah Valley transportation system (Commonwealth of Virginia, 2004). HISTORICAL SETTING The ownership of Natural Bridge between the time of Thomas Jefferson and 1935 is uncertain. However, a deed dated April 18, 1935, transfers certain rights from Natural Bridge of Virginia, Inc., to the Commonwealth of Virginia. As detailed in the deed, the Commonwealth would receive and maintain the right to build and operate what would eventually become U.S. Route 11. The deed expressly limits the rights of the Commonwealth to the “surface” of the land with no right-of-way. In this manner, Route 11 was established as a corridor, and Natural Bridge was never considered an architectural bridge, was never entered into the Virginia Department of Transportation bridge inventory or the federal bridge inventory, and was never subject to any inspections of any sort. This status quo was maintained through several subsequent owners until 2015. In that year, the then current owners transferred ownership to the newly formed Virginia Conservation Legacy Fund. The agreement stipulates that once the fund pays off a $9.1 million loan, most of

the property will be deeded to the Commonwealth of Virginia (Richmond Times-Dispatch, 2014). In the meantime, the Virginia Department of Conservation and Recreation is operating Natural Bridge as a state park (WSLS, 2016), which includes a weekly privately operated sight-and-sound show, including ampliﬁed music. This created concerns regarding safety of the visitors and the traveling public and the integrity of the arch itself. These concerns were exacerbated by the recent history of Natural Bridge. On Saturday, October 23, 1999, a rockfall fatality occurred. A pickup truck–size slab of rock, formed by the intersection of joints and bedding, detached without any noted warning. The slab free-fell onto the pedestrian platform below, partially destroying an interpretive plinth and plaque, shattering in the process. The resulting shrapnel caused the death of one visitor (Watts and Gilliam, 2000). Accordingly, the then owners of the park contracted with Dr. Chester F. Watts of Radford University to provide a geological report and recommendations. Although some of the recommendations were followed (scaling and installation of rock bolts), it is unclear what other risk-limiting actions were taken. Although the investigation (Watts and Gilliam, 2000) proved invaluable for assessing Natural Bridge in light of the new ownership and potential legal positions, it was quickly determined that, due to these new conditions and the availability of new and improved technologies, a comprehensive geological and geotechnical report was required. Accordingly, a new report was commissioned through the partnership of the Virginia Department of Transportation (VDOT), Radford University, the Virginia Department of Conservation and Recreation, and the Virginia Department of Transportation Research Council (which is itself a partnership of VDOT and the University of Virginia). Due to the uncertainty regarding the structure of the arch, it was determined that only nondestructive and noninvasive methods would be employed. The decision to avoid drilling and sampling constituted a signiﬁcant obstacle. Accordingly, an array of geophysical, optical, laser, and noninvasive traditional geological surveying methods were selected for the investigation. RECENT INVESTIGATION Over the course of the fall of 2017 and the winter of 2018, a comprehensive geological and geotechnical investigation was performed. The following methods were used:

r Ground-penetrating radar (GPR)—250-, 1,000-, and 2,000-MHz antennas

r Electrical resistivity

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 141–148

143

Bruckno, Watts, Stephenson, and Mau

Figure 5. GPR (Virginia Department of Transportation, 2018). 1 mi = 1.6 km, 1 in. = 2.54 cm.

r Seismic refraction r Multichannel analysis of surface waves r Unmanned aerial systems (UAS) photography and r r r r r r

videography UAS remote sensing for discontinuity mapping GigaPan imaging Terrestrial light detection and ranging (lidar) Manual discontinuity mapping Vibration monitoring Traditional ground-based observation and measurement RESULTS

All comprehensive analyses of the various data sets were complementary: all of the methods, when compared, yielded agreeing results. No anomalies or outliers were detected, which would have required interpretation of conﬂicting results, a circumstance not always the case in geological investigations. The interpretation of the data yielded the following. Pavement Section, Soil Proﬁle, and Voids The GPR, performed by both VDOT and Radford, showed that the pavement structure is highly variable across the approximately 160-m section of road supported by Natural Bridge. The pavement section consisted of 10–20 cm of asphalt, underlain by 5–45 cm of aggregate and base course. Figure 5 illustrates interpretations of asphalt and aggregate thickness derived from the 2,000-MHz GPR antenna. (Note that, because the funding for this investigation was provided by the Commonwealth of Virginia, imperial units were required. Metric units are used in the text of this arti-

144

Figure 6. Voids at south end of Natural Bridge (Watts and Stephenson, 2018). The orange arrows indicate reﬂectors in the GPR vertical slice (top) and horizontal slice (bottom), interpreted as voids at the same depth in both analyses. 1 ft = 0.3 m.

cle, and conversions for imperial units are given in the figure captions.) Similarly, the lower-frequency antennas revealed a significantly varying depth of asphalt and aggregate as well as varying thicknesses of underlying soil. This variability is perhaps to be expected because the Valley and Ridge Physiographic Province is classic karst geography and varying soil profiles are not unusual, nor is using a varying aggregate thickness to establish a level base for roadway construction. What is perhaps more revealing is the presence of voids revealed by the 250-MHz antenna at the south end of Natural Bridge, illustrated in Figure 6. The voids were also well resolved by the electrical resistivity results. Figure 7 illustrates voids, located at both the south and the north end of Natural Bridge, revealed by the resistivity data. The resistivity revealed voids at both the north and the south end of Natural Bridge. Resistivity values suggest that the voids to the south are air filled, whereas the voids in rock to the north are sediment filled, most likely with residual clay. Although not of concern in terms of the overall stability of Natural Bridge, the voids may create constructability issues where any alternatives involving new construction may be considered. Joints and Weak Zones A well-developed joint system was also revealed by the investigation. These joints, created by regional deformation and locally generated stresses associated with arch formation and Cedar Creek throughcutting, retain some considerable strength resulting from rock cohesion or rock shear strength. The joint system revealed by the investigation appears to be typical of the regional tectonic system; that is, Natural Bridge is not situated on a separate fault block but is

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 141–148

Natural Bridge, Virginia

Figure 7. Voids Natural Bridge (Watts and Stephenson, 2018). “Hot” or darker colors indicate higher resistance, i.e., true, air-ﬁlled voids; “cold” or blue/mid-gray colors indicate lower resistance, i.e., clay-ﬁlled voids. Data were collected using a dipole-dipole array with 3-m spacing and current of up to 2,000 milliamps. Data collected at roadway centerline.

consistent with the regional geology. These joints were revealed by Radford and VDOT lidar, GigaPan imaging, and the UAS-based remote discontinuity mapping. A point-cloud image derived from lidar and a resulting stereonet are included in Figure 8. In addition to constituting planes of weakness, these joints are also conduits for water. Water, while lowering the available shear stress on the plane along which it travels, also serves as a medium for dissolved salts or other minerals, which may then precipitate, forming crystals, which can exert considerable force on these planes.

The joints were also detected by UAV point-cloud data collection and digital photogrammetry. Figure 9 illustrates a false-color image taken from UAS pointcloud data collected for this project (Watts and Stephenson, 2018). The enlarged view in Figure 9 (lower photograph) is of the upstream side of Natural Bridge, taken from the point-cloud animation depicted above. A sensitive feature, nicknamed the “Old Man” by the investigators, is circled. The Old Man is one of a number of sensitive structures identiﬁed from UAV, lidar, and manual data collection methods. The arrows point to solution openings aligned with a

Figure 8. Planes of weakness as revealed by lidar point clouds (top) and analyzed by Split-FX software (Split Engineering, 2020) (bottom). (The stereonet colors do not reﬂect any data.) (Watts and Stephenson, 2018).

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 141–148

145

Bruckno, Watts, Stephenson, and Mau

through-cutting fracture, creating a block that could fall. Point clouds such as this are georeferenced, allowing for quantitative measurements to be made, including discontinuity orientations for stability analyses. Factors of Safety

Figure 9. Planes of weakness and resulting weak masses as revealed by UAS point-cloud data. The lower image is a magniﬁcation of the area indicated in the upper image; arrows indicate planes of weakness, and the area circled has been informally named the “Old Man.” (Watts and Stephenson, 2018).

The authors used RockPack III, a kinematic data analysis package, in order to evaluate factors of safety for the sensitive features described above (RockPack, 2003). Data on the rock blocks and masses that are deﬁned by the intersection of joints and bedding planes can be collected remotely, as discussed above, or manually. Either analog compasses or applications installed on devices such as smartphones can be used for manual data collection. Regardless of the method of data collection, the data can be compiled into a single data set or subsets of data and analyzed for potential for failure and factors of safety. RockPack III uses Mohr–Coulomb shear strength parameters of cohesion and eﬀective friction angle, c and Ø . Field observations indicate that, during rainfall events, a water column may form in the high-angle cracks, thereby lowering the factor of safety. These data indicate that, under adverse conditions, the factor of safety for the Old Man structure and similar sensitive structures falls below 1.0, indicating

Figure 10. Factor of safety analysis for the Old Man structure (Watts and Stephenson, 2018). c (400 psf; note the typographical error in RockPack units) and Ø (29º) were back-calculated from prior failures, and horizontal acceleration was obtained from triaxial vibration sensors (see following section). 1 psf = 47.9 Pa.

146

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 141–148

Natural Bridge, Virginia

damage to any structure, whereas only architecturally unsound structures are likely to experience damage at a peak particle velocity of 0.25 cm/s. However, 0.2 cm/s is the recommended upper limit of vibration to which “ruins and ancient monuments” should be subjected. Although not a ruin or a monument per se, this is the closest analog for Natural Bridge. It can therefore be asserted that the normal traﬃc load is, at least, contributing to the natural forces to which the arch is exposed. Figure 11. Vibration data. The ﬁgure shows a heavy vehicle spike at a rock pinnacle approximately 35 m away from pavement (Watts and Stephenson, 2018).

DISCUSSION

the potential for failures of this and similar structures. Such isolated and singular failures do not necessarily indicate a substantial lowering of the wholesale, or global, factor of safety for Natural Bridge; however, the cumulative eﬀect of such failures will, over time, indeed lower the global factor of safety and should be evaluated within the context of Natural Bridge’s suitability as a transportation corridor.

All the data collected and analyzed for this report were highly complementary and mutually selfsupporting. Rather than being an idealized stratal stack, the outcrop is actually tectonically complex, showing voids, an irregular bedrock profile, several sets of joints, and susceptibility to human-induced vibration at outcrops relatively great from the traffic on Route 11. Specific findings and the method(s) used to reveal these findings include the following:

Vibration Analysis

r Voids at both the northern and the southern end of

Three different types of ground acceleration sensors (measuring peak particle acceleration) were deployed at Natural Bridge in various locations and produced similar results. Included are Infiltec QM4.5V-20HZ vertical seismometers, Radford University Model 03 triaxial accelerometers, and VibSensor triaxial accelerometer application (version 2.0.0) on three iOS smart phones. Examples of results from the Infiltec are presented in Figure 11. These devices are extremely sensitive to vibrations and are documented to have sensitivities on the order of thousandths of a g (Amick et al., 2013), where g represents the Earth’s gravitational pull. The most useful of these sensors thus far at Natural Bridge have been the Infiltec QM-4.5V-20HZ vertical seismometers. This arrangement of vibration sensors allowed the vibrations from traffic to be recorded at the same time by different sensors but at different locations along Natural Bridge. The data show that the pavement structure contributes very little to the attenuation of the vibrations as they propagate through the asphalt and subgrade and through the geological structure. (Analysis of the natural frequency of Natural Bridge was beyond the scope of this study.) The vibration data shown in Figure 11 that the traffic—from automobiles to heavy commercial vehicles—induce a peak particle velocity in the range of 0.05–0.25 cm/s. According to the Whiffen Vibration Criteria for Continuous Vibration (Whiffen and Leonard, 1971), 0.05 cm/s is unlikely to cause any

r r

the arch, with the southern voids appearing to be air filled and the northern voids sediment filled (revealed by both resistivity and GPR) A well-developed joint system consistent with the regional tectonics and structural geology (revealed by resistivity, lidar, unmanned aerial photogrammetry– derived point clouds, and traditional kinematic analysis) Blocks and masses, resulting from the intersection of joints and bedding, that are sensitive to natural and human-induced weathering (revealed by unmanned aerial photogrammetry–derived point clouds, and vibration analysis) Water migration along joint and bedding surfaces throughout the mass (revealed by GigaPan) Traffic and weekly sight-and-sound show vibrations that can be detected in the rock mass at relatively great distances from the vibration source (revealed by vibration analysis) Areas susceptible to rockfall (revealed by lidar, unmanned aerial photogrammetry–derived point clouds, GigaPan, and traditional kinematic analysis)

Natural Bridge is not an engineered structure, and this investigation revealed no geological evidence that the arch is unsuitable for its current use as part of a transportation corridor. In fact, the distant detection of vibratory energy suggests tight rock conditions with little transmission loss across joint planes between source and sensors. However, naturally

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 141–148

147

Bruckno, Watts, Stephenson, and Mau

occurring weathering will continue, and humaninduced weathering may accelerate the processes of erosion, which will eventually render Natural Bridge unsuitable for this purpose. As a first step in protecting Natural Bridge, human-induced weathering should be minimized by realigning U.S. Route 11 off Natural Bridge. As of 2019, several alignments are under consideration, all of which will fully remove commercial and commuter traffic from Natural Bridge, thereby helping to preserve this valuable cultural, historical, and geological asset. ACKNOWLEDGMENTS The authors would like to thank Mr. Ted Dean, CPG, and Draper Aden Associates for contributions to the geophysical data collection and interpretation. REFERENCES Amick, R. Z.; Patterson, J. A.; and Jorgensen, M. J., 2013, Sensitivity of tri-axial accelerometer within mobile consumer electronic devices: A pilot study: International Journal of Applied Science and Technology, Vol. 3, No. 2, pp. 97–100. Commonwealth of Virginia, 2004, Map of Rockbridge County, Virginia: Electronic document, available at http://www. virginiadot.org/travel/resources/county_maps/81_Rockbridge. pdf Commonwealth of Virginia, 2018, Integrator: Electronic document, available at http://integrator Dietrich, R. V., 1970, Geology and Virginia: University Press of Virginia, Charlottesville, VA. 213 p. Google Earth, 2020, Electronic document, available at https://www.google.com/earth Jefferson, T. J., 1787, Notes on the State of Virginia: Prichard and Hall, Philadelphia, PA. 382 p.

148

National Gallery of Art, 2018, Electronic document, available at https://www.nga.gov/collection/art-objectpage.66535.html National Park Service, 2020, Electronic document, available at https://www.nps.gov/subjects/nationalregister/databaseresearch.htm#table Richmond Times-Dispatch, 2014, Natural Bridge Sold, on Path to Becoming State Park: Electronic document, available at http:// www.richmond.com/news/virginia/natural-bridge-sold-onpath-to-becoming-state-park/article_f1e92060-8f61-11e3-83ef0017a43b2370.html RockPack, 2003, User’s Manual, ROCKPACK III for Windows, ROCK Slope Stability Computerized Analysis Package: C. F. Watts & Associates, Radford, VA. Spencer, E. W., 1968, Geology of the Natural Bridge, Sugarloaf Mountain, Buchanan and Arnold Valley Quadrangles, Virginia: Virginia Division of Mineral Resources Report of Investigations 13, scale 1:24,000. Split Engineering, 2020, Electronic document, available at https://www.spliteng.com Virginia Department of Historic Resources, 2018, Electronic document, available at https://www.dhr.virginia.gov/historicregisters/081-0415 Virginia Department of Transportation, 2018, Geophysical Study for Natural Bridge: Virginia Department of Transportation, Richmond, VA. Watts, C. F. and Gilliam, D. R., 2000, Engineering Geologic Evaluation of Rock Slope Stability at Natural Bridge, Virginia: Report prepared for Natural Bridge of Virginia, Inc., Natural Bridge, VA. Watts, C. F. and Stephenson, G. C., 2018, Final Report Natural Bridge Investigation: Virginia Department of Transportation, Richmond, VA. Whiffen, A. C. and Leonard, D. R., 1971, A Survey of TrafficInduced Vibrations: Road Research Laboratory, Design Division, Wokingham, U.K. 57 p. WSLS, 2016, Natural Bridge Officially Becomes a State Park, Affiliated with National Park Service: Electronic document, available at https://www.wsls.com/news/natural-bridge-officiallybecomes-a-state-park

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 141–148

Evidence for Large Holocene Earthquakes along the Denali Fault in Southwest Yukon, Canada ANDRÉE BLAIS-STEVENS* Geological Survey of Canada/Commission gèologique du Canada Natural Resources Canada/Ressources naturelles Canada 601 Booth Ottawa ON K1A 0E8

JOHN J. CLAGUE Department of Earth Sciences Simon Fraser University, 8888 University Drive Burnaby, British Columbia, Canada, V5A 1S6

JANICE BRAHNEY Watershed Sciences, Utah State University, 5210 Old Main Hill, Logan, Utah 84322

PANYA LIPOVSKY Yukon Geological Survey P.O. Box 2703 (K-14) Whitehorse, Yukon, Canada, Y1A 2C6

PETER J. HAEUSSLER US Geological Survey, 4210 University Drive, Anchorage, Alaska 99508

BRIAN MENOUNOS Geography Program and Natural Resources and Environmental Studies Institute University of Northern British Columbia, 3333 University Way Prince George, British Columbia, V2N 4Z9

Key Terms: Denali Fault, Positive Flower Structure, Late Holocene, Paleoseismic, Strike-Slip Fault ABSTRACT The Yukon–Alaska Highway corridor in southern Yukon is subject to geohazards ranging from landslides to ﬂoods and earthquakes on faults in the St. Elias Mountains and Shakwak Valley. Here we discuss the late Holocene seismic history of the Denali fault, located at the eastern front of the St. Elias Mountains and one of only a few known seismically active terrestrial faults in Canada. Holocene faulting is indicated by scarps and mounds on late Pleistocene drift and by tectonically deformed Pleistocene and Holocene sediments. Previous work on trenches excavated against the fault scarp near the Duke River reveals paleoseismic sediment disturbance dated to ∼300–1,200, 1,200–1,900, and 3,000 years ago. Re-excavation of the trenches indicates a fourth event dated to 6,000 years ago. The trenches are interpreted to show a negative ﬂower structure produced by extension of sediments by dextral

*Corresponding author email: andree.blais-stevens@canada.ca

strike-slip fault movement. Nearby Crescent Lake is ponded against the fault scarp. Sediment cores reveal four abrupt sediment and diatom changes reflecting seismic shaking at ∼1,200–1,900, 1,900–5,900, 5,900– 6,200, and 6,500–6,800 years ago. At the Duke River, the fault offsets sediments, including two White River tephra layers (∼1,900 and 1,200 years old). Late Pleistocene outwash gravel and overlying Holocene aeolian sediments show in cross section a positive flower structure indicative of post-glacial contraction of the sediments by dextral strike-slip movement. Based on the number of events reflecting ∼M6, we estimate the average recurrence of large earthquakes on the Yukon part of the Denali fault to be about 1,300 years in the past 6,500–6,800 years.

INTRODUCTION Natural Resources Canada, through its Public Safety Geoscience program, carries out geologic research to identify natural hazards in Canada. A component of this program is the study of faults on the Canadian landmass that might be seismically active and thus pose a potential hazard to life and the built

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 149–166

149

Blais-Stevens, Clague, Brahney, Lipovsky, Haeussler, and Menounos

Figure 1. Denali fault system, the tectonic plates that control it, and the study area near Kluane Lake. The Paciﬁc Plate is subducting beneath the North America Plate at a rate of ∼50 mm/yr. The Denali and Totschunda faults are located at the northeastern edge of the Wrangell subplate. The red arrow marks the intersection of the Denali and Totschunda faults and the division between the western and eastern segments of the Denali fault. The average slip rate on the Denali fault east of this intersection is less than half that to the west (Haeussler et al., 2017). Diagram modiﬁed from Fuis and Wald (2003) and Elliott et al. (2010).

environment. To date, few faults, aside from those oﬀshore of the coast of British Columbia near active plate boundaries, have been shown to be active and thus capable of producing damaging earthquakes. Recently, the Leech River fault, located in southern Vancouver Island, southwestern British Columbia, has been documented as a seismically active fault (Morell et al., 2018). Another seismically active fault is the Denali, a major intracontinental dextral strike-slip fault that has been active since the early Cretaceous. It is more than 2,000 km long, extending from northwest British Columbia to southwest Alaska (Figure 1; Grantz, 1966; Eisbacher, 1976; Lanphere, 1978; Dodds, 1995; and Lowey, 1998). The Alaska section of this fault generated a magnitude 7.9 earthquake in November 2002 and subsequently was well studied (EberhartPhillips et al., 2003; Haeussler et al., 2004, 2017; Matmon et al., 2006; and Haeussler, 2008, 2009). The Yukon section of the fault, in contrast, is less well studied (Clague, 1979; Haeussler et al., 2017) and historically has produced only a few small earthquakes.

150

Bender and Haeussler (2017), however, have recently published a map showing the surface trace of the eastern Denali fault in Alaska and Yukon. The 2002 earthquake ruptured the central section of the Denali fault, which has the highest late Quaternary average slip rate (∼13 mm/yr) west of its intersection with the Totschunda fault (Figure 1, red arrow and thick red line; Matmon et al., 2006; Hauessler et al., 2017). To the west of the 2002 epicenter, over a distance of 575 km, the average slip rate has been estimated at ∼5 mm/yr and to the east, roughly 280 km southeast of the Tochunga–Denali intersection (Figure 1), at ∼2 mm per year (Haeussler et al., 2017). Marechal et al. (2018) estimated dextral strike-slip motion reducing to less than 1 mm/yr along the eastern Denali fault, approximately 80 km south of the Denali– Totschunda junction. In this article, we report evidence for Holocene seismic activity and displacements on the eastern portion of the Denali fault, that is, the Shakwak segment, in the vicinity of Kluane Lake, Yukon (Figure 2).

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 149–166

Evidence for Large Holocene Earthquakes

Figure 2. (a) Study sites A–C (black letters). The red line is the approximate trace of the Denali fault. (b) Location of study area (red rectangle) in Yukon. (c) Locations of sites D–F (black letters) near the Duke River (Google Earth image).

We document geomorphic evidence of these displacements; describe and re-interpret paleoseismic trenches excavated by scientists from the U.S. Geological Survey, Yukon Geological Survey, and Simon Fraser University (Seitz et al., 2008; Lipovsky et al., 2009); describe and interpret sediment cores collected from a small lake impounded by the fault scarp; and, finally, describe deformation of sediments intersected by a strand of the fault in a bluff along the Duke River. We attempt to correlate inferred paleoseismic events in three different but spatially close settings subject to the limits of radiocarbon dating (sites D–F in Figure 2c). SETTING The Denali fault separates the Kluane Ranges, located within the St. Elias Mountains, from the Ruby Range in the Yukon Plateau (Figure 2; Mathews, 1986; Huscroft et al., 2004). Bedrock comprises volcanic, volcaniclastic, and sedimentary rocks ranging in age from Permian to Jurassic (Dodds and Campbell, 1992). Permafrost in the area is sporadic and discontinuous, with less than 10% ground ice to depths of 10–20

m below the ground surface (Heginbottom and Radburn, 1992; Heginbottom, 1995). At the three sites that we studied (D–F in Figure 2c), no obvious permafrost features were observed at the surface or during trenching and coring. The study area is characterized by a subarctic continental climate with long, cold winters and short, warm summers. The climate is semi-arid, with an average of 340 mm of precipitation per year. The average monthly temperatures in January are −22ºC and in July 13ºC (Huscroft et al., 2004; Northern Climate ExChange, 2013). The area has undergone several Quaternary glaciations, the last of which was the McConnell Glaciation (∼22–12.5 ka), during which glaciers filled the Shakwak Trench and flowed toward the northwest (Figure 3; Rampton, 1981; Duk-Rodkin 1999; and Huscroft et al., 2004). Two explosive volcanic eruptions occurred 1,900 and 1,200 years ago from a source near Mt. Bona and Mt. Churchill in the St. Elias Mountains in eastern Alaska (McGimsey et al., 1990; Clague et al., 1995; and Lerbekmo, 2008). The older eruption left a thin tephra layer that extends to the north along the Yukon–Alaska boundary, whereas the younger eruption produced a larger tephra layer that covers southern Yukon and areas to the east and south (McGimsey et al., 1990; Lerbekmo, 2008). Together, the two White River tephras are found over an area of about 340,000 km2 and have a volume of 25–30 km3 . Where the two lobes overlap, the tephra layers are commonly separated by a thin layer of organic sediment or loess. Typically, the older tephra is finer and thinner than the younger one (McGimsey et al., 1990; Lerbekmo, 2008). METHODS We used aerial photographs, Google Earth satellite imagery, and lidar images to document surface displacements along the Denali fault (sites A– C in Figure 2a and c). Seitz et al. (2008) excavated three trenches perpendicular to the fault (site D in Figure 2c). We revisited and deepened U.S. Geological Survey trenches T2 and T3 in 2011 and 2013. Trench 1(T1) was not re-excavated due to a high water table. Trench T3 was the largest and deepest of the three and is the focus of this study. It measured up to 7 m long and 2 m deep. Age control is provided by (1) the two late Holocene White River tephra layers, (2) non-faulted 300-year-old shorelines of Kluane Lake (Clague et al., 2006; Brahney et al., 2010) crossed by the Denali fault, and (3) AMS radiocarbon ages on plant remains in the trench sediments, in Crescent Lake cores, and on the

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 149–166

151

Blais-Stevens, Clague, Brahney, Lipovsky, Haeussler, and Menounos

Figure 3. Trace of the Denali fault mounds in the vicinity of site A. The fault cuts late Pleistocene glacier ﬂutings oriented toward the northwest (white curved arrows).

Duke River bluﬀ. The two samples at the Duke River were collected as follows: one just below the younger tephra and the other just above the older tephra. Elevations of the Duke River samples (Table 1) appear inverted, but this is due to the irregular surface at the top of the bluﬀ.

We aﬃxed strings on the vertical faces of the handexcavated trench walls to establish a grid (0.5 m2 ) to facilitate mapping of sediments and structures. We photographed the trench walls and created a georeferenced orthophoto of the wall that we used for mapping. Two samples of wood and charcoal from trench

Table 1. Radiocarbon ages. Sample No. Ages from USGS trench 3 (Seitz et al., 2008) CL-13 CL-14 CL-16 Ages from revisited USGS trench 3 UCIAMS 134801 UCIAMS 134803 Sample no. Ages from Crescent Lake sediment cores UCIAMS 83763 UCIAMS 83764 UCIAMS 83765 UCIAMS 83766 UCIAMS 79274 UCIAMS 109492 UCIAMS 109493 UCIAMS 109494 Ages from the Duke River bluﬀ UCIAMS 134799 UCIAMS 134800

Radiocarbon Age

Calibrated Age Range (yr BP)*

2,010 ± 35 2,810 ± 30 3,515 ± 40

1,880–2,054 2,844–2,999 3,650–3,895

0.25 0.4 0.5

Plant fibers/peat Plant fibers/peat Plant fibers/peat

3,240 ± 20 5,280 ± 15

3,556–3,495 5,960–6,178

1 1.5

Wood Charcoal

Radiocarbon age

Calibrated age range (yr BP)*

Core depth (cm)

Dated material

5,210 ± 150 5,820 ± 170 5,880 ± 30 5,795 ± 25 2,040 ± 15 430 ± 20 1,840 ± 15 3,735 ± 40

5,660–6,290 6,290–7,150 6,640–6,780 6,500–6,660 1,940–2,050 471–520 1,722–1,821 3,975–4,233

Core 1, 76 Core 1, 80 Core 1, 82.5 Core 1, 85 Core 2, 48 Core 4, 43 Core 4, 54 Core 4, 98

Plant fibers Plant fibers Plant fibers Plant fibers Plant fibers Plant fibers Plant fibers Plant fibers

1,460 ± 15 1,805 ± 15

1,310–1,380 1,700–1,800

0.6† 0.3†

Wood Wood

Trench Depth (m)

Dated Material

* Calibrated ages calculated with OxCal v. 4.3.2 (Bronk Ramsey, 2017) and the IntCal13 atmospheric curve (Reimer et al., 2013). Radiocarbon age is given with 1σ error multiplier, and calibrated age is given with 2σ error multiplier. † Depth from surface (m).

152

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 149–166

Evidence for Large Holocene Earthquakes

T3 were radiocarbon-dated to supplement the radiocarbon ages reported by Seitz et al. (2008). We collected four 7.5-cm-diameter sediment cores from Crescent Lake (site E in Figure 2c), a shallow pond ∼200 m southeast of the trenches (site D). Crescent Lake has an area of about 1.7 ha and a maximum depth of about 4 m. We made two attempts to core the lake. A single short (73 cm) percussion core (labeled core 2) was collected from the center of the lake in the summer of 2008. An additional three vibracores up to 106 cm in length were collected from the frozen surface of the lake in March 2010 along a transect perpendicular to the fault scarp (labeled cores 1, 3, and 4). A fibrous rooty peat that prevented core penetration in 2008 also stopped the vibracorer in 2010. We transported all cores to the University of Northern British Columbia in Prince George, British Columbia, where they were split, photographed, and logged. Radiocarbon ages on eight samples of subfossil plant remains from the cores were determined at the University of California Irvine Keck Carbon Cycle AMS Laboratory. We analyzed two of the four Crescent Lake cores (2 and 4) for diatom concentrations and species composition. Prior to the preparation of each slide, 5 mg of dry sediment were digested in 20% hydrogen peroxide solution and placed in a water bath for 2 days. The digested sediment was rinsed with de-ionized water and left to settle for at least 24 hours before being aspirated and further rinsed (Battarbee et al., 2001). Diatom slides were mounted in Battarbee trays to allow for counts of microfossils per gram of sediment. Diatom species were identified from taxonomic literature (Patrick and Reimer, 1966, 1975; Krammer and Lange-Bertalot, 1985, 1986, 1988, 1991, 2000; and Kelly et al., 2005) and Internet resources (Spaulding et al., 2010). Diatoms were identified and counted using an Olympus BX51 Photomicroscope with DIC optics. We described and photographed deformed late Pleistocene and Holocene sediments in a bluff on the north side of the Duke River just upstream of the Alaska Highway (site F in Figure 2c). RESULTS AND DISCUSSION Geomorphic Expression of the Fault Clague (1979, 1982) identified a nearly continuous Holocene offset along the Denali fault from near the Alaska–Yukon boundary to the southern end of Kluane Lake (Figure 2a). The fault displaces drumlins and flutings produced by northwest-flowing glacier ice during the McConnell Glaciation (Marine Isotope Stage 2; Rampton, 1979, 1981; Klassen, 1987;

Figure 4. Google Earth image showing inferred oﬀsets of mounds along the Denali fault at site B. Seitz et al. (2008) documented oﬀsets at this site, and Haeussler et al. (2017) estimated the total displacements to be 20–30 m.

and Duk-Rodkin, 1999; site A in Figure 2a; see also Figure 3). The fault trace is locally marked by elongate mounds that were interpreted by Seitz et al. (2008) and Hauessler et al. (2017) to be tectonic push-ups formed by shortening between en échelon left-stepping fault strands (Figure 4). Here, we interpret the mounds as positive flower structures and the depressions as negative flower structures that formed, respectively, at sites of local compression and extension between en échelon fault strands during strike-slip movement (Harding, 1985; Woodcock and Fischer, 1986; see the section “Paleoseismic Evidence in Trenches”). We identified and measured 538 mounds over a distance of about 110 km along the length of the fault in Yukon. The mounds average 90 m in length and 60 m in width and are up to 10 m in height; the average spacing between mounds is about 125 m. Very few fault mounds show measurable lateral displacement. However, at site B in Figure 2a, mounds are offset by 20–30 m (Figure 4; Seitz et al., 2008; Haeussler et al., 2017). Examination of aerial photographs (1980s National Air Photo Library) and Google Earth imagery (Digital Globe 2016) shows that the mounds and depressions lie within a fault zone, the strands of which penetrate both bedrock and overlying late Pleistocene and Holocene sediments. Fault displacement is also visible on a lidar hill shade image of the floodplain of the Duke River just southwest of the Alaska Highway (Figure 5a). There it extends across a low terrace less than 1 m above the active braid plain of the river, a surface that is likely less than 1,000 years old (Clague et al., 2006). Marechal

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 149–166

153

Blais-Stevens, Clague, Brahney, Lipovsky, Haeussler, and Menounos

over the Holocene of 2 mm/yr. Haeussler et al. (2017) provide a range of 1.5–2.3 mm/yr for the Yukon segment of the Denali fault but note that the average slip rate for the Holocene is probably closer to the upper end of this range because the mounds may have started to form well after deglaciation. Our ﬁndings support these previously reported slip rate estimates. Paleoseismic Evidence in Trenches

Figure 5. (a) Lidar image of a fault-oﬀset ﬂuvial terrace at the Duke River just upstream of the Alaska Highway. Also shown are the locations of sites D–F (see also Figure 2c for locations). (b) Close-up of lidar image showing the locations of the three trenches. Source: Yukon Geological Survey.

et al. (2018) documented a line of fault mounds crossing the ﬂoodplain of Koidern Creek, ∼70 km northwest of the Duke River. The southernmost morphological expression of the Denali fault in our study area is a series of mounds truncating the surface of a post-glacial alluvial fan south of Kluane Lake (site C in Figure 2a; see also Figure 6). All fault displacements east of the Alaska–Yukon border post-date terminal Pleistocene deglaciation of the Kluane Lake area, which has been dated at about 12,500 years ago (Rampton, 1981). This date and oﬀsets of mounds yield an average horizontal slip rate

Two of the three trenches (T2 and T3) at site D (Figures 2c, 5a and b, and 7a and b) record three large earthquakes in the past 3,000 years (Seitz et al., 2008). To recapitulate, T2 and T3 show evidence of co-seismic deformation, including faulted and folded strata forming the fill within the depression adjacent to the fault scarp (see Appendix 1). The evidence includes the two deformed White River tephra layers, fissure fills, upward truncations on inferred faults, stratigraphic mismatches, and colluvial wedges that likely formed soon after three large earthquakes (Figure 7a and b). The green and blue lines are discontinuous and faulted horizons, and the red lines are faults and/or fissures (Figure 7a). The orange line in Figure 7a was labeled free face (i.e., slope profile) or fault. In Figure 7b, it is interpreted as a fault. We recognize a fourth potential seismic event in faulted sediments exposed when we deepened trench T3 (event 4, black fault line in Figure 7b; see also Table 1). A fourth event was also documented in T2 (Seitz et al., 2008). Further details on all the paleoseismic features in the trenches are provided by Seitz et al. (2008) and included in Appendix 1.

Figure 6. Fault mounds on a postglacial alluvial fan just south of Kluane Lake (site C in Figure 2a). This site is the southernmost geomorphological expression of the fault.

154

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 149–166

Evidence for Large Holocene Earthquakes

Figure 7. Stratigraphy of trench T3 (site D in Figure 2c). (a) Photomosaic of the trench from Seitz et al. (2008), view to the north. Seismic events (Ev) 1–3 (∼0.3–1.2, 2.2, and 3.0 ka) are labeled; Cl-13, Cl-14, and Cl-16 (in red) are 14 C ages reported by Seitz et al. (2008) (Table 1). (b) Interpreted stratigraphy of the same trench showing a fourth seismic event and the full set of radiocarbon ages. The older White River tephra (1.9 ka) is shown as small scattered white blebs in the center of the trench, and the younger tephra (1.2 ka) is colored bright yellow. (c) The trench face is interpreted as a negative ﬂower structure with red dashed lines indicating inferred faults and arrows showing a downward direction of movement.

The older White River tephra is present as small scattered and stretched blebs, which we interpret to record co-seismic sediment deformation. The younger White River tephra layer is thickest in the center of the trench and is discontinuous and has a bleb-like structure adjacent to the scarp. It is oﬀset by a fault strand in trench T2 (Seitz et al., 2008; Appendix 1). We in-

terpret the sediments within the trenches to show extensional deformation in a negative ﬂower structure produced by dextral strike-slip movement (red dashed lines in Figure 7b; see also Figure 7c). Although we do not know how much slip occurred during each earthquake, the fact that the fault ruptured the surface suggests that the events were large,

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 149–166

155

Blais-Stevens, Clague, Brahney, Lipovsky, Haeussler, and Menounos

Figure 8. (a) Stratigraphy of Crescent Lake sediment cores collected in 2008 and 2010. Stippled black and blue lines indicate inferred correlations of units. Comparable diatom assemblages are color coded (see Appendix 2). (b) Locations of the four sediment cores; white dashed line is the Denali fault trace. Based on lithostratigraphic and biostratigraphic correlations, the upper and lower radiocarbon ages in core 4 (italicized) are thought to be too young. Refer to Appendix 2 for color-coded peaks in concentrations of speciﬁc diatom species; the color reﬂects the dominant species.

in the range of magnitude 6–7 or larger (EberhartPhillips et al., 2003; Haeussler et al., 2004; and Matmon et al., 2006). The relatively small number of events in the past several thousand years, however, is consistent with the regional tectonic model that hypothesizes much lower slip rates on the Yukon segment of the Denali fault than on the Alaska segment (Matmon et al., 2006; Haeussler et al., 2017). The most recent rupture occurred after deposition of the 1,200-year-old White River tephra but prior to the formation of the raised Kluane Lake shorelines less than 300 years ago (Clague et al., 2006). Based on the radiocarbon ages from the trenches, Seitz et al. (2008) assigned ages of 300–1,200 (event 1), ∼2,200 (event 2), and 3,000 (event 3) cal yr BP to earthquakes. A fourth inferred earthquake, mentioned above, is assigned an age of ∼5,900–6,200 cal yr BP. Uncertainties are inherent in dating paleoseismic events because of potential errors in calibrated radiocarbon ages and the stratigraphic relationship of dated sediments to earthquake events. Moreover, deformation is likely driven not only by fault slip at depth but also by gravitational movements from the fault scarp in an extensional setting, that is, a negative ﬂower structure (Figure 7c; Woodcock and Fisher, 1986). Paleoseismic Evidence from Crescent Lake Cores Basal sediments in three of the four Crescent Lake sediment cores consist of dark brown, woody, ﬁbrous

156

peat (Figure 8a and b; Table 1). In two of the cores (1 and 4), this peat is abruptly overlain by a terrigenous silt unit (blue unit in Figure 8a), which is in turn overlain by an organic-rich silt (olive grey). Gyttja (beige unit in Figure 8a) is the dominant sediment in Crescent Lake. It overlies the clastic organic-rich silt layer near the base of the cores. Subtle differences in lithology are indicated by color differences and by the amount of clay, silt, and sand in the gyttja. The two White River tephra layers are present in three of the four cores; only the younger tephra was found in core 4 (Figure 8a). Diatom data allowed us to refine stratigraphic correlations between cores 2 and 4 (color-coded dots in Figure 8). They also revealed that the upper and lower radiocarbon ages in core 4 are too young, likely due to sediment mixing, possibly by bioturbation (Figure 8a; Table 1; Appendix 2, Supplementary Figure 1). Dominant diatom species at the base of the two cores (core 2: 69 cm; core 4: 90–96 cm) are pioneering benthic species (Pseudostaurosira brevistriata, Staurosirella pinnata, and Staurosira construens), which commonly constitute the bulk of initial diatom communities in early Holocene subarctic sediments (Appendix 2, Supplementary Figures 2 and 3; Lotter et al., 2010). This assemblage is followed by a sudden rise in planktonic species that require deeper waters to achieve active lake mixing (Aulacoseirs spp.). The transition from species characteristic of wetlands (peat) to planktonic species is consistent with sudden deepening of a wetland into a lake (organic-rich silt). We

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 149–166

Evidence for Large Holocene Earthquakes

attribute the deepening to the formation of a sag pond due to subsidence along a fault scarp during an earthquake (McCalpin et al., 2009). We ruled out climatic or hydrologic factors as potential causes of sudden deepening of the wetland because there is no evidence of surface flow into the lake or thermokarst subsidence at this site. The age of this event (6,500– 6,800 cal yr BP) was estimated from the radiocarbon ages of samples collected from the top of the peat and from the overlying silt layer in core 1 (Figure 8a; Table 1). Marked increases in diatom flux, albeit with decreases in species richness, characterize several levels in the gyttja above the basal sediments in core 4 (88, 55, and 41 cm; Appendix 2, Supplementary Figure 3). These intervals likely represent intermittent disturbances of the lake system, perhaps accompanied by a sudden increase in nutrients or a shift in benthic habitat affecting diatom production; we infer that they are responses to seismic shaking and movements on the nearby fault. As we have argued for the older subsidence event, it is unlikely that climatic and hydrologic factors are responsible for these disturbances due to the cold, dry climate and the absence of streams flowing into the lake. We infer that the silt layer at 88 cm in core 4 correlates with the silt layer in core 1 (5,900–6,200 cal yr BP) and the clay layer in core 2. The silt layer at 55 cm in core 4 is not easily correlated with silt layers in the other cores due to uncertainties in its age. The silt layer between the two White River tephras in cores 1–3 and below the youngest tephra in core 4 has an age range of 1,900–1,200 years ago. Diatoms are rare in this layer in cores 2 and 4, indicating rapid sedimentation of terrigenous rather than organic material. The only diatom species that is present (Eunotia panda), albeit in low numbers, is characteristic of shallow-water or bog habitats (Veselá, 2015), indicating that it was likely transported to the lake floor. Nitzschia amphibia, a benthic species that is sensitive to turbidity and is otherwise common in cores 2 and 4, is absent in the silty layers sampled, whereas other pioneering benthic species increase in abundance in these layers (Appendix 2, Supplementary Figure 2). We infer that this silt layer is the result of seismic shaking. It may correlate with event 2 (∼2,200 cal yr BP) of Seitz et al. (2008); if so, event 2 is younger than 2,200 cal yr BP. Core 4 is closest to the fault scarp and thus most likely to record influxes of sediment from the scarp (e.g., sandy layers colored orange in Figure 8a and b). These coarse layers could also have formed during earthquakes by the partitioning and redeposition of the clastic component of gyttja within the lake, as documented in Lake Témiscaming (northern Québec)

following a magnitude 6.3 earthquake in 1935 (Doig, 1991). Paleoseismic Evidence from the Duke River Bluff Late Pleistocene sediments are exposed continuously over a distance of about 800 m along the north side of the Duke River upstream from the Alaska Highway bridge (site F in Figure 2c; see also Figures 5 and 9). Two Pleistocene outwash gravel units and an overlying blanket of aeolian sand are exposed in cliffs up to 20 m high. The upper gravel unit dates to the McConnell Glaciation. It unconformably overlies a lower gravel unit that pre-dates the McConnell Glaciation (Rampton, 1981; Duk-Rodkin, 1999). The two gravel units are locally separated by a thin unit of laminated to bedded silt (Figure 9a). Three types of paleoseismic evidence are apparent in the river bluff. First, strata in the two outwash gravel units are disturbed at the trace of the Denali fault (Figure 9a). The gravel has been uplifted and the bedding disturbed near the top of the bluff. The ground surface where the fault reaches the top of the bluff shows a positive flower structure/mound produced by upward displacement of sediment due to compression from dextral slip movement during one or more earthquakes (Figure 9a and b). Second, several pebbles and cobbles in the two gravel units within the fault zone are fractured and displaced along the fault trace (Figure 10). Third, the aeolian blanket at the top of the bluff contains the older and younger White River tephras, both of which are faulted and folded (Figure 11). It is likely that many paleo-earthquakes have produced the deformation in the Duke River bluff. At least one of them happened after 1,200 years ago because the younger White River tephra is deformed. Geomorphology and Seismicity Other researchers have documented surface displacements and the generation of positive flower structures during modern earthquakes in strike-slip tectonic settings. Mavroulis et al. (2017), for example, reported such features along the Cephalonia transform fault in western Greece, near Ateras village, after two strong (magnitudes 5.9 and 6.0) crustal earthquakes in 2014. Ulusay et al. (2002) documented fault displacements with flower structures along the North Anatolian fault in Turkey after the 1999 Kocaeli and Düzce earthquakes (magnitudes 7.4 and 7.2, respectively). We are not aware of any earthquakes smaller than the 2014 Cephalonia events that have produced surface or nearsurface deformation similar to that observed along the Yukon section of the Denali fault. We thus conclude that the displacements and deformation we describe are the product of moderate to large earthquakes.

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 149–166

157

Blais-Stevens, Clague, Brahney, Lipovsky, Haeussler, and Menounos

Figure 9. (a) Bluff exposing late Pleistocene outwash gravel on the north side of the Duke River west of the Alaska Highway (site F in Figure 2c; see also Figure 5a). The thicker white dashed line is the approximate trace of the Denali fault plane. Note the difference in the color and structure of the gravel on opposite sides of the fault. The large half arrows indicate fault movement directions. The small half arrows next to the thinner white dotted lines indicate the upward movement of the petals. The red dot indicates the location of the photos in Figure 10, and the blue dot is the location of the photo in Figure 11. (b) Block diagram showing a positive flower structure with petals along a dextral strike-slip fault (from Woodcock and Fisher, 1986).

Timing of Past Seismic Events The Crescent Lake sediment cores and the U.S. Geological Survey trenches provide evidence for at least

ﬁve large earthquakes on the Denali fault over the past 6,500–6,800 years. Seitz et al. (2008) dated the last three earthquakes to ∼300–1,000 (event 1), 2,200 (event 2), and 3,000 (event 3) years ago and linked two

Figures 10. Fractured pebbles and cobbles within the Denali fault zone in the Duke River bluﬀ (see Figure 9a for location).

158

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 149–166

Evidence for Large Holocene Earthquakes Table 2. Ages of inferred earthquakes (cal yr BP).

Figure 11. Deformation of aeolian sediments at the top of the Duke River bluﬀ (see Figure 9a for location). Sediments on the left side of the fault (red stippled line) were elevated during one or more earthquakes. Sediments to the right moved downward toward the viewer. The last earthquake responsible for this deformation happened after deposition of the younger White River tephra 1,200 years ago.

of the three events to documented prehistoric earthquakes in Alaska. Based on our deepening and reexamination of one of Seitz et al.’s (2008) trenches, we suggest that a fourth earthquake happened about 6,000 years ago (event 4) (Figures 7b and 8a). We infer that the sudden formation of Crescent Lake and subsequent deposition of silt 6,500–6,800 years ago records yet another large earthquake (event 5) and that silt and clay units and diatom changes in Crescent Lake cores (1, 2, and 4) provide evidence for an additional earthquake about 5,900–6,200 years ago (event 4). Silt beds deposited between 5,900 and 1,900 years ago may correlate with event 3 of Seitz et al. (2008), which they dated to about 3,000 years ago, although this correlation is tentative due to poor chronological constraints. We infer that an earthquake happened between 1,900 and 1,200 years ago (event 2) based on the silt unit present between the two White River tephras in three of the four cores (1–3) and underlying the younger tephra in one core (4). This event may be linked to the 2,200-year event proposed by Seitz et al. (2008), in which case their event 2 is less than 1,900 years old. Deformation of the younger White River tephra in the

Seismic Event No.

USGS Trench

Crescent Lake Sediment Cores

Duke River Bluﬀ

1 2 3 4 5

300–1,200 ∼2,200 ∼3,000 6,000–6,200 Not observed

Post-1,200 1,200–1,900 1,700–5,900 5,900–6,200 6,500–6,800

300–1,200 Not observed Not observed Not observed Not observed

trenches and the Duke River section indicates that the most recent large earthquake on the Denali fault is less than 1,200 years old (event 1). Table 2 summarizes our correlations of earthquakes inferred from the Crescent Lake sediment cores, U.S. Geological Survey trenches, and the Duke River bluff. Trench events 1–4 can be linked to events in Crescent Lake, albeit with much uncertainty in the case of event 3 due to a lack of reliable radiocarbon ages. If trench event 2 correlates with the aforementioned dated silt layer in Crescent Lake, it must be younger than 1,900 years old. Deformation associated with the most recent event (<1,200 years ago) was not seen in the Crescent Lake cores, possibly because of the narrow diameter of the cores. However, in core 4 close to the fault scarp, coarser sandy layers above the younger White River tephra may represent disturbance due to seismic shaking in the past 1,200 years. Sediment deformation in the trenches is cumulative and due to several large earthquakes but also induced gravitationally from the edge of the fault scarp. At Crescent Lake, paleoseismic activity is displayed in the sudden change from a wetland to a lake and, subsequently, by pulses (or partitioning) of clastic sediment reflected in changes in diatom assemblages in the lake. Hence, closely spaced but different geological settings reveal recurrent paleoseismic activity over the past 6,500–6,800 years. Considered together, the data suggest an average recurrence of large earthquakes on the Yukon section of the Denali fault of about 1,300 years, which is an order of magnitude lower than on the San Andreas fault in California, which has an average recurrence rate of large earthquakes every 150 years (Schulz and Wallace, 2016). CONCLUSION Several independent lines of evidence indicate that the Yukon segment of the Denali fault produced many large earthquakes during the Holocene. Geomorphic observations show that the fault has been active since the area was deglaciated about 12,500 years ago. The active fault trace cuts across glacially streamlined landforms at a low angle and is marked by rectilinear mounds, some of which show right-lateral

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 149–166

159

Blais-Stevens, Clague, Brahney, Lipovsky, Haeussler, and Menounos

displacements. The fault also offsets a late Holocene terrace bordering the modern braid plain of the Duke River near the north end of Kluane Lake and a Holocene alluvial fan south of the lake. Sediments deposited in Crescent Lake, which is impounded against the west side of a Denali fault scarp, also contain evidence of co-seismic disturbance. The lake formed about 6,500–6,800 years ago from movement along the nearby fault. Silty layers record detrital sediment inputs from the adjacent fault scarp 6,500– 6,800 and 5,900–,6200 years ago, possibly several times between about 5,900 and 1,900 years ago, and again between 1,900 and 1,200 years ago. Earthquakes inferred from sediment disturbance in nearby paleoseismic trenches date to ∼6,000, 3,000, 1,900–1,200, and 1,200–300 years ago. The sediments are interpreted as a negative flower structure resulting from extension by dextral strike-slip movement against the fault scarp. Paleoseismic disturbance of sediments is also present in a bluff on the north side of the Duke River. Late Pleistocene outwash gravel is displaced horizontally and vertically in the form of a positive flower structure. Many cobbles and pebbles within the gravel along the trace of the fault are tectonically broken. A fault mound at the top of the exposure is capped by aeolian sediments, including the two White River tephras, both of which are faulted and folded. The formation of positive flower structures in a strike-slip fault setting requires earthquakes with minimum magnitudes of ∼6.0. Hence, the earthquakes that we have documented are assumed to be at least this size. The average recurrence of large earthquakes on the Yukon portion of the Denali fault is estimated to be ∼1,300 years. In comparison, recurrence estimates for the Alaska portion of the fault are ∼1,000 years. A more detailed study of fault mounds in Shakwak Trench using lidar or high-resolution drone imagery would improve understanding of average slip rates. There likely is better measurable evidence of horizontal and vertical displacements in flatter areas of the fault where slope processes have not modified the mounds. In addition, along streams, there may be mounds that are incised, exposing positive flower structures in the late Pleistocene sediments. Finally, better age control on Crescent Lake cores would help refine the earthquake history that we infer using this paleoseismic proxy. ACKNOWLEDGMENTS Our research was funded by Natural Resources Canada’s Public Safety Geoscience Program, its Program for Energy Research and Development (1D00.006), and the Natural Sciences and Engineer-

160

ing Research Council of Canada (NSERC) through Discovery Grants to Clague and Menounos as well as a Canada Foundation for Innovation Grant to Menounos. Tyler Dimenna, Tracy Barry, and Richard Franklin drafted, respectively, Figures 2, 7, and 8. The Yukon Geological Survey provided tools and instrumentation. We thank Johannes Koch for assistance in the field. Comments and suggestions by critical reviewers Greg Brooks, Wouter Bleeker, Kristen Kennedy, Eldon Gath, Roy J. Shlemon, Marti Miller, and Richard Lease and an anonymous reviewer have greatly improved the manuscript. This is Geological Survey of Canada Contribution Number 20190100. We also thank the authors of Seitz et al. (2008), who first dug and logged the trenches. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government. REFERENCES Battarbee R. W.; Carvalho, L.; Jones, V. J., Flower, R. J.; Cameron, N. G.; Bennion, H.; and Juggins, S., 2001, Diatoms. In Smol, J. P.; Birks, H. J. B.; and Last, W. M. (Editors), Tracking Environmental Change Using Lake Sediments, Vol. 3: Kluwer Academic, Dordrecht, Netherlands, pp. 155–202. Bender, A. M. and Haeussler, P. J., 2017, Eastern Denali Fault Surface Trace Map, Eastern Alaska and Yukon, Canada: U.S. Geological Survey Open-File Report 2017-1049, 10 pp. Brahney, J.; Clague, J. J.; Edward, T. W. D.; and Menounos, B., 2010, Late Holocene paleohydrology of Kluane Lake, Yukon Territory, Canada: Journal Paleolimnology, Vol. 44, No. 3, pp. 873–885. Bronk Ramsey, C., 2017, Methods for summarizing radiocarbon datasets: Radiocarbon, Vol. 59, No. 6, pp. 1809–1833. Clague, J. J., 1979, The Denali Fault system in southwest Yukon Territory—A geologic hazard? In Current Research Part A: Geological Survey of Canada Paper 79-1A, pp. 169–178. Clague, J. J., 1982, The role of geomorphology in the identification and evaluation of natural hazards. In Craig, R. G. and Craft, J. L. (Editors), Applied Geomorphology: George Allen and Unwin, London, U.K., pp. 17–43. Clague, J. J.; Evans, S. G.; Rampton, V. N.; and Woodsworth, G. J., 1995, Improved age estimates for the White River and Bridge River tephras, western Canada: Canadian Journal Earth Sciences, Vol. 32, No. 8, pp. 1172–1179. Clague, J. J.; Luckman, B. H.; Van Dorp, R. D.; Gilbert, R.; Froese, D.; Jensen, B. J. L.; and Reyes, A. V., 2006, Rapid changes in the level of Kluane Lake in Yukon Territory over the last millennium: Quaternary Research, Vol. 66, No. 2, pp. 342–355. Dodds, C.J., 1995, Denali fault system. In Gabrielse, H. and Yorath, C. J. (Editors), Geology of the Cordilleran Orogen in Canada: Geological Survey of Canada, Ottawa, ON, Canada, pp. 656–657. Dodds, C. J. and Campbell, R. B., 1992, Geology, Southwest Kluane Map Areas 115G and F E1/2, Yukon Territory: Geological Survey of Canada Open File 2188, 1:250, 000 scale, p. 85. Doig, R., 1991, Effects of strong seismic shaking in lake sediments, and earthquake recurrence interval, Témiscaming, Québec: Canadian Journal Earth Sciences, Vol. 28, No. 9, pp. 1349–1352.

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 149–166

Evidence for Large Holocene Earthquakes Duk-Rodkin, A., 1999, Glacial Limits Map of Yukon Territory: Geological Survey of Canada Open File 3694, 1:1,000,000 scale. Eberhart-Phillips, D.; Haeussler, P. J.; Freymueller, J. T.; Frankel, A. D.; Rubin, C. M.; Craw, P.; Ratchkovski, N. A.; Anderson, G.; Carver, G. A.; Crone, A. J.; Dawson, T. E.; Fletcher, H.; Hansen, R.; Harp, E. L.; Harris, R. A.; Hill, D. P.; Hreinsdóttir, S.; Jibson, R. W.; Jones, L. M.; Kayen, R.; Keefer, D. K.; Larsen, C. F.; Moran, S. C.; Personius, S. F.; Plafker, G.; Sherrod, B.; Sieh, K.; Sitar, N.; and Wallace, W. K., 2003, The Denali fault earthquake, Alaska: A large magnitude, slip-partitioned event: Science, Vol. 300, No. 5622, pp. 1113–1118. Eisbacher, G. H., 1976, Sedimentology of the Dezadeash flysch and its implications for strike-slip faulting along the Denali fault, Yukon Territory and Alaska: Canadian Journal Earth Sciences, Vol. 13, pp. 1495–1513. doi:10.1139/e76-157. Elliott, J. L.; Larsen, C. F.; Freymueller, J. T.; and Motyka, R. J., 2010, Tectonic block motion and glacial isostatic adjustment in southeast Alaska and adjacent Canada constrained by GPS measurements: Journal Geophysical Research, Vol. 115, B09407, 21 p. doi:10.1029/2009JB007139. Fuis, G. and Wald, L. A. (Compilers), 2003, Rupture in South-Central Alaska—The Denali Fault Earthquake of 2002: U.S. Geological Survey Fact Sheet 014-13, 4 p. https://pubs.usgs.gov/fs/old.2003/fs014-03. Grantz, A., 1966, Strike-Slip Faults in Alaska: Unpublished Ph.D. Thesis, Stanford University, Stanford, CA, 158 p. Haeussler, P. J., 2008, An overview of the neotectonics of interior Alaska: Far-field deformation from the Yakutat microplate collision. In Freymuller, J. T.; Haeussler, P. J.; Wesson, R. L.; and Ekström, G. (Editors), Active Tectonics and Seismic Potential of Alaska: American Geophysical Union, Washington, DC, pp. 83–108. Haeussler, P. J., 2009, Surface Rupture Map of the 2002 M7.9 Denali Fault Earthquake, Alaska: Digital Data: U.S. Geological Survey Data Series Report DS-0422, 9 p. Haeussler, P. J.; Matmon, A.; Schwartz, D. P.; and Seitz, G. G., 2017, Neotectonics of interior Alaska and the late Quaternary slip rate along the Denali fault system: Geosphere, Vol. 13, No. 5, pp. 1–19. Haeussler, P. J.; Schwartz, D. P.; Dawson, T. E.; Stenner, H. D.; Lienkaemper, J. J.; Sherrod, B.; Cinti, F. R.; Montone, P.; Craw, P.; Crone, A. J.; and Personius, S. F., 2004, Surface rupture and slip distribution of the Denali and Totschunda faults in the 3 November 2002 M 7.9 earthquake, Alaska: Bulletin Seismological Society America, Vol. 94, No. 6B, pp. S23–S52. Harding, T. P., 1985, Seismic characteristics and identification of negative flower structures, positive flower structures, and positive structural inversion: American Association Petroleum Geologists Bulletin, Vol. 69, No. 4, pp. 582–600. Heginbottom, J. A., 1995, Canada Permafrost: National Atlas of Canada, 5th ed.: Natural Resources Canada, Map MCR 4177, 1:7,500,000 scale. Heginbottom, J. A. and Radburn, L. K., 1992, Permafrost and Ground Ice Conditions of Northwestern Canada: Geological Survey of Canada Map 1691A, 1:1,100,000 scale. Huscroft, C. A.; Lipovsky, P. S.; and Bond, J. D., 2004, A Regional Characterization of Landslides in the Alaska Highway Corridor, Yukon: Yukon Geological Survey Open File Report 2004-18, 65 p. Kelly, M. G.; Bennion, H.; Cox, E. J.; Goldsmith, B.; Jamieson, J.; Juggins, S.; Mann, D. G.; and Telford, R. J., 2005, Common Freshwater Diatoms of Britain and Ireland: An In-

teractive key: CD-ROM, Bristol Environment Agency, Bristol, U.K. Klassen, R. W., 1987, The Tertiary–Pleistocene Stratigraphy of the Liard Plain, Southeastern Yukon Territory: Geological Survey of Canada Paper 86-17, 16 p. Krammer, K. and Lange-Bertalot, H., 1985, Naviculaceae Bibliotheca Diatomologia, Band 9: J. Cramer, Berlin, Germany, 230 p. Krammer, K. and Lange-Bertalot, H., 1986, Bacillariophyceae Süsswasser flora von Mitteleuropa, Band 2/1: Gustav Fischer Verlag, Stuttgart, Germany, 876 p. Krammer, K. and Lange-Bertalot, H., 1988. Bacillariophyceae, Süsswasserflora von Mitteleuropa, Band 2/2: VEB Gustav Fischer Verlag, Jena, Germany, 596 p. Krammer, K and Lange-Bertalot, H., 1991, Bacillariophyceae Süsswasserflora von Mitteleuropa, Band 2/3: Gustav Fischer Verlag, Stuttgart, Germany, 576 p. Krammer, K. and Lange-Bertalot, H., 2000, Bacillariophyceae Süßwasserflora von Mitteleuropa, Band 2/5, Part 5: Gustav Fischer Verlag, Stuttgart, Germany, pp. 2–12. Lanphere, M. A., 1978, Displacement history of the Denali fault system, Alaska and Canada: Canadian Journal Earth Sciences, Vol. 15, No. 5, pp. 817–822. Lerbekmo, J. F., 2008, The White River ash: Largest Holocene Plinian tephra: Canadian Journal Earth Sciences, Vol. 45, No. 6, pp. 693–700. Lipovsky, P. S.; Seitz, G.; Haeussler, P. J.; Crone, A. J.; Schwartz, D. P.; Clague, J. J.; Mazotti, S.; and Cobbett, R., 2009, Neotectonic investigations in southwest Yukon: Abstract, Canadian Quaternary Association Meeting, Vancouver, BC, Canada. Lotter, A.; Pienitz, R.; and Schmidt, R., 2010, Diatoms as indicators of environmental change in subarctic and alpine regions. In Smol, J. P. and Stoermer, E. F. (Editors), The Diatoms: Applications for the Environmental and Earth Sciences, 2nd ed.: Cambridge University Press, Cambridge, U.K., pp. 231–248. Lowey, G. W., 1998, A new estimate of the amount of displacement on the Denali fault system based on the occurrence of carbonate megaboulders in the Dezadeash Formation (JuraCretaceous), Yukon, and the Nutzotin Mountains sequence (Jura-Cretaceous), Alaska: Bulletin Canadian Petroleum Geology, Vol. 46, No. 3, pp. 379–386. Marechal, A.; Ritz, J.-F.; Ferry, M.; Mazzotti, S.; Blard, P.-H.; Braucher, R.; and Saint-Carlier, R., 2018, Active tectonics around the Yukata indentor: New geomorphological constraints on the eastern Denali fault, Totschunda and Duke River faults: Earth Planetary Science Letters, Vol. 482, pp. 71–80. Mathews, W. H., 1986, Physiographic Map of the Canadian Cordillera: Geological Survey of Canada, Ottawa, ON, Canada, 1: 5,000,000 scale. Matmon, A.; Schwartz, D. P.; Haeussler, P. J.; Finkel, R.; Lienkaemper, J. J.; Stenner, H. D.; and Dawson, T. E., 2006, Denali fault slip rates and Holocene-late Pleistocene kinematics of central Alaska: Geology, Vol. 34, No. 8, pp. 645–648. Mavroulis, S.; Carydis, P.; Alexoudi, V.; Grambas, A.; and Lekkas, E., 2017, The January-February 2014 Cephalonia (Ionian Sea, western Greece) earthquakes: Tectonics and seismological facts: Proceedings of the 16th World Conference on Earthquakes, Santiago, Chile, Paper 413, 12 p. McCalpin, J. P.; Rockwell, T. K.; and Weldon, R.J., II, 2009, Paleoseismology of strike-slip tectonic environments. In McCalpin, J. A. (Editor), Paleoseismology, Vol. 95, 2nd ed.:

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 149–166

161

Blais-Stevens, Clague, Brahney, Lipovsky, Haeussler, and Menounos International Geophysics Series, doi:10.1016/S0074-6142(09)95006-9.

pp.

421–496.

McGimsey, R. G.; Richter, D. H.; DuBois, G. D.; and Miller, T. P., 1990, A postulated new source for the White River Ash, Alaska. In Geologic Studies in Alaska: U.S. Geological Survey Bulletin 1999, 7 p.

APPENDIX 1 Photomosaics of U.S. Geological Survey Trenches T2 and T3

Morell, K. D.; Regalla, C.; Amos, C.; Bennett, S.; Leonard, L.; Graham, A.; Reedy, T.; Levson, V.; and Telka, A., 2018, Holocene surface rupture history of an active forearc fault redeﬁnes seismic hazard in southwestern British Columbia, Canada: Geophysical Research Letters, Vol. 45, pp. 11605– 11611. doi:10.1029/2018GL078711. Northern Climate ExChange, 2013, Burwash Landing and Destruction Bay Landscape Hazards: Geological Mapping for Climate Change Adaptation Planning: Yukon Research Centre, Yukon College, Whitehorse, YT, Canada, 111 p. Patrick, R. and Reimer, C. W., 1966, The Diatoms of the United States Exclusive of Alaska and Hawaii, Vol. 1: Monographs of the Academy of Natural Sciences, Philadelphia, PA, No. 13, 688 p. Patrick, R. and Reimer, C. W., 1975, The Diatoms of the United States Exclusive of Alaska and Hawaii, Vol. 2: Monographs of the Academy of Natural Sciences, Philadelphia, PA, No. 13, 213 p.

Figure A1. Photomosaic of trench T2 looking north.

Rampton, V. N., 1979, Surficial Materials of Kluane Lake National Park, Yukon Territory: Geological Survey of Canada Map 1979-14, 1:250,000 scale. Rampton, V. N., 1981, Surficial Materials and Landforms, Kluane National Park, Yukon Territory: Geological Survey of Canada Paper, 37 p. Reimer, P. J.; Bard, E.; Bayliss, A.; Beck, J. W.; Blackwell, P. G.; Ramsey, C. B.; Buck, C. E.; Cheng, H.; Edwards, R. L.; Friedrich, M.; Grootes, P. M.; Guilderson, T. P.; Haidason, H.; Hajdas, I.; Hatté, C.; Heaton, T. J.; Hoffmann, D. L.; Hogg, A. G.; Hughen, K. A.; Kaiser, K. F.; Kromer, B.; Manning, S. W.; Niu, M.; Reimer, R. W.; Richards, D. A.; Scott, E. M.; Southon, J. R.; Staff, R. A.; Turney, C. S. M.; and van der Plicht, J., 2013, IntCal13 and Marine13 radiocarbon age calibration curves 0-50,000 years cal BP: Radiocarbon, Vol. 55, No. 4, pp. 1869–1887. Schulz, S. S. and Wallace, R. E., 2016, The San Andreas Fault: U.S. Geological Survey. https://pubs.usgs.gov/gip/earth3. Seitz, G. C.; Haeussler, P. J.; Crone, A. J.; Lipovsky, P.; and Schwartz, D. P., 2008, Eastern Denali fault slip rate and paleoseismic history, Kluane Lake area, Yukon Territory, Canada: American Geophysical Union, San Francisco, CA, Poster T53B-1947. http://www.geology.gov.yk.ca/ pdf/AGU_Denali_Fault_2009v1.pdf. Spaulding, S. A.; Pool, J.; Castro, S.; and Hinz, F., 2010, Species within the genus Encyonema Kützing, including two new species Encyonema reimeri sp. nov. and E. nicafei sp. nov. and E. stoermeri nom. nov., stat. nov.: Proceedings Academy Natural Sciences Philadelphia, Vol. 160, No. 1, pp. 57–71. Ulusay, R.; Aydan, Ö.; and Hamada, M., 2002, The behaviour of structures built on active fault zones: Examples from the recent earthquakes in Turkey: Structural Engineering/Earthquake Engineering, Vol. 19, No. 2, pp. 149–167. Veselá, J., 2015, Eunotia panda: Diatoms of North America. https://diatoms.org/species/eunotia_panda. Woodcock, N. H. and Fischer, M., 1986, Strike-slip duplexes: Journal Structural Geology, Vol. 8, No. 7, pp. 725–735.

162

Figure A2 and A3. Photomosaic of the trenches T2 and T3 looking south.

REFERENCE Seitz, G. C.; Haeussler, P. J.; Crone, A. J.; Lipovsky, P.; and Schwartz, D. P., 2008, Eastern Denali fault slip rate and paleoseismic history, Kluane Lake area, Yukon Territory, Canada: American Geophysical Union, San Francisco, CA, Poster T53B-1947. http://www.geology.gov.yk.ca/ pdf/AGU_Denali_Fault_2009v1.pdf.

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 149–166

Evidence for Large Holocene Earthquakes

APPENDIX 2 Supplementary Figures, Table, and Interpretation of Stratigraphic Changes in Diatom Assemblages

Supplementary Figure 1. Comparison of diatom community composition in cores 2 and 4 showing distinct communities associated with diďŹ&#x20AC;erent time intervals.

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 149â&#x20AC;&#x201C;166

163

Blais-Stevens, Clague, Brahney, Lipovsky, Haeussler, and Menounos

Supplementary Figure 2. Diatom stratigraphy of core 2 as percent of community composition.

Supplementary Figure 3. Diatom stratigraphy of core 4 as percent of community composition.

Notes: Color-coded diatom species are plotted in Figure 8a to facilitate understanding of the stratigraphy. The biostratigraphic correlations indicate that the upper and lower ages in core 4 (bold italics) are anomalously young. Core 4 Biostratigraphic Interpretations 101.5–98 cm: Lake Inception Lake Inception, Likely Due to Earthquake—There are very few diatom frustules present below 98 cm, and it is likely that this interval represents a change from a terrestrial environment to a lake. 96–0 cm Diatoms cycle though dominant species, indicating a shift from a shallow saline system to a deeper freshwater lake. There are also shifts in dominant species type, indicating a change in benthic habitat with changes in depth. 96–92.5 cm: Dark Brown Peaty Silt Lake Is Shallow, Possibly Brackish—Several observations support this interpretation. (1) There is a peak

164

in sponge spicules at 90 cm (60 cm in core 2). The presence of sponge spicules indicates that the core site was shallower during this period. (2) There are higher concentrations of large benthic species, in particular Stauroneis, a genus that often prefers brackish water, as well as the larger Sellaphora pupula and Pinnularia species. (3) The small benthic species Pseudostaurosira brevistriata, which is present in higher concentrations during this period, prefers shallow conditions in diatom calibration sets from the Yukon (Pienitz et al., 1995). This and other small benthic species are present during this interval and may represent pioneering benthic communities (Lotter et al., 2010). (4) There is an absence of planktonic species throughout this period. (5) During slide making, the sediment was coarse, so much so that it interfered with the binding of the coverslip to the slide. These sediments required sieving prior to mounting on a slide. (6) A white salt-like precipitate formed on the coverslips, indicating a higher soluble salt content. (7) Core 2 contains high concentrations of Eunotia praerupta, which is commonly associated with

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 149–166

Evidence for Large Holocene Earthquakes Appendix Table 1. Peak concentrations of diatom species in samples from cores 2 and 4.

increase in planktonic taxa, the lake may have further deepened. Following this period, from 78 to 60 cm (∼51 in core 2), there is an increase in Humidophyla gallixa and small benthic monoraphids (Sellaphora atomoides, Rossithidium pussillum, and Achnanthidium minutissiumum). In core 2, there is also an increase in small benthic species. In general, many Achnanthidium and benthic species (including Stauroseira, Stauorsirella, and Pseudostaurosira) are not motile and are attached either to sediment grains or to vegetation. Thus, their increase in abundance through this period may indicate changes in substrate and nutrient availability. The shift from benthic to planktonic and then back to benthic suggests that the lake deepened and then perhaps shallowed. 54 and 40 cm: Changes in Diatom Abundance and Composition Possible Earthquake Intervals—At both 54 cm and 40 cm in core 4, there are distinct changes in diatom abundance and composition, suggesting a shift in habitat and productivity following abrupt disturbances. Similarly, there is a silty layer with few diatoms at 41 cm in core 2, suggesting a disturbance that brought terrigenous sediment into the lake. Core 4

near-shoreline habitats, vegetation, or peat (Sawai et al., 2002). 90.5–55 cm: Gyttja with Laminae of Olive Gray Mud; Sharp Lower contact a. Lake Deepens, and Benthic Habitat Experiences Some Changes b. Deepest Period Was Probably 80–70 cm—From approximately 88 cm to 78 cm, there are large increases in concentrations of the epiphytic and/or planktonic Tabellaria flocculosa and then the planktonic Aulacoseira lirata (80–70 cm). This shift in species composition likely indicates an increase in lake depth such that mixing conditions can occur. Other Aulocaoseira species have been linked to strong mixing, as they are a heavily siliciﬁed diatom and require active mixing to stay aﬂoat (Sherman et al., 1998). Thus, the presence of this species likely represents a deepening of the lake such that longer spring circulation occurs (Horn et al., 2011). Anderson et al. (2007) determined that this time period was wetter in southern Yukon based on a lake-level reconstruction from a nearby lake. Core 2 also shows an increase in Aulacoseira lirata at this time although more generally an up-core increase in planktonic taxa. Because both cores show this upward

r r r r

Spike in diatom concentrations. Drops in chrysophyte cysts. Loss of species richness (Simpson’s Index). Declines in some benthic species, speciﬁcally Nitzschia spp. and Achnanthidium spp., and increases in benthic taxa that are more tolerant and are often “pioneers” in new habitats (Lotter et al., 2010).

Core 2

r 41 cm has few diatoms, indicating the introduction of terrigenous material into the lake.

r The main diatom found in this sample is Eunotia panda, which typically lives in shallow nearshore environments or bogs (Veselá, 2015). It is possible that these intervals record earthquake disturbance. An earthquake might be expected to redistribute nearshore nutrient-rich sediments and introduce silt and clay in the water body. Larger benthic species that are sensitive to turbidity (e.g., Nitzschia amphibia) are replaced by small benthic species that are more tolerant of poor light conditions and quickly reproduce to colonize new benthic habitats. The decrease in chrysophyte stomatocysts may be a result of a temporary increase in nutrient concentrations due to the introduction of

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 149–166

165

Blais-Stevens, Clague, Brahney, Lipovsky, Haeussler, and Menounos

terrigenous sediment (Sandgren, 1991; Smol and Stoermer, 2010). 40–0 cm: Dark Brown Organic Rich Sand Silty Mud (30–0 cm in Core 2) Establishment of Current Conditions—From 40 cm to the sediment surface, there is a transition to the modern lake. The slight increase in planktonic communities may suggest a minor deepening of the lake through this period. Throughout this part of core 4, there is a balance of planktonic (Discostella stelligera, Ulnaria ulna) and benthic species (Eunotids, Achnanthidium). Chrysophyte cysts and Cocconeis placentula also increase through this period. Taken together, these changes suggest a deepening of the lake and lower nutrient conditions. Cocconeis placentula can rapidly colonize new areas (Kelly et al., 2005), which supports a deepening lake. In addition, a decrease in epiphytic species and an increase in Nitzschoid species, which are motile among sediment (Kelly et al., 2005), may indicate a reduction in the relative amount of benthic vegetation cover with deepening of the lake. Kingsbury et al. (2012) documented species distributions across depth gradients in a variety of boreal lakes. They found that Discostella stelligera is found mainly in the deepest zone of the lakes they studied, which range from 7 to 25 m in depth. Naviculoids and small Fragillaria species (now reclassiﬁed as Stauroseira, Staurosirella, and Pseudostaurosira) were found at mid-depths (2–15 m) and in shallower water (0–7 m). Summary of the Interpreted Changes in Lake Depth and Water Chemistry in Core 4 1. 96–88 cm: Shallow water, higher salinity/nutrients. Sponges, Pseudostaurosira brevistriata (mesotrophic), and Stauroneis and Sellaphora, which are large unattached motile diatoms. Pioneer species indicate inception of the lake. 2. 86 cm: Deepening, planktonic species emerge: Tabellaria flocculosa (either planktonic or benthic) and Aulacoseira lirata, which requires active mixing to remain aﬂoat. 3. 80–70 cm: Dominance of Aulacoseira lirata in planktonic community. Lake perhaps deepest and/or longer spring turnover. 4. 70–60 cm: Lake possibly shallower. Emergence of epiphytic species (Achnanthidium minutissiumum, Humidophila gallica [often aerophyllic], and Nitzschia spp.). 5. ∼54 and 40 cm: Loss of species richness, particularly epiphytes (e.g., Nitzschia spp., Achnanthidium minutissiumum); proliferation of the episammic pi-

166

oneer species Staurosirellla pinnata as well as Pseudostaurosira brevistriata. These intervals appear to represent disturbances, suggesting that earthquakes caused increased sediment delivery to the lake. 6. 40–0 cm: Establishment of current conditions; lake possibly deepens with a reduction in nutrients. Reemergence of epiphytic species (Achnanthidium minutissiumum, naviculoids, Nizschia, and Cocconeis spp.) as well as planktonic species and Discostella stelligera. Diversity increases again, and productivity decreases. REFERENCES Anderson, L.; Abbott, M. B.; Finney, B. P.; and Burns, S. J., 2007, Late Holocene moisture balance variability in the southwest Yukon Territory, Canada: Quaternary Science Reviews, Vol. 26, No. 1, pp. 130–141. Horn, H.; Paul, L.; Horn, W.; and Petzoldt, T., 2011, Longterm trends in the diatom composition of the spring bloom of a German reservoir: Is Aulacoseira subarctica favoured by warm winters?: Freshwater Biology, Vol. 56, No. 12, pp. 2483–2499. Kelly, M. G.; Bennion, H.; Cox, E. J.; Goldsmith, B.; Jamieson, J.; Juggins, S.; Mann, D. G.; and Telford, R. J., 2005, Common Freshwater Diatoms of Britain and Ireland: An Interactive Key: CD-ROM, Bristol Environment Agency, Bristol, U.K. Kingsbury, M. V.; Laird, K. R.; and Cumming, B. F., 2012, Consistent patterns in diatom assemblages and diversity measures across water-depth gradients from eight boreal lakes from north-western Ontario (Canada): Freshwater Biology, Vol. 57, No. 6, pp. 1151–1165. Lotter, A.; Pienitz, R.; and Schmidt, R., 2010, Diatoms as indicators of environmental change in subarctic and alpine regions. In Smol, J. P. and Stoermer, E. F. (Editors), The Diatoms: Applications for the Environmental and Earth Sciences, 2nd ed.: Cambridge University Press, Cambridge, U.K., pp. 231–248. Pienitz, R.; Smol, J.; and Birks H. J., 1995, Assessment of freshwater diatoms as quantitative indicators of past climatic change in the Yukon and Northwest Territories, Canada: Journal of Paleolimnology, Vol. 13, No. 1, pp. 21–49. Sandgren, C., 1991, Chrysophyte reproduction and resting cysts: A paleolimnologist’s primer: Journal Paleolimnology, Vol. 5, No. 1, pp. 1–9. Sawai, Y.; Nasu, H.; and Yasuda, Y., 2002, Fluctuations in relative sea-level during the past 3000 yr in the Onnetoh estuary, Hokkaido, northern Japan: Journal of Quaternary Science, Vol. 17, No. 5–6, pp. 607–622. Sherman, B. S.; Webster, I. T.; Jones, G. J.; and Oliver, R. L., 1998, Transitions between Auhcoseira and Anabaena dominance in a turbid river weir pool: Limnology Oceanography, Vol. 43, No. 8, pp. 1902–1915. Smol, J. P. and Stoermer, E. F., 2010, The Diatoms: Applications for the Environmental and Earth Sciences. Cambridge University Press, Cambridge, U.K., 667 p. Veselá, J., 2015, Eunotia panda: Diatoms of North America. https://diatoms.org/species/eunotia_panda.

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 149–166

Landslide Susceptibility and Soil Loss Estimates for Drift Creek Watershed, Lincoln County, Oregon DAVID M. KORTE ABDUL SHAKOOR* Department of Geology, Kent State University, 221 McGilvery Hall, 325 S. Lincoln Street, Kent, OH, 44242

Key Terms: Landslide Inventory, Landslide Susceptibility, Siletz River Volcanics, Tyee Formation, Soil Loss Estimates, Logging Activities ABSTRACT Drift Creek watershed, Lincoln County, Oregon, is a source of drinking water as well as a reproductive habitat for endangered salmon and trout species. Landslides, exacerbated by logging, are suspected as a cause of water quality deterioration in the watershed. To investigate the impact of landslides on water quality, we mapped landslide distribution and susceptibility, determined engineering properties of landslide-prone soil and rock, and estimated soil loss resulting from landslide-derived sediment within 30 m of Strahler third-order-or-higher streams in the watershed. We mapped 570 landslides using LiDAR imaging, orthophotographs, and field observations. We used logistic regression to determine the most significant variables contributing to landslide occurrence and to create a watershed-scale landslide susceptibility map. Siletz River Volcanics and the sedimentary Tyee Formation make up 85 percent of the watershed, with the sedimentary Yamhill and Nestucca formations making up the majority of the rest. Sedimentary rocks dominate in the Upper Drift Creek watershed, and volcanic dominate in the lower portion. The largest landslide deposits and the highest susceptibility occur in the sedimentary rock formations. The Siletz River Volcanics has a larger abundance of landslides than the sedimentary rock formations, but they are smaller in size with lower susceptibility of occurrence. The soil loss model indicates that the average annual soil loss from landslide deposits in the Upper Drift Creek watershed is 65 tons/acre/yr compared to 29 tons/acre/yr in the Lower Drift Creek watershed. The model also indicates that soil loss from areas along roads in the watershed is high. INTRODUCTION The U.S. Department of Agriculture Forest Service (1996) has designated the Drift Creek watershed *Corresponding author email: ashakoor@kent.edu

(Figure 1) as a tier 1 key watershed. A key watershed is defined as a watershed that (1) provides habitat for potentially threatened species of anadromous salmonids or other potentially threatened fish or (2) covers more than 6 mi2 of area with high-quality water and fish habitat. The Oregon Department of Fish and Wildlife (2015) lists the watershed as a “source area” for steelhead trout, but populations of Coho salmon and steelhead trout have severely diminished in Drift Creek. The Mid-Coast Watersheds Council (2013) has designated the watershed as “impaired by unknown stressors,” with a high priority for protection and enhancement. The Oregon Department of Environmental Quality (DEQ, 2013) is concerned about the level of total maximum daily loads (the maximum amount of a pollutant that a water body can receive and still meet water quality standards) found in the watershed (Michie, personal communication, 2014). Suspected sediment influxes of unknown origin are changing the ecology of the watershed and stressing fish habitat through increasing turbidity, increasing conductivity, increasing temperature, reducing stream flow, decreasing dissolved oxygen concentrations, and decreasing general water quality (Salmon-Drift Creek Watershed Council, 2006). In addition, Lincoln City and the Confederated Tribes of Siletz Indians obtain drinking water from the watershed. Occasionally, water treatment operations in the watershed require shutting down due to turbidity resulting from excessive sediment, particularly below Strahler third-order streams (van de Wetering, personal communication, 2014). The DEQ, the Oregon Department of Geology and Mineral Industries (DOGAMI), and the Confederated Tribes of Siletz Indians suspect that landslides may be responsible for some of the water quality deterioration (Michie, personal communication, 2014). In the Strahler hierarchy of a stream network, each segment of a stream in the network is treated as a node in a tree, with the next segment downstream as its parent. When two first-order streams come together, they form a second-order stream; when two secondorder streams come together, they form a third-order stream; and so on (Strahler, 1957; Dunne and Leopold, 1978).

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 167–184

167

Korte and Shakoor

Figure 1. The Drift Creek watershed, Lincoln County, Oregon.

The objective of this study is to estimate landslidederived sediment delivered to Strahler third-order-orhigher streams. In order to accomplish this objective, we (1) created an inventory map of landslides present within 30 m of third-order-or-higher streams, (2) determined engineering properties of soil and rock that have the greatest inﬂuence on landslide distribution, and (3) created a landslide susceptibility map. The results of the study will provide DEQ the data necessary to determine where to allocate funds for compliance with the Clean Water Act of the U.S. Environmental Protection Agency (U.S. EPA, 1972). The decision about whether the soil loss estimates are excessive, acceptable, or of no concern will be up to the regulators. Drift Creek Watershed The Drift Creek watershed (Figure 1) is approximately 107 km2 in area. Elevation in the watershed ranges from 0 to 975 m, and slope angles range from 0° to 85° (average slope angle 25°). Streams have heavily dissected the area with no preferred direction of ridgelines. Major streams trend from northeast to southwest and from east to west. Long-term (30-year)

168

average annual and seasonal precipitation data are unavailable for the watershed. Newport, Oregon, 25 mi south of the watershed, is the only precipitation station with long-term records available that we used as a proxy for the watershed. The 30-year average annual precipitation for Newport is 170.7 cm. The seasonal 30-year averages are 40.7 cm for spring, 11.4 cm for summer, 45.2 cm for autumn, and 73.5 cm for winter (National Oceanic and Atmospheric Administration, 2019). The seasonal precipitation concentration leads to most landslides occurring during the winter months (Smith, 1978) as bedrock erosion increases and soil strength decreases due to the ﬂow and inﬁltration of water. The watershed is in its third or fourth forest harvest (van de Wetering, personal communication, 2014). Logging is currently active in the eastern region of the watershed ((Upper Drift Creek watershed) (Figure 2). Foresters are required to reclaim logged areas with conifers per the Forest Practices Act (Oregon Department of Forestry, 2014). The dominant bedrock formations in the watershed are the Siletz River Volcanics (74.8 km2 ) and the Tyee Formation (15.9 km2 ) (Figure 3; DOGAMI, 2014). The Yamhill Formation, Nestucca Formation,

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 167–184

Landslides and Soil Loss, Drift Creek OR

Figure 2. Orthophotograph showing logging activity in Upper Drift Creek watershed (to the east, outlined in red) near Strahler third-orderor-higher streams. Lighter green colors show limited regrowth, while absence of a green color indicates relatively fresh logging activity.

intrusive rocks, Depot Bay Basalt, Alsea Formation, and alluvial deposits (listed in order of age) make up the remaining 16.3 km2 area. The Yamhill, Nestucca, and Alsea formations are generally constrained to the eastern and western edges of the watershed. These units have eroded away from the interior of the water-

shed, leaving exposures of Siletz River Volcanics as the dominant bedrock by area (U.S. Department of the interior, U.S. Geological Survey, MROSD, 2017). The Siletz River Volcanics consist of aphanitic to porphyritic vesicular pillow flows, tuff-breccias, massive lava flows, and sills of tholeiitic and alkali

Figure 3. Bedrock units exposed in the Drift Creek watershed (DOGAMI, 2014).

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 167–184

169

Korte and Shakoor

basalt. The upper part of the sequence contains numerous interbeds of basaltic siltstone and sandstone, basaltic tuff, and locally derived basalt conglomerate. Zeolitization of the rocks and presence of calcite veins are pervasive within the unit. These rocks are of marine origin (oceanic crust and seamounts) and represent an offshore terrain, named “Siletzia,” which was accreted onto the Oregon coast 55–50 million years ago (Wells et al., 1998). Siletzia basalts (Siletz River Volcanics) form the basement of the Oregon Coast Range (OCR), comprising the core of the OCR anticline. The age of the Siletz River Volcanics spans the Paleocene to Eocene epochs (USGS, 2017). The Tyee Formation is a deep-water submarine fan complex that formed on top of the Siletzia terrane in a fore-arc basin 49–48 million years ago. Groove and flute casts indicate deposition by north-flowing turbidity currents. The Tyee Formation is thickest in the south and thinnest in the north. The Drift Creek watershed is at the very distal end of the fan complex, near the end of the Tyee Formation. The Tyee Formation is a sequence of thin, rhythmically bedded, medium- to fine-grained micaceous, feldspathic, lithic, or arkosic marine sandstone and micaceous, carbonaceous siltstone containing minor interbeds of dacite tuff in the upper part. Dip measurements in this watershed ranged between 11° and 17° with northeast to southeast directions. Probable provenance of the unit is the southwest Idaho batholith (Heller et al., 2016). The Tyee Formation is Middle Eocene in age (USGS, 2017). The Yamhill Formation is composed of massive thick- to thin-bedded concretionary marine siltstone, basaltic sandstone, and thin interbeds of arkosic, glauconitic, and basaltic sandstone. The Yamhill Formation locally contains interlayered basalt lava flows and lapilli tuff. The Yamhill Formation dates from Middle to Upper Eocene (USGS, 2017). The Nestucca Formation is composed of thick- to thin-bedded marine tuffaceous mudstone, siltstone, and fine- to coarse-grained friable sandstone. The formation contains calcareous concretions and is carbonaceous and micaceous in some places. The Nestucca Formation dates from Middle to Upper Eocene (USGS, 2017). Runoff potential in the watershed varies from moderately low to moderately high, and erodibility varies from moderate to high. RESEARCH METHODS Preparing a Landslide Inventory Map We prepared a landslide inventory map for the Drift Creek watershed using the protocol for inventory

170

mapping of landslide deposits from LiDAR imagery developed by Burns and Madin (2009). The protocol provides a consistent methodology for collecting and visualizing data, minimizing mapping errors, and enhancing the reliability of landslide hazard maps (Ardizzone et al., 2002; Haneberg, 2006; and Burns and Madin, 2009). DOGAMI provided a 10.8-point/m2 LiDAR-based digital elevation model (DEM) with 15-cm vertical resolution. The inventory map was created by manually digitizing 570 landslide deposits, associated scarps, scarp flanks (the area between the head scarp and the landslide deposit), and fans at 1:24,000, 1:10,000, and 1:4,000 scales. We used geomorphological features, such as arcuate head scarps, hummocky topography, toe bulges, and disrupted drainage patterns, to delineate landslide deposit boundaries. Not all of the landslide deposits exhibit all of these features. For example, some debris flows do not exhibit visible head scarps on the LiDAR imagery. We used polygons to digitize landslide deposits and scarp flanks and lines to digitize head scarps and secondary scarps (scarps within a landslide deposit). DOGAMI and the Oregon Geospatial Enterprise Office provided the database template used to create the inventory map. The database includes spatial and tabular data, such as slope movement type, slope angle, type of material (soil, rock, or debris), slope aspect, deposit area, deposit thickness, deposit volume, bedrock lithology, estimated age, deep- or shallowseated, and location data. Slope angle in the database is the pre-failure slope angle measured on an undisturbed slope adjacent to landslide deposits, using the DEM-derived slope raster. Estimating landslide deposit thickness is necessary to calculate deposit volume and to classify the landslide as either deep- or shallowseated. We estimated deposit thickness using the following trigonometric relationship between slope angle and head scarp height according to the mapping protocol (Burns and Madin, 2009): t = h cos (∝) ,

(1)

where t = landslide deposit thickness normal to the slope, h = head scarp height, and = slope angle. The protocol establishes a failure depth less than 4.5 m for shallow-seated landslides. Previous studies have shown that 4.5-m failure depth is representative for shallow-seated landslides in the OCR (Burns and Madin, 2009). Shallow-seated landslides have their failure surface through soils. Deep-seated landslides have failure depths greater than or equal to 4.5 m and have their failure surfaces in bedrock. We measured fan heights for debris ﬂows as the diﬀerence in elevation between the top and the bottom of the fan deposit. Landslide age was estimated as either historic (<150 years) or prehistoric (>150 years) according to the

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 167–184

Landslides and Soil Loss, Drift Creek OR

protocol. Historic landslides exhibited sharply defined features. Prehistoric landslides exhibited subdued features due to erosional processes and developed drainage channels. We also assigned a confidence level to landslide identification using the point-based system suggested in the DOGAMI protocol (Burns and Madin, 2009). The confidence of interpretation is based on the visual clarity of characteristic landslide topographic features, such as head scarps, flanks, toes, and internal deposit features (internal scarps, sag ponds, closed depressions, gullying, and so on) in the LiDAR and orthoimagery. In addition, we used field observations and orthophotographs to verify the inventory map where possible. Because of dense vegetation, only a handful of landslides could be verified. The Drift Creek watershed contains many complex slope movements (combination of more than one movement type), such as debris flows on earth flows or rotational slides on earth flows. Inventory mapping was limited to slope movements within 30 m of Strahler third-order-or-higher stream channels because landslide-derived sediment entering these stream channels was the primary concern of this study (Michie personal communication, 2014; van de Wetering, personal communication, 2014). For example, if there was a complex landslide, such as a large earth flow extending beyond 30 m upslope of a stream channel with two debris flows within 30 m of a stream channel, only the two debris flows near the stream channel were mapped. Individual landslides where the toes were within 30 m of a stream channel were mapped regardless of the distance between head scarp (or initiation zone) and stream channel. Other mapping exceptions included deposits associated with abandoned stream channels (no longer connected to other streams) and landslides farther than 30 m from stream channels. We report two landslide counts for inventory purposes: total landslide count for the entire watershed within 30 m of a third-order-or-higher stream channel and landslide count within 10 m of a third-order-orhigher stream channel, which accounts for the landslides closest to the stream channels.

used a penetrometer to record unconfined compressive strength adjacent to two landslide scarps in both Siletz River Volcanics and the Tyee Formation, taking 20 measurements at each site and averaging them for each soil type. Field cohesion (c) was estimated as c = qu /2. We used a torvane to measure shear strength in the field for Siletz River Volcanics and Tyee Formation soils, taking 20 measurements for each soil and averaging them. We measured bulk (wet) density of three Siletz River Volcanics soils and two Tyee Formation soils by pressing a coffee can into the soil, digging around the can, and slicing the soil at the bottom of the can. The thin walls of the cans caused minimal sample disturbance. The volumes and weights of the samples provided data for calculating bulk density of the soils in situ. Soil and rock samples were collected from accessible landslide scarps for laboratory testing. Soil samples were collected from 0.5- to 1-m depth, and rock samples were collected from 2- to 3-m depth. Grain size distribution, natural water content, Atterberg limits (liquid limit, plastic limit, plasticity index, and liquidity index), and shear strength parameters (friction and cohesion) were determined for the soil samples. We used the direct shear test for determining shear strength parameters for soil alone as well as along soil–rock contact. Soil samples for direct shear testing were compacted close to the field bulk density (2.05 and 2.11 g/cm3 for Siletz River Volcanics and Tyee Formation soils, respectively). For strength parameters along soil–rock contact, a slab of rock was cut to fit in the lower half of the shear box, and soil was placed in the upper half. Additionally, a secondcycle slake durability test was performed on rock samples from Siletz River Volcanics, the Tyee Formation, and the Yamhill Formation. Results of grain size distribution and Atterberg limits tests were used to classify soils according to the Unified Soil Classification System (USCS) (Casagrande, 1948). The slake durability test results were used to evaluate rock durability according to the classification of the International Society for Rock Mechanics (ISRM, 2007).

Determining Engineering Properties of Soil and Rock

Preparing the Landslide Susceptibility Map

We used both field and laboratory tests to determine engineering properties of soil and rock in the watershed. Where applicable, we performed all tests following standardized procedures by the American Society for Testing and Materials (ASTM, 2010). We used a Schmidt hammer to determine unconfined compressive strength (qu ) for the dominant rock types (Siletz River Volcanics and the Tyee Formation) at three accessible field sites, taking 20 measurements at each site and averaging them for each rock type. For soils, we

We created a new database for the landslide susceptibility map by adding the DEM-derived variables of curvature (plan and profile), slope aspect, terrain elevation, and water flow direction to the DOGAMI database template. Plan curvature is the curvature perpendicular to the slope, and profile curvature is the curvature in the direction of the slope. We obtained soil-type and erodibility data from a Custom Soil Resource Report for Lincoln County, Oregon, prepared by the U.S. Department of Agriculture

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 167–184

171

Korte and Shakoor

Natural Resources Conservation Service (USDA NRCS, 2017). For the land cover data, we used the U.S. Department of the Interior, U.S. Geological Survey, Multi-Resolution Land Characteristics Consortium National Land Cover Database (USDI USGS MRLCC, 2017). We used logistic regression to evaluate landslide susceptibility. Previous research shows that statistical methods can accurately predict the likelihood of landslide occurrence (Haneberg, 2004; Guzzetti et al., 2005; Guzzetti, 2006; Nandi and Shakoor, 2008; van Westen et al., 2008; and Nandi and Shakoor, 2009). The data set in this study includes non–normally distributed data, continuous data, categorical data, and binary data (presence or absence of a landslide). Unlike regular regression, logistic regression analysis can accommodate all types of problematic data (Costanzo et al., 2014; Budimir et al., 2015; Zhang et al., 2016; and Raja et al., 2017). For instance, a logistic regression model can use the occurrence of a landslide as a binary dependent variable where a value of 1 would indicate the presence of a landslide and a value of 0 would indicate the absence of a landslide. Values between 0 and 1 indicate the likelihood (probability) of landslide occurrence. Independent variables in the model are the factors contributing to landslides based on the data from the study area. Logistic regression calculates coefficients (weight factors) and significance (p) of each independent variable through an iterative process by evaluating each independent variable on its own, in combination with other independent variables, and with all independent variables. The significance level set for each individual independent variable in this study is 0.05. The null hypothesis is that each independent variable has no effect on the overall model. If p < 0.05, then the null hypothesis is rejected and the variable is significant to the model. We used the area under the receiver operating characteristic (ROC) curve to test the statistical validity of our model (Lombardo and Mai, 2018). The area under the curve is equal to the probability that the observer will correctly identify the positive case (landslide occurrence) when presented with a randomly chosen pair of cases in which one case is positive and one case is negative (no landslide occurrence). The method plots the average sensitivity (true positive fraction) against the entire range of average specificity (1 − specificity = false positive fraction). If the combination of observer and test are perfectly accurate, with 100 percent sensitivity and 100 percent specificity, then the ROC curve would consist of two straight-line segments encompassing a unit square. The area under the curve would be 1, and we would correctly “call” whether a landslide would or would not occur (Eng, 2005). The model with area under the ROC curve closest to 1 was chosen

172

as the most valid model. Quantitatively, the regression equation takes the form (2) P = 1/ 1 + e−z , where P is the probability of landslide occurrence and z is the combination of coeﬃcients and independent variables responsible for landslide occurrence. The parameter z can be expressed as z = β0 + β1V1 + β2V2 + . . . + βnVn ,

(3)

where β0 is a constant and βn is the coefficient for each respective independent variable Vn . The parameter z ranges from − to ; therefore, the range of P is from 0 to 1. We added landslide non-occurrence to the database by randomly selecting 570 points within 30 m of stream channels where landslides did not occur. We tagged these points with the same attributes as landslide occurrence. The process tests for P = 0 and P = 1 while generating coefficients βn and the significance of the independent variables in the model. We used variables with high significance (p < 0.05) in the model to maximize the area under the ROC curve. The model was then rasterized to create a 10-m2 -resolution landslide susceptibility map of the study area. We chose this resolution because none of the landslide deposits mapped were less than 10 m2 in area. For mapping purposes, we classified susceptibility as very high (P = 0.76–1), high (P = 0.51–0.75), medium (P = 0.26– 0.50), and low (P = 0.01–0.25). Estimating Average Annual Soil Loss We used the U.S. Department of Agriculture Agricultural Research Service (USDA ARS, 2017) RUSLE 2 equation in ArcGIS for estimating the soil loss in the Drift Creek watershed. The equation estimates average annual soil loss in tons per acre per year. We used ArcGIS instead of the USDA ARS RUSLE 2 v 2.6.8.4 program (USDA ARS, 2017) because of the variable soil types, attributes, and land cover in the Drift Creek watershed. This technique allowed for more accurate soil loss estimates by accounting for the spatial variations in soil types and conditions as well as variations in slope lengths, slope angles, and land cover attributes. The RUSLE equation is A = R × K × Ls × C × P,

(4)

where A = average annual soil loss, R = rainfall-runoﬀ erosivity factor, K = soil erodibility factor, Ls = slope length and steepness factor, C = cover management factor, and P = support practice factor. The R factor is a measure of the erosive force and intensity of rain in a normal year. R was calculated using the U.S. EPA (2017) rainfall erosivity factor calculator for a 30-year period from January 1, 1987, to

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 167–184

Landslides and Soil Loss, Drift Creek OR Table 1. C factors for land cover designations. Land Cover Value 11 21 22 23 24 31 41 42 43 52 71 90 95

Description

0 0.003 0.013 0.2 0.45 1 0.003 0.003 0.003 0.009 0.013 0.001 0.003

Open water Developed, open space Developed, low intensity Developed, medium intensity Developed, high intensity Barren land Deciduous forest Evergreen forest Mixed forest Shrub/scrub Herbaceous Woody wetlands Emergent herbaceous wetlands

October 20, 2017 (R = 4,260), and for January 1, 2016, to December 31, 2016 (R = 151), for Lincoln County, Oregon, because the calculator does not use watershed shapes or polygons for input extent. The K factor varies by soil type. There were 37 different K factor values in this study, corresponding to the 37 soil types. The K factor values are included in custom soil reports from the USDA NRCS but not included in the NRCS Soil Survey Geographic Database. We created a table of the soil types and respective K factors and added it to this database. The Ls factor quantifies the combined effect of slope length and steepness. The Ls factor values for different slope angles and slope lengths are available literature in the form of tables (Curtis et al., 2007) or can be calculated from the LiDAR DEM (Pelton et al., 2007). For this study, we calculated the Ls factors for each 10-m2 pixel because of the many different slope lengths and slope angles in the study area. The equation for Ls uses a flow accumulation (“flowacc”) and slope (“slope”) raster multiplied together. The individual equations for L and s are L = (“flowacc” × [resolution]/22.1)0.4

(5)

s = (sin(“slope” × 0.01745)/0.09)1.4 × 1.4.

(6)

and

The C factor (Table 1) reduces the soil loss estimate according to how eﬀective vegetation is at preventing detachment and transport of soil particles. C factor values were rasterized to 10 m2 using Table 1 generated from RUSLE 2 guidelines for the state of Oregon and land cover data sourced from the USDI USGS MRLCC (2017). The P factor is the ratio of soil loss with a given surface condition to soil loss with uphill and downhill plowing. P is >1 for compacted soils and <1 for loose soils. The neutral condition, P = 1, was used in this study.

We estimated and mapped the average annual soil loss at 10-m2 resolution over a 30-year period and for the year 2016. We also estimated the overall mean soil loss statistics for the entire watershed. The relationship between estimated average annual soil loss, mapped landslide deposits, and a 30-m stream buffer was investigated by intersecting the respective map layers and using zonal statistics to estimate the sum of how much soil sediment is delivered to the stream network by the landslide deposit stream buffer. The methodology developed in this study will enable soil loss estimates anywhere in the watershed and from landslide deposits in similar watersheds of similar size and physiographic location. RESULTS This study documents landslide distribution within 30 m of Strahler third-order-or-higher streams in the Drift Creek watershed, investigates the likelihood of landslide occurrence throughout the watershed, characterizes landslide-prone soil and rock, and estimates the average annual soil loss from landslide deposits that contributes sediment to streams channels. The results indicate that landslide-derived sediment entering stream channels is much greater in the currently active logging areas, including those in the process of restoration, than near-stream channels without adjacent logging activity. The DEQ should target streams near logging areas and logging roads in the Upper Drift Creek watershed (Drift, Fowler, Smith, and Sampson creeks) for water quality studies. Factors Controlling Landslide Distribution and Landslide Inventory Map The steep and converging topography of the Drift Creek watershed concentrates surface and sub-surface water flow toward gullies and valleys where thick packages of sediment accumulate. Gravity, steep slope angles, and precipitation events contribute to landslide initiation in these thick unconsolidated sediment packages. Soils developed on Siletz River Volcanics tend to retain larger amounts of water because of the relatively plastic nature of their fines (plasticity index = 12) before changing to a viscous liquid state. In contrast, the Tyee Formation soils have almost no capacity to retain water before changing to a viscous liquid state (plasticity index = 2). This makes Tyee Formation soils more susceptible to flow type movement than soils developed on Siletz River Volcanics during a precipitation event. On the other hand, soils derived from Siletz River Volcanics can creep with the addition of water and removal of surface vegetation. The Tyee Formation soils do not appear to creep when denuded.

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 167–184

173

Korte and Shakoor

Larger landslides in the watershed occur where one bedrock unit is in contact with another. These landslides are mostly complex landslides with two or more types of movement involved. The Tyee Formation fails where it is in contact with the more durable Siletz River Volcanics. The Yamhill Formation fails where it is in contact with the more durable Tyee Formation. This is supported by the bedrock geology map of the watershed (Figure 3) because most of the younger and less durable sedimentary rock has eroded away leaving the durable Siletz River Volcanics basement rock exposed over most of the region. The Siletz River Volcanics, Tyee, and Yamhill formations also result in slope failures where they are in contact with intrusive dikes in the watershed. These dikes appear to have cooled rapidly as indicated by their heavily fractured nature. The fractures allow water to inﬁltrate which increases pore pressures, facilitating slope failure. The total estimated landslide deposit volume exceeds 345.5 × 106 m3 , excluding 71 landslides whose thickness could not be determined. Two hundred and forty (42.1 percent) of the landslides are within 10 m of a third-order-or-higher stream channel and thus have a greater potential for contributing sediment to the streams. The estimated volume for these 240 landslides exceeds 277.5 × 106 m3 , excluding 33 landslides whose thicknesses could not be determined. Therefore, 80.3 percent of the volume of mapped landslides falls within 10 m of a third-order-or-higher stream. All of the mapped landslides had sharply deﬁned features

Table 2. Movement class and type of material for all 570 mapped landslides. Both rotational and translational slides can be either deepseated or shallow-seated. Type of Material Movement Class Fall Flow Rotational slide Translational slide Complex

Rock

Soil

Debris

1 0 6 1

1 50 188 0 160

2 160 1 0

and no observable drainage lines. Therefore, they all classified as historic landslides. Figure 4 shows the inventory map of the 570 landslides, including scarps and scarp flanks, within 30 m of third-order-or-higher streams. Using the classification of Cruden and Varnes (1996), we categorized all 570 landslides by type of movement and type of material, (Table 2). Of the 570 landslides, 299 are deep-seated and 271 shallowseated. We assigned high confidence to 465 landslides and moderate confidence to 105 landslides according to the DOGAMI protocol (Burns and Madin, 2009). One hundred and fifteen of the 240 landslides within 10 m of a third-order-or-higher stream channel are deep-seated, and 125 are shallow-seated, with 199 being high-confidence landslides and 41 moderateconfidence landslides.

Figure 4. Landslide inventory map, counts, and landslide slope statistics.

174

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 167–184

Landslides and Soil Loss, Drift Creek OR

Figure 5. Distribution of pre-failure slope angles for landslide deposits in Siletz River Volcanics (n = 448) and the Tyee Formation (n = 88).

Landslide area, normalized by bedrock area of individual formations, is the highest in the Yamhill Formation (0.82), followed by the Nestucca Formation (0.26), Tyee Formation (0.20), and Siletz River Volcanics (0.09). The factors influencing the normalization ratio are (1) the large area of the Yamhill Formation slides compared to a small area of the Yamhill Formation bedrock, (2) small bedrock areas of Nestucca and Tyee formations, and (3) mostly small landslides in the large bedrock area of the Siletz River Volcanics. Pre-failure slope angles by bedrock type are Siletz River Volcanics: range 15°–55° with most occurring between 31° and 35° (mean = 30°); Tyee Formation: range 15°–35° with peaks occurring between 22° and 25° (mean = 23°); Yamhill Formation: range 20°–22° (mean = 21°); and Nestucca Formation: range 17°– 33° with a peak at 25° (mean = 24°). Figure 5 shows the distributions of the pre-failure slope angles for the Siletz River Volcanics and Tyee Formation landslides. These two formations cover 85 percent of the watershed area. The dips of the sedimentary rock (Tyee Formation) in this watershed were less than the residual friction angles determined in the laboratory (see the next section) and therefore did not have a significant influence on landslide distribution. The dip aspect of the sedimentary formations (northeast to southeast) probably did not influence landslide distribution

because the dips were shallow and most of the mapped landslides were anti-dip in aspect. Engineering Properties of Dominant Rock and Soil Materials Present in the Watershed The results of in situ measurements of unconﬁned compressive strength for the Siletz River Volcanics and Tyee Formation as well as cohesion, shear strength, and bulk density values for the overlying soils are summarized in Table 3. Table 4 shows the percentages of gravel, sand, and ﬁnes (<0.074 mm) for the three samples of colluvial soil developed on the Siletz River Volcanics (0.75 m thick) and the three samples of colluvial soil derived from the Tyee Formation (1.2 m thick). Figures 6 and 7 show the respective grain size Table 3. Results of in situ measurements of rock and soil properties.

Field Measurements

Siletz River Volcanics

Tyee Formation

19,988

20,998

94.4

77.2

47.2 75.9 2.05

38.6 81.4 2.11

Rock unconﬁned compressive strength (kPa) Soil unconﬁned compressive strength (kPa) Soil cohesion (kPa) Soil shear strength (kPa) Soil bulk density (g/cm3 )

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 167–184

175

Korte and Shakoor

Figure 6. Grain size distribution curves for the three soil samples derived from Siletz River Volcanics (SRV).

distribution curves. There is much more gravel in the soil derived from the Siletz River Volcanics than in the soil derived from the Tyee Formation. Table 5 summarizes the overall USCS classification of the soils. The average size of landslide deposits in the Siletz River Volcanics (11,895 m2 ) is smaller than that in the Tyee Formation (14,450 m2 ). This could be because of the higher frictional resistance of the Siletz River Volcanics soil, as it contains larger amount and size of angular gravel and is relatively more well-graded (Figure 6) than the Tyee Formation soil. Well-graded soils have a higher angle of friction than poorly graded soils (Holtz et al., 2011). Table 6 summarizes the laboratory results of natural water content and Atterberg limits for the finegrained fractions (passing a 0.074-mm sieve) of the soil samples. Based on Atterberg limits, the fine-grained fractions of both Siletz River Volcanics and Tyee Formation soils classify as silts of low plasticity (ML) according to the USCS. The negative values of the liquidity index indicate that both soils are likely to beTable 4. Percentages of gravel, sand, and fines in the soil derived from Siletz River Volcanics and Tyee Formation. Soil Sample Siletz River Volcanics 1 Siletz River Volcanics 2 Siletz River Volcanics 3 Tyee Formation 1 Tyee Formation 2 Tyee Formation 3

176

Gravel (%)

Sand (%)

Fines (%)

42.2 42.4 43 1.8 1.5 1.2

44.8 45.1 44.8 88.1 87.9 86.9

13 12.5 12.2 10.1 10.6 11.9

have as brittle material when sheared at their respective natural water contents (Holtz et al., 2011). The high percentage of gravel in the Siletz River Volcanics soil and the silty nature of fines explain the abundance of debris flows in the landslide inventory. The contoured LiDAR imagery shows evidence of creep in the Siletz River Volcanics soils in the logging areas, which could be attributed to a higher plasticity index (12) of the fines in these soils and their greater ability to hold water. The second-cycle slake durability index (Id2 ) values for the Siletz River Volcanics, Tyee Formation, and Yamhill Formation samples, along with the type of material remaining after the second cycle per ASTM procedure (ASTM D4644; ASTM, 2010) and their durability classification, are shown in Table 7. Siletz River Volcanics rock is resistant to erosion, as indicated by its higher Id2 values It underlies the majority of the watershed (74.8 km2 ). The Tyee Formation is primarily sandstone with secondary siltstone and tuff Table 5. Classification of coarse and fine fractions of Siletz River Volcanics and Tyee Formation soils according to the USCS. USCS Classification Soil Type Siletz River Volcanics Tyee Formation

Particle Size 0.075 mm SM, poorly graded silty sand SP–SM, poorly graded sand–silty sand

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 167–184

Particle Size <0.074 mm ML, silts of low plasticity ML, silts of low plasticity

Landslides and Soil Loss, Drift Creek OR

Figure 7. Grain size distribution curves for the three soil samples derived from the Tyee Formation (TF).

and is not as durable as the Siletz River Volcanics. The Yamhill Formation is primarily thinly bedded and concretionary siltstone with secondary sandstone and tuﬀ. The very low durability of the Yamhill Formation makes it highly erodible, weathering to silts and clays (DOGAMI, 1973; USDA Forest Service, 1996). The landslide inventory map reﬂects the relative durability of these formations. On average, the durable Siletz River Volcanics landslide deposits are the smallest, whereas the very low-durability Yamhill Formation landslide deposits have the largest areas per deposit, with the Tyee Formation falling in between. Peak and residual friction angle and cohesion values for Siletz River Volcanics and Tyee Formation soils, sheared through soil and along soil– rock interfaces in the laboratory, are shown in Table 8. The results in Table 8 agree with those by Schroeder and Alto (1983), who tested soils from 11 sites in the Oregon–Washington coastal region using consolidated-undrained triaxial shear tests. Shearing Table 6. Natural water content and Atterberg limits for the finegrained fractions (<0.075 mm) of soils derived from Siletz River Volcanics and Tyee Formation.

Property Natural water content Liquid limit Plastic limit Plasticity index Liquidity index

Siletz River Volcanics Soil

Tyee Formation Soil

13 41 29 12 − 1.3

11 45 43 2 − 16

through soil resulted in higher friction angle values than shearing along a soil–rock interface (Table 8). Failure surfaces through soils were irregular compared to the soil–rock interface tests (rock blocks for interface tests were saw cut). We expect in situ friction angle values between soil and rock interfaces to be higher than laboratory values because of the nonplanar weathered bedrock surface. Siletz River Volcanics soils have higher peak and residual friction angle values and lower cohesion than Tyee Formation soils (Table 8). The higher friction angle values for Siletz River Volcanics soils translate into lower landslide potential (smaller landslide deposits) for these soils. Landslide Susceptibility The empirical probabilistic model used for evaluating landslide susceptibility in this study included non–normally distributed data, continuous data, categorical data, and binary data. The model used occurrence of a landslide as a binary dependent variable, used limited data because of the limited Table 7. Results of slake durability tests for Siletz River Volcanics, Tyee Formation, and Yamhill Formation rock samples.

Sample Material Siletz River Volcanics Tyee Formation Yamhill Formation

Slake Durability Material Durability Index (Id2 ) (%) Type Classiﬁcation 91 64 28

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 167–184

I II II

Medium-high Medium Very low

177

Korte and Shakoor Table 8. Direct shear test results for Siletz River Volcanics and Tyee Formation soils sheared through soil and along soil–rock interface. Friction Angle (°) Direct Shear Test

Peak

Siletz River Volcanics soil Siletz River Volcanics soil on Siletz River Volcanics rock Tyee Formation soil Tyee Formation soil on Tyee Formation rock

Cohesion (kPa)

Residual

Peak

Residual

37 23

36 23

32.4 24.1

9.7 6.2

29 16

19 16

69 66.9

39.3 15.2

observational opportunities and a partial landslide inventory, and did not consider precipitation data. Nevertheless, the model exhibited a high degree of statistical validity. The probability of landslide occurrence was higher along logging roads than along stream channels. This suggests (1) that a need exists to further inventory and characterize landslides along logging roads and (2) that soil loss resulting from roads may be a potential concern. The model can easily locate additional landslides in the LiDAR DEM that are not in proximity to the stream network. Running the model using only the independent variables with p < 0.05 (the most signiﬁcant contributors to landslide occurrence) calculates the parameter z as z = −15.383 + (0.096 × slope angle) + (0.003 × terrain elevation) + (36.536 × soil erodibility factor) + (−0.063 × plan curvature).

(7)

Figure 8 shows the distributions, statistical significance (p), coefficient value (β), and the standard error (SE) of the independent variables with p < 0.05. The coefficients in Figure 8, when multiplied by their respective independent variable, maximize the probability of landslide occurrence and non-occurrence (P = 1 and P = 0). Therefore, they maximize the number of true positive results and minimize the number of false positive results. This drives the area under the ROC curve toward a unit square (area = 1). The area under the ROC curve, using Eq. 2, is 0.8761 (Figure 9). Adding the insignificant variables to the model increased the area under the ROC curve by only 0.0033. Table 9 summarizes the significance of all of the independent variables used in the logistic regression model from most significant to least significant. The variables Bedrock, Land Cover, Slope Aspect, Flow Direction, and Profile Curvature contribute essentially nothing to the prediction in this study due to possible redundancy among themselves and, in the case of DEMderived variables, between themselves. This finding is not unique to this study. Remondo et al. (2003) found

178

Table 9. Significance of independent variables used in the logistic regression model. Independent Variable

Signiﬁcance (p)

Slope angle Terrain elevation Soil erodibility factor Plan curvature Bedrock Land cover Slope aspect Flow direction Proﬁle curvature

<0.001 <0.001 <0.001 0.034 0.127 0.281 0.330 0.478 0.881

that more variables do not necessarily contribute to a better result for probabilistic landslide susceptibility modeling in a watershed with similar characteristics to the Drift Creek watershed. Korte (2018) provides the landslide susceptibility parameter database developed for this study. We created a landslide susceptibility map by displaying the probability of landslide occurrence thematically (Figure 10). Table 10 summarizes the probability of landslide occurrence for the watershed in terms of area. Thirty-three percent of the watershed has a high to very high probability of landslide occurrence. The areas of high and very high probability of landslide occurrence are primarily above 390 m of terrain elevation with slope angles >25° (which exceeds the average slope angle for the entire watershed), along stream channels where landslide deposits were mapped, logging areas, and along roads leading to those logging areas. Probability of landslide occurrence is higher along logging roads (65 percent) than along stream channels (28 percent). Anthropogenic activities, such as construction and logging, often trigger landslides (Cruden and Varnes, 1996; Cornforth, 2005; Guzzetti, 2006; Schuster and Highland, 2007; Highland and Bobrowsky, 2008; and Hungr et al., 2014). The logging roads are poorly maintained gravel roads, typically constructed using balanced cut and fill and full bench construction methods (Oregon Department of Forestry, 2000). The roads are temporary and therefore not carefully engineered. As a result, these roads contribute a significant amount of sediment to the stream network by inhibiting infiltration and Table 10. Probability of landslide occurrence with corresponding watershed area. Watershed Area (m2 ) 16.1 17.1 21.4 45.4

Probability of Landslide Occurrence 0.76–1, very high 0.51–0.75, high 0.26–0.50, medium 0.01–0.25, low

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 167–184

Landslides and Soil Loss, Drift Creek OR

Figure 8. Statistical distributions and signiﬁcance (p), model coeﬃcient value (β), and the standard error (SE) of the independent variables with p < 0.05.

increasing surface runoﬀ during precipitation events (Montgomery et al., 2000). High and very high probability of landslide occurrence correlates with material properties determined in the laboratory. The areas of highest susceptibility occur in the very low durability Yamhill Formation and medium-durability Tyee Formation. The mapped landslides are largest in area in these formations. In contrast, the more durable Siletz River Volcanics has mostly low to medium probability of landslide occurrence and smaller landslide size.

There is one exception for the Siletz River Volcanics, the area near 44°55 N, 123°55 W (Figure 10). For this speciﬁc area, the model shows high to very high probability of landslide occurrence. The slope angles in this area are >40°, which is higher than the friction angle for soil derived from Siletz River Volcanics as determined in the laboratory (Table 8). The plan curvature (curvature perpendicular to the slope) in this area oscillates from very high to very low. Thus, this area has the geometry of steep and highly incised hills. The incisions result from gullies in the landscape due to

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 167–184

179

Korte and Shakoor

Figure 9. ROC curve for landslide susceptibility model.

rainfall runoﬀ and lead the model to predict high to very high probability of landslide occurrence. We mapped several landslides within 30 m of Strahler third-order-or-higher streams in this area (Figure 10). These landslides have the potential to deliver sediment and debris to the stream network. The landslide

susceptibility map (Figure 10), constructed from our landslide inventory of the watershed (within 30 m of Strahler third-order-or-higher streams), can serve as an eﬀective means to quickly and easily ﬁnd additional landslides in other areas of the watershed designated as areas of high landslide susceptibility. The mean probability of landslide occurrence within 30 m of Strahler third-order-or-higher streams along the entire 122-km stream network is 28 percent (medium). While this statistic may be useful on a watershed scale for characterization purposes, it is biased by large areas of low probability of landslide occurrence in the west and south of the watershed (Figure 10). Slope angles, curvature, and terrain elevation are lower in the west and the south than they are to the north and to the east of the watershed. High elevations, steep slope angles, weak and nondurable sedimentary bedrock, high soil erosion factors, and recently logged areas and roads characterize areas of high landslide susceptibility. The Upper Drift Creek watershed has all of these characteristics, and 65 percent of it maps as having high to very high susceptibility. The Upper Drift Creek watershed also has the highest average annual soil loss estimate. The most important omission in the model is water. Although the area of the watershed is nearly 107 km2 , data regarding variation of precipitation across the watershed are not available. Additionally, there have been no published pore pressure studies in the study area.

Figure 10. Landslide susceptibility map of the Drift Creek (Siletz) watershed showing probability of landslide occurrence.

180

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 167–184

Landslides and Soil Loss, Drift Creek OR

Figure 11. Landslide deposits and associated soil loss estimates within 30 m of Strahler third-order streams.

Precipitation monitoring stations and pore pressure studies could help further refine the landslide susceptibility model and provide essential data when assessing a specific area of interest. Estimated Average Annual Soil Loss We estimated soil loss specifically from landslide deposits that were within 30 m of stream channels (Figure 11). The highest soil loss occurs within recent logging areas in the Upper Drift Creek watershed. The soil erodiblity factors in these areas are among the highest in the watershed, ranging from 0.37 to 0.4. The combination of high soil erodibility factor and logging activity (forest removal and time to restore land cover) contributes to greater soil loss estimates than in the Lower Drift Creek watershed. Currently, there is little logging activity in the Lower Drift Creek watershed. In addition, the soils derived from the Siletz River Volcanics have lower erodibility factors than the soils derived from the sedimentary bedrock in the Upper Drift Creek watershed. The area of the landslide deposits within 30 m of the stream network in Upper Drift Creek is 128 acres, and the total estimated annual soil loss for those 128 acres is 65 tons/acre/yr. Lower Drift Creek has 85 acres of landslide deposits within 30 m of the stream network, and the estimated annual soil loss is 29 tons/acre/yr. This analysis shows that landslide deposits where there is logging activity contribute substantially more

sediment to streams in the Upper Drift Creek watershed than landslide deposits in the Lower Drift Creek watershed, where there is little logging activity. The model points out the exact landslide deposits and stream channels where the DEQ should focus initial studies of deteriorating water quality resulting from landslide-derived sediment entering stream channels. We estimated the average annual soil loss (specific sediment yield) over a 30-year period (1986–2016) to be 27 tons/acre/yr over the entire extent of the watershed. Our estimate of the average soil loss for 2016 was 28 tons/acre/yr, which is about the same as the 30-year annual average. Specific sediment yield is the sediment yield per unit area of the drainage basin, in this case the watershed, and is equivalent to the average annual soil loss. Milliman and Syvitski (1992) summarized the relationship between watershed area and specific sediment yield using data from 280 rivers classified into seven different groups based on the maximum elevation in each basin. According to Milliman and Syvitski (1992) data, the specific sediment yield for a watershed the size of the Drift Creek watershed is approximately 29.5 tons/acre/yr. The estimate of average annual soil loss, using the RUSLE equation in this study, is within 5 percent of Milliman and Syvitski’s results. Therefore, the RUSLE equation provides reasonable soil loss estimates for the entire Drift Creek watershed. The average annual soil loss model used in this study allows examination of small or large areas of interest. The model can estimate average annual soil loss

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 167–184

181

Korte and Shakoor

anywhere in the watershed, at any distance from the stream network, at any specific landslide deposit, for any specific land cover condition, and near any road. Debris and earth flows (102 in total) are the most common types of landslide movement within 10 m of third-order-or-higher stream channels and therefore contribute the most fine-grained sediment to the stream network. The two dominant rock types, Siletz River Volcanics and Tyee Formation, weather into poorly graded sandy and silty soils. The silt content in these soils poses a water quality problem for drinking water by blocking filtration. It also causes an ecological problem for salmon and trout by blanketing gravel beds needed for reproduction after entering a stream channel entrained in debris and earth flows (SalmonDrift Creek Watershed Council, 2006). The study does not address sediment transport because stream flow or stream geometry data are not available for the streams in the Drift Creek watershed. Stream flow and stream geometry data are essential for estimating the capacity of a stream to transport bedload and suspended load. One can use grain size distributions from this study to estimate the amount of landslide-derived sediment transported and its effect on water quality and ecological habitat (temperature, turbidity, effect on oxygen concentration, and total maximum daily load of pollutants) in the Drift Creek watershed. There are turbidity measurements for the nearby streams (Cooper, 2005; National Water Quality Monitoring Council, 2017). However, because of different conditions. such as terrain elevation, bedrock, and soils, one cannot use these measurements as a proxy for streams in the Drift Creek watershed.

Limitations of the Study The study presented herein has the following limitations: 1. Landslides smaller than 100 m2 were excluded because of the resolution of LiDAR and aerial orthophotograph. 2. Calculating landslide deposit volume in the Burns and Madin (2009) protocol is based on the assumptions of a planar failure surface with uniform depth. This assumption may not be valid for the heterogeneous nature of the rock and soil in the Drift Creek watershed. Moreover, rotational failures typically vary in depth from head to toe with the deepest part near the center of the failure surface. 3. The role of water in landslide susceptibility was not considered because of a lack of data regarding precipitation variability and pore pressure measurements.

182

CONCLUSIONS The conclusions of this study are as follows: 1. Among the 570 mapped landslides, 299 are deepseated (>15-ft depth), and 271 are shallow-seated. The pre-failure slope angles range from 31°–35° for the Siletz Formation, 22°–25° degrees for the Tyee Formation, 21° for the Yamhill Formation, and 24° for the Nestucca Formation. The average landslide areas are 11,894 m2 for the Siletz Formation and 14,450 m2 for the Tyee Formation. The largest landslides, in terms of landslide deposit area, occur in the Upper Drift Creek watershed, which has erodible sedimentary rock (Tyee and Yamhill formations), high terrain elevations, and areas where there has been recent logging activity and limited regrowth of forest. Smaller landslides occur in the Siletz River Volcanics in both the upper and the lower portions of the Drift Creek watershed. Landslide deposit area, normalized by bedrock area, is the highest for the Yamhill Formation (low durability), intermediate for the Tyee Formation (medium durability), and the lowest for the Siletz River Volcanics (as medium-high durability). Debris flows and earth flows are the most common types of landslides and contribute most to the finegrained sediment of the streams. Larger landslides occur where two formations are in contact with one another. 2. The most significant variables affecting landslide susceptibility in the Drift Creek watershed, indicated by the logistic regression model, are slope angle, terrain elevation, soil erodibility factor, and plan curvature. The landslide susceptibility model exhibits high statistical validity with an area under the ROC curve of 0.8761. Probability of landslide occurrence is higher along logging roads (65 percent) than along stream channels (28 percent). The highest susceptibility occurs in the low- and medium-durability Yamhill and Tyee formations, which also have the largest landslides. 3. The Siletz River Volcanics soils classify as SM (poorly graded sand), whereas the Tyee Formation soils classify as SP-SM (poorly graded sand to silty sand). 4. The 30-year average annual soil loss for the entire Drift Creek watershed is 27 tons/acre/yr using the RUSLE equation. The average annual soil loss estimate resulting from this study is 28 tons/acre/yr in 2016. The Upper Drift Creek section of the watershed exhibits more than twice the estimated soil loss within 30 m of Strahler third-orderor-higher stream network than the Lower Drift Creek section. The highest soil loss estimates from

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 167–184

Landslides and Soil Loss, Drift Creek OR

landslide-derived sediment occur in logging areas adjacent to the stream channels. 5. The DEQ should target the Upper Drift Creek watershed (Drift, Fowler, Smith, and Sampson creeks) for water quality studies. ACKNOWLEDGMENTS The authors would like to thank the Oregon Department of Environmental Quality and the Confederated Tribes of Siletz Indians for the opportunity to conduct this study. The Oregon Department of Geology and Mineral Industries provided the LiDAR digital elevation data for this study and other logistical help that the authors appreciate greatly. REFERENCES American Society for Testing and Materials (ASTM), 2010, Annual Book of Standards: Section 4, Construction, 4.08, Soil and Rock (I), ASTM International, West Conshohocken, PA. Ardizzone, F.; Cardinali, M.; Carrara, A.; Guzzetti, F.; and Reichenbach, P., 2002, Impact of mapping errors on the reliability of landslide hazard maps: Natural Hazards and Earth System Sciences, Vol. 2, pp. 3–14. Budimir, M. E. A.; Atkinson, P. M.; and Lewis, H. G., 2015, A systematic review of landslide probability modelling using logistic regression: Landslides, Vol. 12, pp. 419–436. Burns, W. J. and Madin, I. P., 2009, Protocol for Inventory Mapping of Landslide Deposits from Light Detection and Ranging (LiDAR) Imagery: Oregon Department of Geology and Mineral Industries Special Paper 42, 36 p. Casagrande, A., 1948, Classification and identification of soils: Transactions, American Society of Civil Engineers, Vol. 113, pp. 901–930. Cooper, R., 2005, Estimation of Peak Discharges for Rural, Unregulated Streams in Western Oregon: U.S. Geological Survey Scientific Investigations Report 2005–5116, 134 p. Cornforth, D. H., 2005, Landslides in Practice: John Wiley and Sons, Hoboken, NJ, 596 p. Costanzo, D.; Chacon, J.; Conoscenti, C.; Irigaray, C.; and Rotigliano, E., 2014, Forward logistic regression for earthflow landslide susceptibility assessment in the Platani river basin (southern Sicily, Italy): Landslides, Vol. 11, pp. 639–653. 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, pp. 36–75. Curtis, N. M.; Darrach, A. G.; and Sauerwein, W. J., 2007, Estimating Sheet-Rill Erosion and Sediment Yield on Disturbed Western Forest and Woodlands: U.S. Department of Agriculture, Soil Conservation Service, Portland, OR, 36 p. Dunne, T. and Leopold, L., 1978, Water in Environmental Planning: W. H. Freeman and Company, San Francisco, CA, 818 p. Eng, J., 2005, Receiver operating characteristic analysis: A primer, Academic Radiology, Vol. 12, No. 7, pp. 909–916. Guzzetti, F., 2006, Landslide Hazard and Risk Assessment: Ph.D. Thesis, University of Bonn, Bonn, Germany, http://hss.ulb.uni-bonn.de/2006/0817/0817.htm

Guzzetti, F.; Reichenbach, P.; Cardinali, M.; Galli, M.; and Ardizzone, F., 2005, Probabilistic landslide hazard assessment at the basin scale: Geomorphology, Vol. 72, pp. 272–299. Haneberg, W. C., 2004, A rational probabilistic method for spatially distributed landslide hazard assessment: Environmental and Engineering Geoscience, Vol. 10, pp. 27–43. Haneberg, W. C., 2006, Eﬀects of digital elevation model errors on specially distributed seismic slope stability calculations: An example from Seattle, Washington: Environmental and Engineering Geoscience, Vol. 12, pp. 247–260. Heller, P. L.; Peterman, Z. E.; O’Neil, J. R.; and Shafiqullah, M., 2016, Isotopic provenance from the Eocene Tyee Formation, Oregon Coastal Range, Geological Society of America Bulletin, Vol. 96, No. 6, pp. 770–780. Highland, L. and Bobrowsky, P., 2008, The Landslide Handbook—A Guide to Understanding Landslides: U.S. Geological Survey Circular 1325, 129 p. Holtz, R. D.; Kovacs, W. D.; and Sheaham, T. C., 2011, An Introduction to Geomechanical Engineering: Pearson Education, Upper Saddle River NJ, 853 p. Hungr, O.; Leroueil, S.; and Picarelli, L., 2014, The Varnes classiﬁcation of landslide types, an update: Landslides, Vol. 11, p. 167. International Society for Rock Mechanics (ISRM), 2007, The Complete ISRM Suggested Methods for Rock Characterization, Testing and Monitoring, 1974–2006: Suggested Methods Prepared by the Commission of Testing Methods, International Society for Rock Mechanics, Ulusay R. and Hudson, J. A. (Editors): Ankara, Turkey, compilation arranged by the ISRM Turkish National Group, 293 p. Korte, D. M., 2018, Landslide Distribution and Susceptibility, Material Properties, and Soil Loss Estimates for the Drift Creek Watershed (Siletz River), Lincoln County, Oregon: Ph.D. Dissertation, Kent State University, Kent, OH, 145 p. Lombardo, L. and Mai, P.M., 2018, Presenting logistic regression– based landslide susceptibility results: Engineering Geology, Vol. 244, pp. 14–24. Michie, R., 2014, personal communication, Oregon Department of Environmental Quality, Portland, OR. Mid-Coast Watersheds Council, 2013, Mid-Coast Watersheds Biological Monitoring Results Survey: Electronic document, available at http://www.midcoastwatersheds.org Milliman, J. D. and Syvitski, J. P. M., 1992, Geomorphic/tectonic control of sediment discharge to the ocean: The importance of small mountain rivers: Journal of Geology, Vol. 100, pp. 525–544. Montgomery, D. R.; Schmid, T.; Kevin, M.; Greenberg, H. M.; and Dietrich, W. E., 2000, Forest clearing and regional landsliding: Geology, Vol. 28, No. 4, p. 311–314. Nandi, A. and Shakoor, A., 2008, Application of logistic regression model for slope instability prediction in Cuyahoga River Watershed, Ohio, USA: Georisk: Assessment and Management of Risk for Engineered Systems and Geohazards, Vol. 2, No. 1, pp. 16–27. Nandi, A. and Shakoor, A., 2009, A GSI-based landslide susceptibility evaluation using bivariate and multivariate statistical analyses: Engineering Geology, Vol. 110, No. 1–2, pp. 11–20. National Oceanic and Atmospheric Administration, 2019, Climate Data Online (CDO): Electronic document, available at https://www.ncdc.noaa.gov/cdo-web/datatools/normals National Water Quality Monitoring Council, 2017, Water Quality Portal: Elecronic document, available at https://www.waterqualitydata.us

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 167–184

183

Korte and Shakoor Oregon Department of Environmental Quality (DEQ), 2013, Construction of Stormwater Erosion and Sediment Control Manual: Water Quality Division, Surface Water Management, 22 p. Oregon Department of Fish and Wildlife, 2015, ODFW Home Page: Electronic document, available at http://www.dfw.state.or.us Oregon Department of Forestry, 2000, Forest Road Construction: Oregon Department of Forestry, Veneta, OR, 11 p. Oregon Department of Forestry, 2014, Forest Practice Administrative Rules and Forest Practices Act: Chapter 629. Forest Practices Administration, 94 p. Oregon Department of Geology and Mineral Industries (DOGAMI), 1973, Stratigraphic and biostratigraphic relationships of the Tyee and Yamhill Formations in central-western Oregon: The Ore Bin, Vol. 35, No. 11, pp. 172–174. Oregon Department of Geology and Mineral Industries (DOGAMI), 2014, Oregon Geologic Data Compilation (OGDC), Vol. 5: Electronic document, available at https://www.oregongeology.org/gis Pelton, J.; Frazier, E.; and Pickilingis, E., 2007, Calculating slope length factor (LS) in the Revised Universal Soil Loss Equation (RUSLE): Resonance, Vol. 12, No. 7, pp. 7. Raja, N. B.; Cicek, I.; Turkoglu, N.; Aydin, O.; and Kawasaki, A., 2017, Landslide susceptibility mapping of the Sera River Basin using logistic regression model: Natural Hazards, Vol. 85, pp. 1323–1346. Remondo, J.; González-Díez, A.; De Terán, J. R. D.; and Cendrero, A., 2003, Landslide susceptibility models utilizing spatial data analysis techniques: A case study from the Lower Deba Valley, Guipuzcoa, Spain: Natural Hazards, Vol. 30, No. 3, pp. 267–279. Salmon-Drift Creek Watershed Council, 2006, Volunteer Water Quality Monitoring: Final sport and accounting, Oregon Watershed Enhancement Board grant 204-074, 21 p. Schroeder, W. L. and Alto, J. V., 1983, Soil properties for slope stability analysis—Oregon and Washington coastal mountains: Forest Service, Vol. 29, p. 823–833. Schuster, R. L. and Highland, L. M., 2007, Overview of the effects of mass wasting on the natural environment: Environmental and Engineering Geoscience, Vol. 13, p. 25–44. Smith, E., 1978, Determination of Coastal Changes in Lincoln County, Oregon Using Aerial Photographic Interpretation: Master’s Thesis, Portland State University, Portland, OR, 29 p.

184

Strahler, A. N., 1957, Quantitative analysis of watershed geomorphology: Transactions of the American Geophysical Union, Vol. 38, No. 6, pp. 913–920. U.S. Department of Agriculture (USDA) Agricultural Research Service (ARS), 2017, Watershed Physical Processes Research: Oxford, MS: Electronic document, available at https://www.ars.usda.gov/southeast-area/oxfordms/national-sedimentation-laboratory/watershed-physicalprocesses - research / research / rusle2 / revised - universal - soilloss-equation-2-overview-of-rusle2 U.S. Department of Agriculture Forest Service, 1996, Drift (Siletz) Watershed Analysis: General Technical Report PNW8, 417 p. U.S. Department of Agriculture National Resources Conservation Service (USDA NRCS), 2017, Web Soil Survey: Electronic document, available at https://websoilsurvey. sc.egov.usda.gov/App/WebSoilSurvey.aspx U.S. Department of the Interior, U.S. Geological Survey (USGS), 2017, Mineral Resources On-Line Spatial Data (MROSD): Electronic document, available at http:// mrdata.usgs.gov/geology/state/ﬁps-unit.php?code=f41041 U.S. Department of the Interior, U.S. Geological Survey, Multi-Resolution Land Characteristics Consortium (USDI USGS MRLCC), 2017, Electronic document, available at https://www.mrlc.gov/data U.S. Environmental Protection Agency (U.S. EPA), 1972, Summary of the Clean Water Act: Electronic document, available at https://www.epa.gov/laws-regulations/summary-cleanwater-act U.S. Environmental Protection Agency (U.S. EPA), 2017, Rainfall Erosivity Factor Calculator: Electronic document, available at https://www.epa.gov/npdes/rainfall-erosivityfactor-calculator-small-construction-sites#getTool van De Wetering, S., 2014, personal communication, Confederated Tribes of Siletz Indians. van Westen, C. J.; Castellonos, E.; and Kuriakose, S. L., 2008, Spatial data for landslide susceptibility, hazard, and vulnerability assessment: An overview: Engineering Geology, Vol. 102, pp. 241–255. Wells, R. E., Weaver, C. S., and Blakely, R. J., 1998, Fore-arc migration in Cascadia and its neo-tectonic signiﬁcance: Geology, Vol. 26, No. 8, pp. 759–762. Zhang, M.; Cao, X.; Peng, L.; and Niu, R., 2016, Landslide susceptibility mapping based on global and local logistic regression models in the Three Gorges Reservoir area, China: Environmental Earth Sciences, Vol. 75, pp. 958–969.

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 167–184

Landslide Susceptibility Mapping in the Commune of Oudka, Taounate Province, North Morocco: A Comparative Analysis of Logistic Regression, Multivariate Adaptive Regression Spline, and Artiﬁcial Neural Network Models SAID BENCHELHA* HASNAA CHENNAOUI AOUDJEHANE GAIA Laboratory, Hassan II University of Casablanca, Faculty of Sciences, Aïn Chock, Morocco

MUSTAPHA HAKDAOUI LGAGE Laboratory, Hassan II University of Casablanca, Faculty of Sciences, Ben M’sik, Morocco

RACHID EL HAMDOUNI Department of Civil Engineering, University of Granada, Granada 18071, Spain

HAMOU MANSOURI Laboratoire Public d’Essai et d’Etudes (LPEE), Casablanca, Morocco

TAOUFIK BENCHELHA GAIA Laboratory, Hassan II University of Casablanca, Faculty of Sciences, Aïn Chock, Morocco

MOHAMMED LAYELMAM Hassan II Agronomic and Veterinary Institute, Rabat, Morocco

MUSTAPHA ALAOUI Laboratory of Management and Valorization of Natural Resources, Faculty of Science and Technology, Sultan Moulay Slimane University, Beni Mellal, Morocco

Key Terms: Landslide Susceptibility Mapping, Multivariate Adaptive Regression Splines, Logistic Regression, Artificial Neural Networks, Oudka, Taounate (Morocco) ABSTRACT Landslide susceptibility indices were calculated and landslide susceptibility maps were generated for the Oudka, Morocco, study area using a geographic information system. The spatial database included current landslide location, topography, soil, hydrology, and lithology, and the eight factors related to landslides

*Corresponding author email: said1.benchelha@gmail.com

(elevation, slope, aspect, distance to streams, distance to roads, distance to faults, lithology, and Normalized Difference Vegetation Index [NDVI]) were calculated or extracted. Logistic regression (LR), multivariate adaptive regression spline (MARSpline), and Artificial Neural Networks (ANN) were the methods used in this study to generate landslide susceptibility indices. Before the calculation, the study area was randomly divided into two parts, the first for the establishment of the model and the second for its validation. The results of the landslide susceptibility analysis were verified using success and prediction rates. The MARSpline model gave a higher success rate (AUC (Area Under The Curve) = 0.963) and prediction rate (AUC = 0.951) than the LR model (AUC = 0.918 and AUC = 0.901) and the ANN model (AUC = 0.886

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 185–200

185

Benchelha, Aoudjehane, Hakdaoui, El Hamdouni, Mansouri, Benchelha, Layelmam, and Alaoui

Figure 1. Flow diagram showing the methodology used in this study.

and AUC = 0.877). These results indicate that the MARSpline model is the best model for determining landslide susceptibility in the study area. INTRODUCTION A landslide is the movement of a mass of rock, debris, or soil on a slope under the inﬂuence of gravity (Varnes, 1978; Guzzetti, 2005). In recent decades, landslides have received considerable attention because they are the most widespread disasters in the world in terms of loss of life and damage to social economies (Nefeslioglu et al., 2008; Shahabi et al., 2014). In Morocco, areas subject to landslides are mainly the Rif chain and, to a lesser extent, the Middle Atlas due to the existence of relatively young relief, which has very important dynamics compared to other regions. The dynamics associated with the formation of the Rif chain (from Alpine tectonics) were accompanied by instabilities mainly related to tectonic movements. The construction of major infrastructures (including roads and highways) is a triggering factor and promotes landslides. These landslides aﬀect populations, infrastructure, and the environment through injury or loss of life, economic losses, and disrupted habitats and corridors. For this study, ArcGIS and RStudio were used to construct and validate landslide susceptibility models us-

186

ing logistic regression, multivariate adaptive regression spline (MARSpline), and artificial neural network (ANN) models for the Oudka, Morocco, study area. A flow chart outlining the methodology is shown in Figure 1. For the application and verification of landslide susceptibility models, the study area was randomly divided into two parts: the first for the establishment of the model, and the second for its validation. In situ landslides were mapped in the study area using aerial photographs, geologic map data, and data obtained with field surveys using a global positioning system (GPS). Topographic, pedological, hydrological, and geological databases were included in the analysis. From these databases, eight factors were extracted: elevation, slope, aspect, distance to streams, distance to roads, distance to faults, lithology, and Normalized Difference Vegetation Index (NDVI). Landslide susceptibility models were created using logistic regression, MARSpline, and ANN in RStudio using the current landslide inventory and the eight factors. Finally, the analysis results were verified using the relative operating characteristic curve (ROC), including success and prediction rates (van Westen et al., 2003). STUDY AREA The study area is located in the northern region of Morocco, in the northwest area of Taounate

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 185–200

Landslide Susceptibility Mapping, North Morocco

Figure 2. Geographic location and distribution of landslides in Oudka commune, Taounate Prefecture, Morocco.

Province, because it is one of the areas most exposed to landslides in Morocco. The commune of Oudka is situated between the longitudes 4°42 11.40 W and 4°56 53.70 W and the latitudes 34°42 35.05 N and 34°42 52.43 N, and it covers a surface of 89 km2 (Figure 2). This area is a continuation of the Rif Cordillera and is characterized by mountainous terrain with no plains, except near the Aoulai River along the western boundary of the town. Jbel Oudka is considered to be the most important mountain of the Taounate Province, and its altitude reaches 1600 m. This mountain is characterized by signiﬁcant vegetation cover, including the Oudka Forest. In the commune of Oudka, olive trees occupy most of the arboreal surface area (92%). Olive is followed by the cultivation of ﬁg and other crops. Between 1977

and 2018, the Jbel Oudka station recorded an average annual rainfall of 1455 mm (data provided by the Hydraulic Basin Agency of Fes). The territory of the commune of Oudka is part of the producing area of the Ouergha watershed, which is crossed by several rivers of the oued Aoulai such as oued Elil, oued Elmaleh, and oued Assenou. There are several lakes in this territory, especially in the Oudka Forest. The most important of them is Afrat N’joum, which is located in the northern region of the Oudka and which has an area of 13,000 m2 . The average annual temperature in the region is between 15°C and 16°C. The average maximum temperature of the hottest month is approximately 34.2°C, and the average minimum of the coldest month is 0.5°C. The average extreme thermal amplitude is in the whole

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 185–200

187

Benchelha, Aoudjehane, Hakdaoui, El Hamdouni, Mansouri, Benchelha, Layelmam, and Alaoui

Figure 3. Overview of the landslides at the level of the Oudka commune: (a), (b), (c) landslide at douar Tissoufa; (d) landslide at the chaaba located east of douar de Tissoufa; (e) mass slide of limestone mixture into pebbles and marl; (f) landslide at RP 5302.

pre-Rifaine area between 30°C and 32°C, which corresponds to a semi-continental climate (data provided by the Hydraulic Basin Agency of Fes). DATABASE CONSTRUCTION Image Data Landsat OLI8 image data were downloaded from the U.S. Geological Survey web page and preprocessed by layer stacking of bands 2, 3, 4, 5, 6, and 7. Landsat imagery that was collected along the same satellite path were mosaiced into a single image. However, atmospheric correction was not necessary for images taken on the same calendar date (Song et al., 2001). Landslide Inventory Map Old landslide data were obtained from aerial photography, the database of geological maps, and ﬁeld surveys using GPS. Several studies have shown that the best calculation model is one in which the ratio of landslides to non-landslide points is equal to 1 (Bai et al., 2010). In total, 105 landslide polygons (Figure 3) and 43 randomly sampled polygons of stable surfaces mapped from diﬀerent sources were transformed into 8,911 cells with a resolution of 30 m for landslide areas and

188

9,005 cells for stable areas (i.e., without landslides). The 8,911 cells of the landslide grid and the 9,005 cells of the stable zone (without landslides) were randomly divided using RStudio into two subsets: Half of the cells of the grid were used for the realization of the landslide susceptibility model, while the other half was used for the validation of the model.

Factors In this study, we divided the conditioning factors (elevation, slope, aspect, distance to faults, distance to streams, distance to road, lithology, and NDVI) into five data sets, including topography, hydrology, land use, lithology, and human activity data sets. The landslide conditioning factors from these data sets were extracted from different sources and stored in the spatial database with a pixel size of 30 m. Topographic parameters, such as elevation, slope, and aspect, were derived directly from a digital elevation model (DEM). Hydrological data, including distance to streams, were derived indirectly from the DEM. The distance to road parameter reflects the influence of human activities. We used the geologic map to represent the lithology and distance to faults. The land-use data used in this study were represented by the NDVI derived from the Landsat OLI8 image of the study area.

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 185–200

Landslide Susceptibility Mapping, North Morocco

Elevation

Distance to Roads

Elevation is a determining factor, and it is a topographic parameter derived from the DEM, with resolution of 30 m. In our study area, the values of this factor vary between 192 m and 1600 m and can be divided into 10 classes: ࣘ350 m, 351–450 m, 451–550 m, 551–650 m, 651–750 m, 751–850 m, 851–950 m, 951– 1100 m, 1101–1300 m, and >1300 m (Figure 4a).

The distance to the road is considered to be one of the factors responsible for the occurrence of landslides and is frequently used for landslide susceptibility analysis (Yilmaz, 2010; Pourghasemi et al., 2012; and Nourani et al., 2014). The distribution of landslides along roads is very common, mainly because the natural state of the slope is damaged during the road construction process. In this study, distance from roads was taken into account to map landslide susceptibility and was classiﬁed into 10 classes (Figure 4f): ࣘ250 m, 251–500 m, 501–750 m, 751–1000 m, 1001–1250 m, 1251–1500 m, 1501–1750 m, 1751–2000 m, 2001–2250 m, and >2250 m.

Slope Slope is one of the key factors for slope stability and is considered to be one of the important factors in landslide susceptibility (Kanungo et al., 2006). This factor has been widely used in the literature (Kanungo et al., 2006; Lee et al., 2007; and Pourghasemi et al., 2012). In this study, the slope map was extracted from the DEM and divided into 10 classes: ࣘ5°, 5°–10°, 11°–15°, 16°–20°, 21°–25°, 26°–30°, 31°–35°, 36°–40°, 41°–45°, and >45° (Figure 4b).

Normalized Difference Vegetation Index (NDVI) NDVI is a measure of surface reflectance and provides a quantitative estimate of vegetation and biomass growth (Hall et al., 1995; Akgun et al., 2012). The NDVI highlights the difference between the visible red band and the near-infrared band according to the following formula (Eq. 1):

Aspect

NDVI = (IR − R) /IR + R),

The aspect is considered to be one of the main factors of landslide conditioning in landslide susceptibility. The aspect is frequently used in the assessment of landslide susceptibility (Kanungo et al., 2006; Lee et al., 2007). For this study, the slope aspect was classiﬁed into nine classes (Figure 4c): ﬂat (−1°), north (0°–22.5°, 337.5°–360°), northeast (22.5°– 67.5°), east (67.5°–112.5°), southeast (112.5°–157.5°), south (157.5°–202.5°), southwest (202.5°–247.5°), west (247.5°–292.5°), and northwest (292.5°–337.5°).

Distance to Faults The distance to faults was extracted from the geologic map of the study area at a scale of 1:50,000. The distance was calculated using ArcGIS software and classiﬁed into 10 classes (Figure 4d): ࣘ150, 151–300 m, 301–450 m, 451–600 m, 601–750 m, 751–900 m, 901– 1050 m, 1051–1200 m, 1201–1350 m, and >1350 m.

Distance to Streams The distance to streams is also a very important factor for landslide susceptibility analysis. In this area, the distance to the rivers was classiﬁed into 10 classes (Figure 4e): <50 m, 51–100 m, 101–150 m, 151–200 m, 201–250 m, 251–300 m, 301–350 m, 351–400 m, 401– 450 m, and >450 m.

(1)

where IR is the infrared band of the electromagnetic spectrum, and R is the red band of the electromagnetic spectrum. NDVI values range from −1 to +1, with negative values for areas other than plant cover, such as snow, water, or clouds, where red reflectance is greater than near infrared. For bare soils, the reflectances are approximately the same order of magnitude in the red and the near-infrared bands, and the NDVI has values close to 0. The vegetated areas have positive NDVI values, generally between 0.1 and 0.7. The highest values correspond to the densest vegetation cover. For this study, the NDVI value was reclassified into 10 categories (Figure 4g): ࣘ0.1, 0.11–0.15, 0.16–0.20, 0.21–0.25, 0.26–0.30, 0.31–0.35, 0.36–0.40, 0.41–0.45, 0.46–0.50, and >0.50. Lithology The occurrence of a landslide in the context of geomorphological studies is related to lithology (Pourghasemi et al., 2012), which is considered to be a very important conditioning factor because each lithological unit has a different influence on the susceptibility to landslides. In this study, the lithological map of Oudka was extracted from the Rhafsai geologic map at a scale of 1:50,000. The general lithologic setting of our study area is shown in Figure 4h. The resulting map

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 185–200

189

Benchelha, Aoudjehane, Hakdaoui, El Hamdouni, Mansouri, Benchelha, Layelmam, and Alaoui

Figure 4. Maps of some landslide conditioning factors: (a) elevation; (b) slope angle; (c) aspect; (d) distance to faults; (e) distance to streams; (f) distance to road; (g) NDVI; (h) lithology.

190

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 185â&#x20AC;&#x201C;200

Landslide Susceptibility Mapping, North Morocco Table 1. Frequency ratio values of the diﬀerent lithological components of the Oudka commune.

Type

Study Area (points)

Percent of Class

Landslide Points

Landslide Deposit Area per Class (%)

Landslide Frequency (%)

Alluvium, silt and marne Blue Marne Argillite and sandstone Marne, limestone marne and limestone Limestones and dolomites Red Marne

47,968 5,154 8,000 35,663 255 1,295

48.780 5.241 8.135 36.267 0.259 1.317

7,012 80 1 1,816 1 1

78.689 0.898 0.011 20.379 0.011 0.011

1.613 0.171 0.001 0.562 0.043 0.009

Class A B C D E F

contained categorical data that were transformed into numerical data to lighten the model (Bai et al., 2010). We applied the frequency ratio (FR), which is represented by the following formula (Eq. 2): FR =

N i=1

Di /Ai , N Di Ai

(2)

i=1

where Di is the area of a landslide of the ith category, Ai is the area of the ith category for a given parameter, and N is the category number of the parameter. The different lithological components of the Oudka commune and their FR values can be found in Table 1. LANDSLIDE SUSCEPTIBILITY MAPPING Logistic Regression Model The logistic regression (LR) model is a mathematical method that is used to establish the relationship between independent factors and landslides occurrence (or non-occurrence) (Bai et al., 2010; Das et al., 2010; and Nandi and Shakoor, 2010). The LR model is useful for predicting the presence or absence of a characteristic or outcome based on values of a set of predictor variables. Past studies compared LR to support vector machines, classification trees, and likelihood ratios and found that logistic regression was more accurate (Nandi and Shakoor, 2010; Atkinson and Massari, 2011; Shahabi et al., 2014; and Demir et al., 2015). An LR landslide susceptibility model predicts the probability of landslide occurrence. The predicted values range from 0 to 1 and can be defined by the following equation (Eq. 3): 1 1 + e−C(X ) when C (X ) = b0 + b1 X1 + b2 X2 + . . . + bn Xn , (3)

P (Y = 1/X ) =

where P is the probability of landslide occurrence (i.e., the landslide susceptibility index), C(X) is the linear logistic model, b0 is the intercept of the model, n is

the number of landslide-conditioning factors, bi is the weight of each factor, and xi is the landslide conditioning factor. In the case of the LR model, the multicollinearity should be checked. To quantify multicollinearity, there are several methods, such as Pearson’s correlation coefficients (Booth et al., 1994), variance decomposition proportions (Schuerman, 1983), conditional index (Belsley, 1991), variance inflation (VIF), and tolerances (Hair et al., 2009; Liao and Valliant, 2012). In our case, the variance inflation factor (VIF) was used.

Artificial Neural Network Model An ANN is a computational mechanism that can acquire, represent, and compute a map of information from one multivariate space to another (Garrett, 1994) using a data set representing that mapping. The purpose of an ANN is to build a model of the datagenerating process so that the network can generalize and predict outputs from inputs that it has not previously seen (Lee et al., 2007). The multi-layer perceptron (MLP) neural network, which has been described by Rumelhart and McClelland (1986), is one of the most widely used ANNs. The MLP network consists of three layers (input, hidden, and output layers), and it can identify relationships that are non-linear in nature (Pijanowski et al., 2002). The MLP networks are trained by error-correction learning, which means that the desired response of the system must be known; a back-propagation (BP) algorithm must also be known. The S-shaped sigmoid function is commonly used as the transfer function. The BP algorithm randomly selects the initial weights. Then, the difference between the expected and calculated output values across all observations is summarized using the mean-square error. After all observations are presented to the network, the weights are modified according to a generalized delta rule (Rumelhart and McClelland, 1986). This process of feeding forward signals and backpropagating errors is repeated iteratively until the error stabilizes at a low level (Pijanowski et al., 2002).

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 185–200

191

Benchelha, Aoudjehane, Hakdaoui, El Hamdouni, Mansouri, Benchelha, Layelmam, and Alaoui Table 2. The value of the weights and biases of the neural network model between the input layer and hidden layer 1. Weights between Input Layer and Hidden Layer 1 Input Layer Hidden Layer 1

Elevation

A1 A2 A3 A4

−1.790 0.875 −0.977 0.073

Slope −0.623 2.528 4.062 3.110

Distance to Fault −2.293 −0.175 1.533 0.077

Distance to Streams −2.676 −0.402 0.566 0.495

After integrating the training data (dependent and independent) in the MLP model using RStudio, a network architecture was constructed that consisted of eight neurons in the input layer, four and two neurons for hidden layers, and a neuron for the output layer (Figure 5). This model can identify non-linear relationships (Pijanowski et al., 2002). The transfer function used between neurons is a sigmoid function (Eq. 4):

Input: C (input) = −0.208 − 3.392 × B1 (output) +1.782 × B2 (output) ; Output: LSM = C (outputlayer) = C (landslide) = 1/ (1 + exp (−C (input))) . (4) The update of the initial weights is performed using a BP algorithm that minimizes the error between the predictive and calculated output values until the error stabilizes at a low level (Pijanowski et al., 2002). The ﬁnal weights and biases are shown in Tables 2 through 4.

Distance to Road −1.821 −0.370 0.004 −0.244

Aspect −0.655 0.009 0.739 0.220

Table 3. The value of the weights and biases of the neural network model between hidden layer 1 and hidden layer 2.

192

0.635 −33.848 79.93 15.01

ai fi (x),

(5)

where fi (x) is a basis function, n is the number of basis functions in the model, and f0 (x) is the constant basis function, the coeﬃcient of which is a0 . All of the coeﬃcients are calculated using ordinary least squares (OLS). The basis functions are represented by the following equation (Eq. 6): di s ji Xv ( j,i) − t ji ,

(6)

where di is the number of variables (interaction order) in the ith basis function Sji , Xv(j,i) is the vth variable, where 1 ࣘ v(j,i) ࣘ d, and tji is the knot location for each of the corresponding variables. The implementation of MARSpline was carried out with RStudio software. Table 4. The value of the weights and biases of the neural network model between hidden layer 2 and the output layer. Weights between Hidden Layer 2 and Output Layer

Hidden Layer 1

−0.246 −92.602 −1.414 99.254

−1.706 71.11 57.743 0.498

i=1

Weights between Hidden Layer 1 and Hidden Layer 2

B1 B2

−2.150 207.302 −60.442 −142.801

j=1

Multivariate adaptive regression splines (MARSpline) is a form of regression analysis introduced by Friedman (1991). MARSpline is a non-parametric regression technique that can be seen as an extension of

Bias

F (x) = a0 +

MARSpline Model

NDVI

linear models that automatically model non-linearities and interactions between variables. The MARSpline approach to regression modeling effectively uncovers important data patterns and relationships that are difficult, if not impossible, for other regression methods to reveal. The model is automatically determined by the data through a forward/backward iterative approach (Felicísimo et al., 2013). MARSpline estimates the function through a set of adaptive piecewise linear regressions called the “basis functions”. The MARSpline model can be defined as a sum of basis functions (Eq. 5):

fi (x) =

Hidden Layer 2

Geology

Bias

8.177 −181.785

287.988 −304.666

10.879 −12.698

Output Layer C

Hidden Layer 2 B1

Bias

−3.392

1.782

0.208

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 185–200

Landslide Susceptibility Mapping, North Morocco Table 5. Multicollinearity diagnosis indexes for independent variables. Factors

VIF

Elevation Slope Dist_fault Dist_streams Dist_road Aspect Geology NDVI

2.0219 1.1242 1.1312 1.0983 1.1296 1.1667 1.1745 1.8110

Results of Landslide Susceptibility Models Models were run using the database containing a dependent variable (landslide) and the eight independent variables (altitude, slope, aspect, distance to faults, distance to streams, distance to the road, NDVI, and lithology). The resulting values of VIF, as shown in Table 5, were all less than 4, indicating that there was no collinearity problem to explore. In cases where landslide factors had a VIF value greater than 4, these factors were not applied to the LR model. The Hosmer and Lemeshow test (a test to evaluate the effectiveness of the training data sets) showed that the LR model was statically significant. The variable coefficients for the LR model are listed in Table 6. From Table 6, the distance to fault, distance to streams, and NDVI were negatively related to landslide susceptibility; that is, when the values of these factors increased, the susceptibility to landslides decreased. Figure 6 shows the landslide susceptibility map based on the LR model. The area of very high susceptibility is the area where there is a high landslide probability (P > 0.8). Compared to statistical methods, the neural networks make it possible to define classes by taking into account their distribution in the corresponding domain of each data source (Zhou, 1999). After completion of the training phase and after reaching the network objective, the data from the study area were introduced into the network to estimate Table 6. Coefficients of the LR model. Parameters Elevation Slope Distance to fault Distance to streams Distance to road Aspect Geology NDVI Constant

Table 7. Coeﬃcients and basis functions of the MARSpline model. MARSpline Model Basis Functions BF1 max(0,Elevation-265) BF2 max(0,373-Elevation) BF3 max(0,Elevation-373) BF4 max(0,Elevation-801) BF5 max(0,429.534637-Dist_fault) BF6 max(0,295.465729-Dist_streams) BF7 max(0,Dist_streams-295.465729) BF8 max(0,582.494629-Dist_road) BF9 max(0,Dist_road-582.494629) BF10 max(0,1.607985-Geology) BF11 max(0,0.192658-NDVI) BF12 max(0,NDVI-0.192658) BF13 max(0,NDVI-0.25172) BF14 max(0,NDVI-0.312795) Constant

MARSpline Model Coeﬃcients − 0.213647 − 0.204117 0.213144 0.007225 0.002557 0.004196 − 0.028934 − 0.003606 − 0.001306 − 13.387262 − 30.793324 4.826796 − 47.973135 47.533463 24.812086

landslide susceptibility. After obtaining susceptibility values, the ﬁnal map of landslide susceptibility was produced (Figure 7). The area of very high susceptibility is the area where there is a high landslide probability (P > 0.8). For the MARSpline model, the implementation was performed with RStudio based on the training database and the 14 basic functions (Table 7). The equation for the probability of landslide presence (Y) using the MARSpline model is: Y = 24.812 − (0.214 × BF1) − (0.204 × BF2) + (0.213 × BF3) + (0.007 × BF4) + (0.003 × BF5) + (0.004 × BF6) − (0.029 × BF7) − (0.004 × BF8) − (0.001 × BF9) − (13.387 × BF10) − (30.793 × BF11) + (4.827 × BF12) − (47.973 × BF13) + (47.533 × BF14) . (7) The resulting landslide susceptibility map is shown in Figure 8. The area of very high susceptibility is the area where there is a high landslide probability (P > 0.8).

LR Model Coeﬃcients

VALIDATION AND COMPARISON

0.003 0.004 − 0.002 − 0.005 0.0002 0.005 3 − 2.314 − 6.002

The landslide susceptibility maps resulting from the application of the diﬀerent statistical models (logistic regression, ANN, and MARSpline models) were divided into ﬁve classes. The accuracy of these landslide susceptibility maps was evaluated by calculating the relative operating characteristic (ROC) and the percentage of landslide points observed in various susceptibility categories (Nandi and Shakoor, 2010).

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 185–200

193

Benchelha, Aoudjehane, Hakdaoui, El Hamdouni, Mansouri, Benchelha, Layelmam, and Alaoui

Figure 5. Neural network architecture.

The area under the ROC curve (AUC) represents the quality of the probabilistic model (i.e., its ability to predict the occurrence or not of an event) (Yesilnacar and Topal, 2005). The ideal model shows an AUC value close to 1.0, whereas a value close to 0.5 indicates an inaccurate

model (Fawcett, 2006; Nandi and Shakoor, 2010; and Akgun, 2012) or no diagnostic accuracy (Rosner et al., 2015) or poor discrimination (Hanley and McNeil, 1982; Rosman and Korsten, 2007). In this study, all landslide susceptibility models were validated using success rate and prediction rate

Figure 6. Landslide susceptibility map produced from the LR model.

194

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 185â&#x20AC;&#x201C;200

Landslide Susceptibility Mapping, North Morocco

Figure 7. Landslide susceptibility map produced from the ANN model.

Figure 8. Landslide susceptibility map produced from MARSpline model.

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 185â&#x20AC;&#x201C;200

195

Benchelha, Aoudjehane, Hakdaoui, El Hamdouni, Mansouri, Benchelha, Layelmam, and Alaoui

Figure 9. ROC curve evaluation of the LR, ANN, and MARSpline models: (a) success rate curves and (b) prediction rate curves.

methods. The success rate results were obtained by comparing landslide susceptibility maps with landslides in the training data set, while the prediction rate results for the susceptibility models were evaluated using the validation data set independent of that used in the landslide model construction process and using RStudio. The ROC curves of this study are shown in Figure 9. The AUC values obtained from the susceptibility maps show that the MARSpline model gave the highest success rate (AUC = 0.963) and the best prediction rates (AUC = 0.951). The results of the LR model showed success and prediction rates with values of AUC = 0.918 and AUC = 0.901. The application of the ANN model yielded the lowest success and prediction rates, which were AUC = 0.886 and AUC = 0.877, respectively. The landslide susceptibility maps were verified by landslides covering 4,463 pixels of the municipality Oudka. These landslides were not used in the construction of the models. The landslide susceptibility maps of the three models were divided into five categories (Figure 10): very low (0 < LSI (Landslide Susceptibility Index) ࣘ 0.2), low (0.2 < LSI ࣘ 0.4), medium (0.4 < LSI ࣘ 0.6), high (0.6 < LSI ࣘ 0.8) and very high (LSI > 0.8). The comparison between the verification landslides (4,463 pixels) and the landslide susceptibility maps resulting from the LR, ANN and MARSpline models allowed us to determine the percentages of test landslide points falling into different susceptibility categories (Figure 10). In the very low susceptibility class, we found just 1% of the observed landslides for the LR and MARSpline methods and 3% for the ANN

196

method. In the very high susceptibility class, we found 56%, 68%, and 77% of the observed landslides for the LR, ANN, and MARSpline methods, respectively. However, in the high and very high susceptibility classes, we found 85%, 68%, and 91% of the observed landslides with the LR, ANN, and MARSpline methods, respectively. By comparing the results of the LR, ANN, and MARSpline analyses, we determined that the MARSpline and LR methods were better than the ANN method. The MARSpline method is the best approach for the assessment of landslide susceptibility for the Oudka commune. Results and Discussion of the MARSpline Model The analysis of the landslide susceptibility map based on the MARSpline statistical model in relation to the diﬀerent factors allowed us to obtain the following results. Elevation The landslides are mostly concentrated with a high frequency between elevations of 851 m and 1300 m (intermediate elevation; Figure 11a). This may be due to:

r The terrain in this area is dominated by marls and pelites with poor geomechanical characteristics (angle of friction and cohesion). r The marls constitute the base of the water table contained in the limestone and the sandstone with an elevation higher than 1300 m. This induces saturation and loading of the ground material.

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 185–200

Landslide Susceptibility Mapping, North Morocco

Figure 10. Percentages of test landslide points falling into diﬀerent susceptibility categories using LR, ANN, and MARSpline.

r The water channeled by faults creates additional

Slope Analysis of the slope graph (Figure 11b) shows that landslides are concentrated mainly between 5° and 20°. Actually, pressiometric drilling carried out in situ (Figure 12) showed low values of the limit pressure parameter in a soil depth of 5 to 10 m.

loading and increases eﬀective stress along the slope.

r The material in the fault can slide and the upstream soil remains without lower support (low slope side).

r The hazard as a function of the distance to faults has a logarithmic form. Distance to Streams

Aspect The area of Oudka has south-, southwest-, southeast-, and west-oriented slopes with respective percentages of 23%, 20%, 20%, and 14%, and these are consistent with the analysis of Figure 11c, which shows that most landslides are oriented south, southeast, southwest, and west. Note that the north-facing slope was not part of the study area. Distance to Faults The occurrence of landslides increases considerably when approaching the faults (Figure 11d) (with a very strong hazard when the distance to a fault is lower than 450 m) and then attenuates gradually when the distance is greater than 600 m. In fact, closer to the fault, the following concomitant eﬀects contribute to the occurrence of landslides:

r The zones of faults are formed of crushed material, thus losing their initial structures and their cohesion.

The occurrence of landslides increases as the distance to stream channels decreases (Figure 11e). The influx of water induces the loading of the ground and the suppression of upstream land abutment, similar to the faults. Normalized Difference Vegetation Index Landslides occur when the NDVI is between 0.15 an 0.30. This corresponds mainly to the central part of the Oudka Massif (Figure 11f), and revegetation contributes to the fixation of the sliding surfaces for a NDVI greater than 0.30 (high density). Lithology Most of the landslides are on class A (alluvium, silt, and marl) (Table 1) soils, since the surface indexes constituting the database are covered by this facies. We find that mainly, in addition to the Quaternary sediments, it is the marls and weathered pelites that trigger the landslides, and these are grouped into the remodeled lands of class D (Figure 11g).

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 185–200

197

Benchelha, Aoudjehane, Hakdaoui, El Hamdouni, Mansouri, Benchelha, Layelmam, and Alaoui

Figure 11. Percentages of very high susceptibility values in the susceptibility map based on results from the MARSpline model for: (a) elevation; (b) slope; (c) aspect; (d) distance to faults; (e) distance to streams; (f) NDVI; (g) lithology.

CONCLUSIONS The demolition of ﬁve buildings by the Tissoufa landslide, northern Oudka, caused by heavy rainfall in 2013 highlighted the need for the

198

establishment of a landslide susceptibility map in the Oudka commune in the northern region of Morocco. Diﬀerent researchers have proposed various methodologies for landslide susceptibility mapping.

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 185–200

Landslide Susceptibility Mapping, North Morocco

Figure 12. Schemas illustrating: (a) locations of the pressure meter tests, and (b) results of the pressure meter tests.

In this study, logistic regression, ANN, and MARSpline methods were applied to landslide susceptibility mapping in the Oudka commune. This study included three main stages: (1) landslide inventory, (2) analysis, and (3) evaluation of the susceptibility map. The accuracy of each landslide susceptibility map was evaluated by calculating the relative operating characteristic (ROC). The results showed that the MARSpline model had a higher success rate (AUC = 0.963) and prediction rate (AUC = 0.951) than the LR and ANN models. The calculation of the percentages of test landslide points using the LR, ANN, and MARSpline models showed that the results obtained for the two high and very high susceptibility classes (combined) were 85%, 68%, and 91%, respectively. Both results indicate that the map obtained by the MARSpline method is better than maps from the LR and the ANN methods. The results presented in this study can assist developers, planners, and engineers in slope management and land-use planning. ACKNOWLEDGMENTS This work was made possible thanks to the support of Pr. M. Lhaya, president of the Moulay Rachid Arrondissement, Mr. F. Mounaji, director of services of the Moulay Rachid Arrondissement, and all functionary and elected oﬃcials of the Moulay Rachid Arrondissement, and in collaboration with Taounate’s PDELTW (Provincial Directorate of Equipment, Logistics, Transport and Water), the Moroccan Public Laboratory for Testing and Studies (LPEE), and the Hydraulic Basin Agency of Fes.

REFERENCES Akgun, A., 2012, A comparison of landslide susceptibility maps produced by logistic regression, multi-criteria decision, and likelihood ratio methods: A case study at İzmir, Turkey: Landslides, Vol. 9, No. 1, pp. 93–106. https://doi.org/10.1007/s10346-011-0283-7. Akgun, A.; Sezer, E. A.; Nefeslioglu, H. A.; Gokceoglu, C.; and Pradhan, B., 2012, An easy-to-use MATLAB program (MamLand) for the assessment of landslide susceptibility using a Mamdani fuzzy algorithm: Computers & Geosciences, Vol. 38, No. 1, pp. 23–34. https://doi.org/10.1016/j.cageo.2011.04.012. Atkinson, P. M. and Massari, R., 2011, Autologistic modelling of susceptibility to landsliding in the Central Apennines, Italy: Geomorphology, Scale Issues in Geomorphology, Vol. 130, No. 1, pp. 55–64. https://doi.org/10.1016/j.geomorph.2011.02.001. Bai, Shi-Biao; Jian Wang; Guo-Nian Lü; Ping-Gen Zhou; Sheng-Shan Hou; and Su-Ning Xu, 2010, GIS-based logistic regression for landslide susceptibility mapping of the Zhongxian segment in the Three Gorges area, China: Geomorphology, Vol. 115, No. 1, pp. 23–31. https://doi.org/10.1016/j.geomorph.2009.09.025. Belsley, D. A., 1991, A guide to using the collinearity diagnostics: Computer Science in Economics and Management, Vol. 4, No. 1, pp. 33–50. https://doi.org/10.1007/BF00426854. Booth, G. D.; Niccolucci, M. J.; and Schuster, E. G., 1994, Identifying Proxy Sets in Multiple Linear Regression: An Aid to Better Coeﬃcient Interpretation: U.S. Department of Agriculture Forest Service Research Paper INT-470. Das, I.; Sahoo, S.; van Westen, C.; Stein, A.; and Hack, R., 2010, Landslide susceptibility assessment using logistic regression and its comparison with a rock mass classiﬁcation system, along a road section in the northern Himalayas (India): Geomorphology, Vol. 114, No. 4, pp. 627–637. https://doi.org/10.1016/j.geomorph.2009.09.023. Demir, G.; Aytekin, M.; and Akgun, A., 2015, Landslide susceptibility mapping by frequency ratio and logistic regression methods: An example from Niksar–Resadiye (Tokat, Turkey):

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 185–200

199

Benchelha, Aoudjehane, Hakdaoui, El Hamdouni, Mansouri, Benchelha, Layelmam, and Alaoui Arabian Journal of Geosciences, Vol. 8, No. 3, pp. 1801–1812. https://doi.org/10.1007/s12517-014-1332-z. Fawcett, T., 2006, An introduction to ROC analysis. Pattern Recognition Letters, Vol. 27, No. 8, pp. 861–874. https://doi.org/10.1016/j.patrec.2005.10.010. Felicísimo, Á. M., Cuartero, A.; Remondo, J.; and Quiró, E., 2013, Mapping landslide susceptibility with logistic regression, multiple adaptive regression splines, classification and regression trees, and maximum entropy methods: A comparative study: Landslides, Vol. 10, No. 2, pp. 175–189. https://doi.org/10.1007/s10346-012-0320-1. Friedman, J. H., 1991, Multivariate adaptive regression splines: The Annals of Statistics, Vol. 19, No. 1, pp. 1–67. Garrett, J. H., 1994, Where and Why Artificial Neural Networks Are Applicable in Civil Engineering: Journal of Computing in Civil Engineering: 8 (2), pp. 129–130. http://dx.doi.org/10.1061/(ASCE)0887-3801(1994)8:2(129) Guzzetti, F., 2005, Landslide Hazard and Risk Assessment: Ph.D. Thesis, University of Bonn, Bonn, Germany, 371 p. Hall, D. K.; Riggs, G. A.; and Salomonson, V. V., 1995, Development of methods for mapping global snow cover using moderate resolution imaging spectroradiometer data: Remote Sensing of Environment, Vol. 54, No. 2, pp. 127–140. https://doi.org/10.1016/0034-4257(95)00137-P. Hair, J. F.; Black, W. C.; Babin, B. J.; and Anderson, R. E., 2009, Multivariate Data Analysis: Prentice Hall, New York. Hanley, J. A. and McNeil, B. J., 1982, The meaning and use of the area under a receiver operating characteristic (ROC) curve: Radiology, Vol. 143, No. 1, pp. 29–36. Kanungo, D. P.; Arora, M. K.; Sarkar, S.; and Gupta, R. P., 2006, A comparative study of conventional, ANN black box, fuzzy and combined neural and fuzzy weighting procedures for landslide susceptibility zonation in Darjeeling Himalayas: Engineering Geology, Vol. 85, No. 3, pp. 347–366. https://doi.org/10.1016/j.enggeo.2006.03.004. Lee, S., Ryu, J.-H.; and Kim, I.-S., 2007, Landslide susceptibility analysis and its verification using likelihood ratio, logistic regression, and artificial neural network models: Case study of Youngin, Korea: Landslides, Vol. 4, No. 4, pp. 327–338. https://doi.org/10.1007/s10346-007-0088-x. Liao, D. and Valliant, R., 2012, Variance inflation factors in the analysis of complex survey data: Survey Methodology, Vol. 38, pp. 53–62. Nandi, A. and Shakoor, A., 2010, A GIS-based landslide susceptibility evaluation using bivariate and multivariate statistical analyses: Engineering Geology, Vol. 110, No. 1, pp. 11–20. https://doi.org/10.1016/j.enggeo.2009.10.001. Nefeslioglu, H. A.; Gokceoglu, C.; and Sonmez, H., 2008, An assessment on the use of logistic regression and artificial neural networks with different sampling strategies for the preparation of landslide susceptibility maps: Engineering Geology, Vol. 97, No. 3–4, pp. 171–191. Nourani, V.; Pradhan, B.; Ghaffari, H.; and Sharifi, S. S., 2014, Landslide susceptibility mapping at Zonouz Plain, Iran, using genetic programming and comparison with frequency ratio, logistic regression, and artificial neural network models: Natural Hazards, Vol. 71, No. 1, pp. 523–547. https://doi.org/10.1007/s11069-013-0932-3.

200

Pijanowski, B. C.; Brown, D. G.; Shellito, B. A.; and Manik, G. A., 2002, Using neural networks and GIS to forecast land use changes: A land transformation model: Computers, Environment and Urban Systems, Vol. 26, No. 6, pp. 553–575. https://doi.org/10.1016/S0198-9715(01)00015-1. Pourghasemi, H. R.; Pradhan, B.; and Gokceoglu, C., 2012, Application of fuzzy logic and analytical hierarchy process (AHP) to landslide susceptibility mapping at Haraz Watershed, Iran: Natural Hazards, Vol. 63, No. 2, pp. 965–996. https://doi.org/10.1007/s11069-012-0217-2. Rosman, A. S. and Korsten, M. D., 2007, Metaanalysis comparing CT colonography, air contrast barium enema, and colonoscopy: The American Journal of Medicine, Vol. 120, No. 3, pp. 203–210. https://doi.org/10.1016/j.amjmed.2006.05.061. Rosner, B.; Tworoger, S.; and Qiu, W., 2015, Correcting AUC for measurement error: Journal of Biometrics & Biostatistics, Vol. 6, No. 5, pp. 270. https://doi.org/10.4172/21556180.1000270. Rumelhart, D. E. and McClelland, J. L., 1986, Parallel Distributed Processing: Explorations in the Microstructure of Cognition, Vol. 1: MIT Press, Cambridge, MA. https://www.osti.gov/biblio/5838709. Schuerman, J. R., 1983, Principal Components Analysis. In: Multivariate Analysis in the Human Services. International Series in Social Welfare, Vol. 2. Springer, Dordrecht. pp. 93–119. DOI: https://doi.org/10.1007/978-94-009-6661-1_6. Shahabi, H.; Khezri, S.; Bin Ahmad, B.; and Hashim, M., 2014, Landslide susceptibility mapping at central Zab Basin, Iran: A comparison between analytical hierarchy process, frequency ratio and logistic regression models: Catena, Vol. 115, pp. 55– 70. https://doi.org/10.1016/j.catena.2013.11.014. Song, C.; Woodcock, C. E.; Seto, K. C.; Lenney, M. P.; and Macomber, S. A., 2001, Classification and change detection using Landsat TM data: When and how to correct atmospheric effects?: Remote Sensing of Environment, Vol. 75, No. 2, pp. 230–244. van Westen, C. J.; Rengers, N.. and Soeters, R., 2003, Use of geomorphological information in indirect landslide susceptibility assessment: Natural Hazards, Vol. 30, No. 3, pp. 399–419. https://doi.org/10.1023/B:NHAZ.0000007097. 42735.9e. Varnes, D. J., 1978, Slope Movement Types and Processes: Transportation Research Board Special Report 176, Transportation Research Board, Washington, D.C., pp. 11–33. Yesilnacar, E. and Topal, T., 2005, Landslide susceptibility mapping: A comparison of logistic regression and neural networks methods in a medium scale study, Hendek Region (Turkey): Engineering Geology, Vol. 79, No. 3, pp. 251–266. https://doi.org/10.1016/j.enggeo.2005.02.002. Yilmaz, I., 2010, Comparison of landslide susceptibility mapping methodologies for Koyulhisar, Turkey: Conditional probability, logistic regression, artificial neural networks, and support vector machine: Environmental Earth Sciences, Vol. 61, No. 4, pp. 821–836. https://doi.org/10.1007/s12665-009-0394-9. Zhou, W., 1999, Verification of the nonparametric characteristics of backpropagation neural networks for image classification: IEEE Transactions on Geoscience and Remote Sensing, Vol. 37, No. 2, pp. 771–779. https://doi.org/10.1109/36.752193.

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 185–200

Characterization and Dynamic Analysis of the Devils Castle Rock Avalanche, Alta, Utah PATRICIA PEDERSEN JEFFREY R. MOORE* BRENDON J. QUIRK University of Utah, Department of Geology and Geophysics, 115 S 1460 E, Salt Lake City, UT 84112

RICHARD E. GIRAUD GREG N. MCDONALD Utah Geological Survey, 594 West North Temple, P.O. Box 146100, Salt Lake City, UT 84114

Key Terms: Rock Avalanche, Landslide Runout Simulation, Cosmogenic Exposure Dating, Coseismic Mass Wasting, Wasatch Mountains

allow us to better understand the processes and controls of large-scale mass movements in the region.

ABSTRACT

INTRODUCTION

Rock avalanches are large-magnitude mass movements with high mobility and fluid-like runout; however, because of their scarcity, little information is typically available to describe the hazard posed by these events. Geologic records thus provide key data regarding rock avalanche size, timing, and dynamics. Here we present a detailed case history analysis of the Devils Castle rock avalanche located near the town of Alta in the Wasatch Mountains of Utah. The deposit is ∼1.5 km in length with a Fahrboeschung angle of 14 degrees (heightto-length ratio = 0.25). Through topographic reconstruction, we calculated a deposit volume of 1.7 million m3 with a maximum thickness of 25 m and an average thickness of 7 m. Cosmogenic surface exposure dating of six deposit boulders indicates a failure age of 14.4 ± 1.0 ka. The Devils Castle headwall displays no obvious evidence indicating precise source location and geometry; therefore, we reconstructed two plausible source volumes and performed numerical runout simulations for each. Results agree well with mapped deposit boundaries for both source scenarios; however, the east source model better represents material and dynamic characteristics of the deposit observed in the field. While the region is seismically active, the Late Pleistocene age for the rock avalanche precludes ascribing direct correlation with any currently known surface-rupturing paleoearthquakes. We identified and describe five similar events in the region highlighting the extent of the potential hazard. Individual case history analyses such as this

With rapid and ﬂow-like runout, rock avalanches represent a high-magnitude, low-frequency landslide hazard (e.g., Dunning et al., 2007; Ivy-Ochs et al., 2009; Wolter et al., 2015; Aaron and Hungr, 2016; Coe et al., 2016; and Moore et al., 2017). While these mass wasting events tend to be rare, they can have severe consequences, making insights into their behavior crucial in areas where the potential for loss of life or property is large (Christenson and Ashland, 2006; Willenberg et al., 2009; Castleton et al., 2016; and Loew et al., 2017). Generating and compiling geologic records of prehistoric rock avalanche case histories thus provides critical data supporting modern hazard and risk assessments, including event parameters such as volume, age, and failure dynamics (e.g., Douglass et al., 2005; Prager et al., 2008; Crosta et al., 2017; IvyOchs et al., 2017; and Köpﬂi et al., 2018). Geologic case histories provide input to calibrate numerical rock avalanche runout models that can be used in support of hazard assessments to predict behavior in future events and are especially critical in areas with no historical data. Rock avalanche runout behavior depends on the source and runout path materials and substrate, topography, and additional effects, such as entrainment of water (Hungr and Evans, 2004; Dufresne et al., 2010; and Aaron et al., 2017). Calibrated numerical runout models in turn provide a basis for quantitative prediction of rock avalanche reach and impact velocities, supporting risk assessment for similar event scenarios (Hungr and Evans, 1996; Hungr and McDougall, 2009; and Aaron and Hungr, 2016). Additionally critical for evaluating rock

*Corresponding author email: jeﬀ.moore@utah.edu

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 201–215

201

Pedersen, Moore, Quirk, Giraud, and McDonald

avalanche hazard is quantifying the age of infrequent events and ultimately, with suﬃcient data, establishing recurrence intervals and magnitude-frequency relations (e.g., Hungr et al., 1999). Preparation and triggering of large rock slope failures involve time-dependent damage processes often linked to the morphological development of alpine landscapes (Stock and Uhrhammer, 2010; Ballantyne et al., 2014; and Grämiger et al., 2016). Glacial erosion in particular undercuts valley rock walls (Sanders et al., 2012), and while the eﬃcacy of glacial debuttressing has recently been called into question (McColl et al., 2010; McColl, 2012), rock debuttressing through glacial undercutting has been shown to be an important factor generating rock mass damage and conditioning future slope failures (Grämiger et al., 2017). Ultimate triggering factors for rock avalanches may include earthquakes and heavy precipitation or snowmelt (Stock and Uhrhammer, 2010; Coe et al., 2016), although many historical events have occurred without a recognizable trigger (e.g., Lipovsky et al., 2008; Coe et al., 2017; and Moore et al., 2017). The Wasatch Range has a long record of seismic activity (DuRoss, 2008) that may contribute to the development and triggering of proximal rock slope failures. Detailed past studies from paleoseismic trench sites along the Salt Lake City segment of the Wasatch Fault reveal an average late Holocene recurrence interval of ∼1,350 years for surface-rupturing earthquakes (McCalpin, 2002; DuRoss, 2008), and earthquake hazard scenarios have been developed due to the large population residing in the seismically active region (Pankow et al., 2015). Meanwhile, several large rock avalanche deposits are known to exist along the Wasatch Front (Hooper, 1951; Cardoso, 2002; and Ashland and McDonald, 2008), and detailed study of these landslides may be of critical importance for seismic hazard scenarios if they are thought to be mainly coseismic in origin. Here we analyze the Devils Castle rock avalanche located in the Albion Basin near Salt Lake City, Utah. We begin by describing the rock avalanche deposit extents, lithology, and surrounding Quaternary landforms. We then reconstruct the pre-slide topography to estimate the deposit volume and project that material onto the headwall in two likely source areas. We use these topographic reconstructions as the basis for three-dimensional runout modeling, which provides insights on the landslide mobility and dynamics of the failure. We determine the age of the Devils Castle rock avalanche from 36 Cl cosmogenic surface exposure dating of deposit boulders. Finally, we discuss possible preparatory and triggering factors and describe key details of other, similar events discovered in the region.

202

STUDY AREA The Devils Castle rock avalanche is located in the Albion Basin at the head of Little Cottonwood Canyon in the Wasatch Mountains near Salt Lake City. The area is an alpine basin with a ﬂoor at approximately 2,800 m above sea level (a.s.l.) and the summit of Devils Castle reaching 3,330 m a.s.l. The rock avalanche released from the headwall of Devils Castle and cascaded into a north-trending, low-relief strike gully formed between two east-dipping quartzite ridges. The Albion Basin and neighboring town of Alta are highly utilized recreation areas during all seasons. With a network of hiking trails, visitation in the summer can be upwards of 10,000 people per week (David Evans and Associates, 2011), with the most heavily accessed trail passing directly over the rock avalanche deposit. Also located in the area is the Uinta-WasatchCache National Forest Albion Basin campground, as well as several privately owned cabins (Figure 1a). The Alta Ski Area maintains multiple ski runs below Devils Castle, and a restaurant lies directly at the toe of the rock avalanche deposit. Little Cottonwood Canyon was extensively glaciated during the last glacial period (locally termed Pinedale Glaciation) (Laabs et al., 2011; Quirk et al., 2018). Valley glaciers retreated following the Last Glacial Maximum at ∼22 ka, leaving a large lateral moraine and several recessional features at the mouth of the canyon (Laabs and Munroe, 2016). Retreat up the canyon was likely rapid, with evidence from neighboring Big Cottonwood Canyon suggesting that high cirque basins were deglaciated by ∼15 ka (Quirk et al., 2018). In the Albion Basin, radiocarbon evidence suggests that glaciers likely disappeared by the start of the Holocene (and potentially earlier), with a small successive re-advance proximal to the Devils Castle headwall constrained by a calibrated radiocarbon age inside nested moraines of ∼8.4 ky cal. BP (Madsen and Currey, 1979; Quirk et al., 2018). Geological Mapping We generated a detailed geologic map (Figure 2) providing a description of key deposit characteristics, including lithology, boundaries, and Quaternary geomorphological features. We also mapped the Devils Castle headwall, where we identiﬁed two possible source areas for the rock avalanche (referred to as the “east source” and the “west source”: in the text; see Figure 1a for locations). We utilized available resources, including aerial photography, previously published geological maps (Calkins and Butler, 1943; Richmond, 1964; and Baker et al., 1966), and

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 201–215

Devils Castle Rock Avalanche

Figure 1. Overview of the study area. (a) Mapped rock avalanche deposit in Albion Basin located near Alta, Utah, approximately 30 km southeast of Salt Lake City. For clarity, only the main continuous deposit body is shown. Photo locations indicated. (b) Limestone boulder transported by the rock avalanche. (c) View of Devils Castle looking south. The rock avalanche originated from this headwall, likely from the east source area. (d) Typical deposit character. The deposit is found largely in wooded areas. View is to the southeast.

2-m bare-earth lidar (Utah Automated Geographic Reference Center, 2006) to aid field mapping. Important Quaternary features observed in the field surrounding the rock avalanche deposit were also incorporated. The two main lithologies comprising the Devils Castle headwall are the Mississippian Fitchville Formation and Deseret/Gardison Limestone (Figure 2) (Baker et al., 1966). The Fitchville Formation consists of parallel and persistent cliff-forming dolomite beds with few joints and good rock mass quality. The Deseret/Gardison Limestone, on the other hand, is more fractured and weathered, forming short cliffs and intervening ledges. Bedding dips ∼25º–30º to the east, not favoring sliding or toppling kinematic fail-

ure modes in either possible source area. A fault cuts through the cliﬀ face, forcing the Fitchville Formation to abut the Deseret/Gardison Limestone through an up-thrown block (Figure 2) (Baker et al., 1966). Comparing lithologies of the east and the west source areas of the headwall (Figure 2), we note that the east source consists mainly of the Deseret/Gardison Limestone, whereas the west source has an outcrop of Fitchville Formation in the lower part of the cliﬀ. The rock avalanche deposit (Qls; Figure 2) consists mainly of Deseret/Gardison Limestone with large boulders measuring up to 6 m in height (Figure 1b). Blocks are distributed continuously throughout the deposit and are typically weathered, often appearing broken and crumbled (Figure 1d). Boulders of the

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 201–215

203

Pedersen, Moore, Quirk, Giraud, and McDonald

Figure 2. Geologic map of Albion Basin (adapted from Baker et al., 1966). Bedrock and some glacial units adapted; other surficial units mapped by the authors. The rock avalanche originated from Mississippian limestone and partially covered existing glacial deposits. Later glacial and periglacial deposits covered the upper portion of the slide. Note the thin lateral deposits east of the main body. Rock units include undifferentiated talus (Qtu), glacial deposits (Qm); periglacial deposits (Qmp); landslide deposits (Qls); landslide deposit, thin layer (Qlt); glacial deposits, undifferentiated till (Qmu); Alta Stock (Tag); Deseret and Gardison Limestone, undivided (Mdg); Fitchville Formation (Mf); Maxfield Limestone (Cm); Ophir Formation (Co); Tintic Quartzite (Ct); and Mineral Fork Tillite (pCmf).

Fitchville Formation tend to be more massive. These boulders are rarely found within the main deposit but occur frequently within the rock glacier feature that overrides the rock avalanche. Very large boulders tend to be located toward the deposit boundaries. The rock avalanche deposit and its perimeter were easily identified in most places by boulder concentration during field inspection. However, above the head of the deposit, rock avalanche debris has been overrun by glacial and periglacial landforms as well as talus, while the toe has been modified by construction of a restaurant. Features of the rock avalanche deposit are controlled by topography. Erosion-resistant bedrock outcrops throughout the basin create parallel strikegullies, and the deposit is located within one such gully. The observable length of the rock avalanche deposit is approximately 1 km (Figure 1a). From topographic

204

reconstruction (detailed in a later section), we estimate the original length of the deposit to be approximately 1.5 km. The landslide width is at maximum 195 m at its head and reduces to 80 m (Figure 1a) where the bedrock gully narrows. At the toe of the rock avalanche, a dense concentration of large boulders is observed that terminates abruptly before the ski area restaurant. We did not identify signs of the deposit downstream from the restaurant during field inspection or from aerial photographs pre-dating the restaurant’s construction. Periglacial and glacial landforms cover the head of the deposit, with the uppermost glacial deposits consisting of a set of nested moraines with several smaller adjacent outboard moraines (Qm; Figure 2). Periglacial deposits are interpreted as rock glaciers, the largest being an apparently inactive feature approximately 250 m in width (Qmp; Figure 2). Because these deposits are often comingled without obvious boundaries, we show their combined extent in Figure 2, indicating crests where field evidence was clear. Undifferentiated talus (Qtu) at the base of the headwall consists of steep block fragments representing episodic rock wall retreat. Along the east side of the main continuous rock avalanche deposit, a smaller area of boulders was discovered extending from the base of the nested moraines (unit Qlt; Figure 2), which we estimate averages at most 2 m thick. In this area, large boulders cover and comingle with undifferentiated glacial till (Qmu; Figure 2). Thin debris follows a gully paralleling the main slide for up to 500 m and ends above the Albion Basin campground. One additional small area of thin deposit (unit Qlt; Figure 2) was identified east of the narrowest neck of the main rock avalanche body (Figure 2), consisting of a few large boulders on the surface and subsurface boulders up to 1 m across visible in recent construction exposures. METHODS Volume and Runout In order to quantify the volume of the Devils Castle rock avalanche and perform runout simulations, we reconstructed a plausible pre-failure basal topography. We interpreted the subsurface along 42 cross profiles and one longitudinal profile by extrapolating existing surface features, such as the slope of current topography (Pedersen, 2018). Internal consistency was maintained through comparison of key features, such as width and depth of the deposit, between each profile. After creating the profiles, we re-gridded the basal elevation data at 10-m resolution, generating a plausible pre-failure topographic surface. Differenc-

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 201–215

Devils Castle Rock Avalanche

ing the re-created pre-failure topography from current topography allowed us to determine the thickness and volume of the rock avalanche. The headwall of Devils Castle exhibits a near vertical face in some areas and is almost 1 km across (Figure 1c); however, a well-defined evacuated source for the rock avalanche is not immediately apparent. Inspecting for possible source areas on the headwall, we identified two possibilities: an east source and a west source (Figure 1a). As a result, we independently created and analyzed two source areas in our runout modeling. The first contains an evacuated area and a west-facing wall toward the eastern side of the headwall (east source), while the second consists of a long, north-facing cliff toward the western side of the headwall (west source). We modeled the reconstructed cliff similar to the existing headwall by extending the cliff face outward and increasing the elevation of the peaks and ridges. We then re-created the topography of each possible source using the same methods as the deposit reconstruction. Numerical runout simulation provides insight into the dynamics of rock avalanche runout behavior and allows us to evaluate failure scenarios involving the east and west potential source areas. We used DAN3D for runout simulation (Hungr and Evans, 1996), which is an “equivalent fluid” runout model with several rheologies available for describing basal flow resistance (McDougall and Hungr, 2004; Hungr and McDougall, 2009). We modeled the Devils Castle rock avalanche using a Voellmy rheology, which provided the best match to observed deposit extents and estimated thickness. Voellmy rheology describes basal shear resistance (τzx ) as (Hungr and McDougall, 2009) ρgv 2 τzx = − σz f + , (1) ξ where σz is the bed-normal total stress, f is the friction coefficient, ρ is density, v is the depth-averaged velocity, and ξ is the turbulence coefficient. Both f and ξ are calibrated parameters (Hungr and Evans, 1996) that we systematically adjusted until the model results best matched field observations. Using the re-created paleopath topography and source geometries for the east and west source scenarios, we modeled runout of the rock avalanche using parameters from previously published studies as a starting point (Hungr and Evans, 1996; Hungr and McDougall, 2009; and Aaron and Hungr, 2016). Surface Exposure Dating We determined the failure age of the Devils Castle rock avalanche using cosmogenic 36 Cl surface exposure dating of carbonate boulders. In total, we col-

lected samples from six boulders distributed across the deposit surface (Figure 2). We searched for large (>1.0 m and typically 2–3 m tall), flat-topped boulders located in areas of high internal relief within the deposit, such as super-elevated portions along the margins. We collected samples using a tile saw, hammer, and chisel, extracting the upper 2–3 cm of the rock surface at least 30 cm from every edge of the boulder. Topographic shielding data were generated in the field by measuring clear-sky obstructions (e.g., valley walls, mountain peaks) as the angle above the horizon at 10º or 20º increments around the full 360º view. Target boulders were selected to minimize internal shielding and cosmic ray scattering from nearby boulders. We did not apply corrections for snow cover (see IvyOchs et al., 2009). All samples were prepared for measurement of in situ bulk rock cosmogenic 36 Cl concentration following methods modified from Sharma et al. (2000). Chlorine isotopic ratios were determined by accelerator mass spectrometry (AMS) at the Purdue Rare Isotope Measurement (PRIME) Laboratory at Purdue University. Samples were first crushed using a jaw crusher set to 500 μm and were not sieved to avoid grain size biasing. Approximately 100 g of sample were progressively leached twice in 5 percent HNO3 solution for 24 hours in an ultrasonic bath, decanted, and triple rinsed in deionized water in order to remove meteoric 36 Cl. Sample aliquots of the leached sample fraction were reserved for subsequent elemental analyses. Prior to dissolution in concentrated nitric acid, leached samples were spiked with a chlorine carrier enriched in 35 Cl (relative to terrestrial 35 Cl/37 Cl). The chlorine fraction of each sample was then chemically isolated to AgCl for AMS measurement of 35 Cl/37 Cl and 36 Cl/37 Cl. Measured 36 Cl/37 Cl ratios ranged from 5.88 × 10−13 to 6.93 × 10−13 . The procedural blank and chlorine carrier 36 Cl/37 Cl ratios were 4.583 × 10−15 and 4.397 × 10−15 , respectively, suggesting that sample contamination of 36 Cl was insignificant. One sigma (1σ) AMS measurement uncertainties ranged from 1.7 to 3.0 percent. The uncertainties associated with calculation of blank-corrected 36 Cl atoms per gram of sample ranged from approximately 4.0 to 11.0 percent. These large 36 Cl concentration uncertainties relative to the AMS uncertainties can be attributed to high chloride concentration (e.g., sample DC1606). Major and minor element concentrations in each sample were determined using X-ray fluorescence spectroscopy and inductively coupled plasma mass spectroscopy/atomic emission spectroscopy at SGS Minerals (Canada). Geochemical analyses were performed on both unleached (bulk rock) and leached (target rock) sample fractions (Table 1). Geochem-

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 201–215

205

206

0.4 0.4 0.5 0.2 0.6 0.4 0.1 0.0 0.0 0.0 0.0 0.1 30.5 30.7 30.6 30.1 30.5 31.2 0.2 0.1 0.1 0.0 0.0 0.4 0.0 0.0 0.0 0.0 0.0 0.0 30.0 10.0 0.0 10.0 20.0 20.0 1.1 0.1 0.1 0.1 0.1 1.4 1.0 0.9 0.3 1.0 0.9 0.8 0.7 0.4 0.2 0.3 0.3 0.7 0.8 0.3 0.2 0.2 0.3 0.8 0.0 0.0 0.0 0.0 0.0 20.0 39.7 46.3 45.6 45.7 45.6 43.6 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.1 0.1 0.1 0.1 0.1 31.2 30.8 30.4 30.7 30.7 31.5 6.1 1.0 1.2 1.8 1.6 4.6 DC1601 DC1602 DC1603 DC1604 DC1605 DC1606

0.1 0.0 0.0 0.0 0.0 0.1

1.7 0.3 0.3 0.3 0.3 1.3

0.3 0.2 0.4 0.1 0.2 0.3

0.0 0.0 0.0 0.0 0.0 0.0

18.8 21.1 20.4 21.2 20.8 18.1

0.2 0.0 0.0 0.0 0.0 0.4

TiO2 Fe2 O3 (%) (%) CaO (%) B Sm Gd U Th Cr Li K2 O (ppm) (ppm) (ppm) (ppm) (ppm) (ppm) (ppm) (%) CO2 (%) P2 O5 (%) K2 O (%) Na2 O (%) CaO (%) TiO2 Al2 O3 Fe2 O3 MnO MgO (%) (%) (%) (%) (%) SiO2 (%) Sample ID

Unleached Rock

Table 1. Oxide and trace element data for leached and unleached sample fractions used in cosmogenic 36 Cl exposure age calculations.

Leached Rock

Pedersen, Moore, Quirk, Giraud, and McDonald

ical data collected from the leached fraction were used to calculate total 36 Cl production in each sample, and geochemical data collected from the unleached fraction were used to determine neutronproducing, -moderating, and -absorbing properties for each sample. Chloride concentrations were determined for leached sample fractions using a standard isotope dilution method at the PRIME Lab and ranged from 249 to 1,531 ppm. We calculated cosmogenic exposure ages (Table 2) using CRONUSCalc (Marrero et al., 2016a) version 2.0 (http://cronus.cosmogenicnuclides.rocks/2.0). The calculator implements recent improvements for determining cosmogenic exposure ages, including the Lifton-Sato-Dunai nuclide and time-dependent scaling scheme (LSDn; SA in CRONUSCalc; Lifton et al., 2014) and updated 36 Cl production rate estimates (Marrero et al., 2016b). We estimated erosion rates for the sampled carbonate boulders using a combination of ﬁeld evidence and regression analysis linking denudation rates and local mean annual precipitation (Levenson et al., 2017). In the ﬁeld, small chert nodules were occasionally found to extend from the sample boulders measuring ∼3 cm above the surrounding carbonate ground mass. Previous studies in the Wasatch Range using cosmogenic exposure dating of quartz-bearing lithologies have estimated erosion rates between 1.0 and 3.0 mm kyr−1 (Laabs et al., 2011; Quirk et al., 2018), giving an order-of-magnitude estimate for chert erosion rates. Meanwhile, Levenson et al. (2017) established a linear regression relating long-term carbonate erosion rates (mm kyr−1 ) to modern mean annual precipitation (MMAP; mm yr−1 ) as

y = 0.02x + 6.4,

(2)

where y is denudation and x is precipitation. MMAP less than 3 km away from the rock avalanche at the Alta Cooperative Observer Program weather station (Station #420072-5) is approximately 1,370 mm yr−1 with over half falling as snow. The estimated erosion rate for this MMAP using Eq. 1 is 33.8 mm kyr−1 . However, because of the relatively poor ﬁt of the linear regression to the available data and uncertainty of the eﬀect that solid–liquid precipitation partitioning has on the erosion rate estimate, we use this only as an order-of-magnitude estimate and select an erosion rate of 10 mm kyr−1 for calculating exposure ages. It should be noted that calculated exposure ages overlap within analytical uncertainty whether an erosion rate of either 10 or 33.8 mm kyr−1 is used.

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 201–215

2.0 2.5 2.6 2.2 3.6 2.7 — 0.6 1.0 0.5 0.6 0.8 1.5 1.0 13.0 14.2 15.4 14.3 21.2 15.3 14.4 0.98714 0.98826 0.99018 0.98930 0.98897 0.99256 — 10 10 10 10 10 10 — 2.5 2 2.5 2 2 3 — 2.49E+02 9.62E+02 1.53E+03 5.15E+02 1.02E+03 5.15E+02 — 1.69E+05 9.27E+05 6.19E+05 3.22E+05 5.10E+05 6.00E+05 —

Cl Error Cl (ppm)

3.36E+06 1.03E+07 1.60E+07 6.29E+06 1.06E+07 5.53E+06 — −111.6158 −111.6146 −111.6148 −111.6153 −111.6161 −111.6174 — 40.5731 40.5737 40.5765 40.5785 40.5801 40.5800 — 2,873 2,913 2,870 2,856 2,139 2,839 — DC1601 DC1602 DC1603 DC1604 DC1605 DC1606 Mean age

Cl (Atoms g Sample−1 )

Cl Error

1.35E+01 8.56E+01 3.92E+01 2.66E+01 4.36E+01 5.77E+01 —

Erosion Topographic Exposure Internal Rate Shielding Age Error (mm/kyr) Factor (ka) (ka) Sample Thickness (cm)

RESULTS AND DISCUSSION

Longitude (ºW) Latitude (ºN) Elevation (m a.s.l.) Sample ID

Table 2. Cosmogenic 36 Cl exposure age calculation input and results using the CRONUS calculator (Marrero et al., 2016a, 2016b).

LSDn (SA)

External Error (ka)

Devils Castle Rock Avalanche

Volume and Runout We calculated the rock avalanche deposit volume as 1.7 × 106 m3 , a value we presume to be accurate within ±25 percent through analysis of error sources (described below). Topographic reconstruction further allowed us to estimate the spatial distribution of deposit thickness (Figure 3a). We found thicker deposits at the distal and proximal ends of the rock avalanche, where topography flattens, while thickness in the middle of the deposit is lower, corresponding to an increase in steepness of the basal long profile. The maximum thickness of the deposit is 25 m, and the average thickness is 7 m. To determine the mobility of the rock avalanche from field and topographic data, we measured the Fahrboeschung angle, or the ratio of the fall height to path length connecting the highest point of the source and most distal point of the deposit (Heim, 1932). This value is 14º for the Devils Castle rock avalanche (equivalent height-to-length ratio = 0.25). Furthermore, two points of deposit super-elevation were identified from field observations (Figure 1a). The first occurs onethird of the way down the deposit where the surface is super-elevated approximately 10 m rising to the west. The second occurs at the lowermost bend just before the toe of the deposit, where the surface is superelevated approximately 10 m to the east. Minimum rock avalanche velocity needed to attain the measured super-elevation can be estimated as (Jibson et al., 2006) ghr , (3) Vmin = w where Vmin is the minimum velocity (neglects basal resistance), g is gravity, h is the super-elevation height of the deposit at the bend, r is the radius of the bend, and w is the width of the channel. Using Eq. 2, we determined the minimum speed of the rock avalanche to be ∼27 m/s at the upper point and ∼15 m/s at the lower bend. The Devils Castle rock avalanche displays high mobility for its volume. Our calculated Fahrboeschung angle of 14º is low compared to deposits of similar volume (Figure 4), indicating that most landslides of this volume have comparably shorter runout distances. One possible mechanism that may explain the low Fahrboeschung is entrainment of water during runout (Hungr and Evans, 2004; Dufresne et al., 2010). We hypothesize there may have been a small tarn at the head of the basin that, when overrun by the rock avalanche, was entrained and helped increase mobility. Cecret Lake, located less than 1 km to the west, provides an example analog of such a tarn (Figure 1a). The de-

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 201–215

207

Pedersen, Moore, Quirk, Giraud, and McDonald

Figure 3. Comparison of deposit thickness from reconstruction and modeling. (a) Estimated thickness of the deposit obtained from topographic reconstruction. (b and c) Final runout simulation thickness for the west and east source areas, respectively. Runout simulation time for both scenarios is 100 seconds. Final modeled thickness has been bulked by 25 percent for comparison with mapped and reconstructed values. Paleoboundary is the estimated boundary of the original deposit. Base image from 2-m bare-earth lidar (Utah Automated Geographic Reference Center, 2006).

posit is relatively narrow and long and constrained by low-relief topography, which further supports possible high ﬂuid content for the debris aided by entrainment of water. Runout over a small ice body, while also

Figure 4. Fahrboeschung vs. volume plot for the six rock avalanches found in the region (see Table 3). Each slide, including Devils Castle, is plotted (red dots with error bars) together with other reported global rock avalanche events (gray dots; data from Davidson, 2011). Smith and Morehouse (SM), White Pine (WP), Kelsey Peak (KP), Devils Castle (DC), South Fork (SF), and Grandview (GV).

208

possible, is in contrast suspected to lead to thinning and lateral spreading of the deposit (e.g., Jibson et al., 2006). Topographic channeling of the debris also likely contributed to the comparably long runout distance (Nicoletti and Sorriso-Valvo, 1991), as is entrainment of water from runout over snow. When intact rock masses collapse and run out, they undergo fragmentation bulking (Hungr and Evans, 2004), which increases their volume. In order to recreate the topography of the source area, we accounted for fragmentation by debulking the calculated deposit volume by 25 percent, a typical fragmentation volume change (Sherard et al., 1963; Hungr and Evans, 2004), which also reﬂects our chosen bulk density of 22 kN/m3 for runout simulation. This resulted in an estimated source volume of 1.36 × 106 m3 . Selecting a diﬀerent bulking factor would require us to either know the bulk density of the rock avalanche deposit or have an independent measure of the source volume, which we do not have. Increasing or decreasing the assumed bulking factor would lead to an equivalent change in estimated source volume. We believe the assumed fragmentation bulking factor may reasonably vary within ±5 percent, giving a source volume range

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 201–215

Devils Castle Rock Avalanche

of 1.3–1.4 × 106 m3 . We assumed no entrainment of colluvium or glacial till substrate during runout, as there was no evidence of entrained debris observable within the rock avalanche deposit (Hungr and Evans, 2004). The largest source of error in our volume calculation likely arises from topographic reconstruction; however, this error is difficult to quantify. A small change of ±2 m in the mean paleovalley depth, for example, could result in a volume change of ±29 percent. Moreover, the upper part of the deposit is concealed by glacial and periglacial debris and talus, which makes determining the precise extents and shape of former topography difficult. Further, we observed two small, thin lateral deposits located parallel and east of the main rock avalanche deposit. These are included in our mapping (Figure 2); however, we did not include their volume in our calculations. These lateral deposits are no more than 2–3 m thick and have a combined surface area of 41,500 m2 , giving an estimated total volume of 83,000–124,000 m3 (∼6 percent of the deposit volume). Also not included in our volume calculation is material lost by post-slide erosion. Throughout the deposit, a stream has eroded a small gully, and we estimate that approximately 95,000 m3 of the rock avalanche has been removed through erosion, or 5 percent of the deposit volume. We use these error calculations to estimate ±25 percent confidence bounds on our reported deposit volume of 1.7 × 106 m3 . Runout Modeling Geological evidence, combined with the overall good match between field mapping and runout modeling results, indicates that the rock avalanche was a single event failure sourced from the Devils Castle headwall. However, there is no obvious evidence on the headwall to indicate the precise source geometry. We identified two possibilities for the source volume, each reconstructed with a plausible pre-failure topography, and then simulated the resulting runout. The goal was to extract hazard-relevant parameters describing the dynamics of the failure and evaluate the likelihood of the failure originating from either the east or the west source areas. Runout parameters, such as the selected rheology and values describing basal and internal shear resistance, were identical between models for the east and west source areas, and the results of each model gave similar final values (Figure 3). Through trial and error, we discovered that each of the two source scenarios can be modeled successfully using the same Voellmy rheology with resistance parameters of f = 0.09 and ξ = 300 m/s2 , a unit weight of 22 kN/m3 , and an internal friction angle of 35º.

Models for the east and west source areas have similar dynamics despite the different initial geometries (Figure 5); the source collapse is funneled into the main gully and continues to run out, reaching generally similar extents (Figures 3b and 3c). However, there are noticeable differences between the two scenarios. After collapse, debris from the east source spreads and separates into three lobes, with the majority of debris traveling into the main gully (Figure 5c). One of the three lobes falls into a gully parallel to and east of the main lobe and continues downstream, while the other lobe proceeds into a smaller gully to the west and eventually terminates. In the west source model, the mass collapses and splits into two lobes and runs out in a similar manner as the east source but does not enter the eastern gully (Figure 5c). Another difference is seen ∼30 seconds into the simulation (Figure 5c) where the east source model super-elevates as it enters the main gully, while the west source model does not super-elevate until 250 m above the toe of the deposit. We determined values for maximum runout velocity and super-elevation velocities from DAN3D simulations of the east and west source failure scenarios. The east source failure attains a maximum velocity of ∼62 m/s and has two super-elevation locations closely matching those discovered during field mapping. The upper super-elevation point has a modeled velocity of 25–30 m/s, while the lower super-elevation point has a modeled velocity of 10–15 m/s. The west source failure attains a maximum velocity of ∼57 m/s but has only one super-elevation point toward the toe of the deposit, in the same location as the east source model, with a modeled velocity of 10–15 m/s. For comparison, field evidence indicated minimum super-elevation velocities of 27 and 15 m/s at the upper and lower points, respectively (using Eq. 3). The simulated thickness and distribution of deposits from DAN3D closely match our thickness results from topographic reconstruction (Figure 3). Both east and west source models produce similar deposit thickness. However, when compared to the reconstructed thickness, the models tend to place more debris toward the middle area and thinner deposits toward the toe. The east source model matches the mapped extents well and includes the lateral deposit found east of the slide observed in the field (Figure 3c). However, it lacks the depth we estimated from topographic reconstruction. For example, the deepest part we reconstructed is found near the top of the deposit at a depth of over 25 m, whereas the deepest part of the modeled east source deposit in the same area is 10–12 m. The west source model also matches the mapped extents well but is missing the upper lateral deposit observed in the field (Figure 3b).

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 201–215

209

Pedersen, Moore, Quirk, Giraud, and McDonald

Figure 5. Runout model comparison. (a–e) Snapshots of the simulations at the indicated times in seconds. Top row: east source; bottom row: west source. Base image: 2-m bare-earth lidar (Utah Automated Geographic Reference Center, 2006).

Both east and west source runout models indicated that a small portion of the rock avalanche descended a side gully parallel to and west of the main channel, resulting in ∼2 m of simulated deposit (Figure 3). However, in field visits, we could not identify evidence of this lateral deposit and thus believe that the result is an artifact of our topographic reconstruction. The area is adjacent to glacial and periglacial deposits that obscure path topography, making reconstruction difficult, and we re-created a gully guided by downstream forms that might have inadvertently funneled some of the deposit to the west. While both the east and the west source models produced reasonable results, field evidence suggests that the rock avalanche super-elevated toward the west near the head of the main strike gully (Figure 5c, east source). Analyzing the two runout models, we found that the east source slide super-elevates at this point, whereas the west source slide does not. Field observations further show a thin layer of boulders located east of the main body (Figure 2). Our east source runout model places boulders in the same area as observed,

210

whereas the west source model lacks this lateral deposit. Although uncertainties remain, evidence thus suggests that the east area is the most likely source of the Devils Castle rock avalanche. Age Six cosmogenic 36 Cl surface exposure ages constrain the failure age for the Devils Castle rock avalanche (Figure 6; Table 2). Individual exposure ages range from 13.0 to 21.2 ka. However, sample DC1605 with an exposure age of 21.2 ka does not overlap with any other sample within analytical uncertainty; we interpret this outlying older age as likely to result from isotope inheritance during prior exposure. The remaining five samples cluster relatively well, with four of the five samples overlapping within analytical uncertainty, and range from 13.0 to 15.4 ka. Therefore, we have confidence that all but one of the individual sample exposure ages are part of the same statistical population and that the mean of the remaining five boulder exposure ages (±1σ) can be used to assign a

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 201–215

Devils Castle Rock Avalanche

Figure 6. Cosmogenic 36 Cl exposure ages for the six rock avalanche boulder samples. Individual sample error is the analytical (internal) uncertainty. Sample DC1605 is interpreted as an old outlier likely due to isotope inheritance and was excluded from landform age calculation. The dark horizontal line represents the mean age of the remaining ﬁve samples, while the gray box represents the standard deviation of individual samples around the mean age.

failure age of 14.4 ± 1.0 ka to the Devils Castle rock avalanche. This age places the rock avalanche near the beginning of the Bølling-Allerød North Atlantic warming interval, which preceded the Younger Dryas (Shakun et al., 2012), a time when the Wasatch Range was likely almost completely deglaciated (Munroe and Laabs, 2017; Quirk et al., 2018). Similar ages throughout the deposit support our hypothesis that this was a single event failure, although the age resolution is insuﬃcient to detect multiple events in close succession. Previously reported radiocarbon ages along with cosmogenic dating presented here bracket the possible age of the nested moraines near the headwall (Figure 2). Madsen and Currey (1979) found a calibrated 14 C age of 8,162–8,484 cal BP (recalculated calibrated 1σ range using IntCal13; Reimer et al., 2013) for humic colluvium sampled from inside the nested moraines (sample GX-4644), and 10,571–11,205 cal BP (sample RL-695) and 10,397–11,103 cal BP (sample GX-4736) from wood and peat at the base of a bog overlying the rock avalanche deposit (McCoy, 1977). Combining this information with the age of the deposit, we estimate that the moraines were deposited between approximately 10,000 and 14,000 years ago.

2012) that may have contributed to development of the rock avalanche. Glacial erosion of the cirque headwall progressively undermined the source area, debuttressing the rock wall and reorienting stresses (McColl, 2012; Grämiger et al., 2017), producing rock mass damage that conditioned the slope for future failure. Throughout the Late Pleistocene, climate cycles likely continued to prepare the slope for failure. Weathering processes, such as frost cracking (Sanders et al., 2012) and thermo-mechanical fatigue (Gischig et al., 2011), may have contributed to further rock mass damage and progressive failure surface development. The age of the rock avalanche places it during a time of relatively rapid Northern Hemisphere warming (Shakun et al., 2012). Geological evidence of superposed moraines suggests that the rock avalanche occurred under nearly ice-free conditions (Madsen and Currey, 1979). While this result permits a possible interpretation of glacial debuttressing in causing the rock avalanche (cf. McColl et al., 2010; Grämiger et al., 2017), it also holds potentially important implications for glacier and climate reconstructions in the Wasatch Mountains during the Late Glacial interval helping constrain de-glaciation timing for the highest cirques (cf. Quirk et al., 2018). The Wasatch Mountains are situated at the easternmost end of the Basin and Range Province in a seismically active region (DuRoss, 2008). An earthquake trigger is therefore feasible for the Devils Castle rock avalanche; however, our calculated age is beyond the record of documented prehistoric earthquakes (ca. 6,000 years). Still, we cannot rule out the possibility of a seismic trigger due to the active nature and epicentral distance of the Wasatch Fault. Available simulation results for an magnitude 7 earthquake scenario on the Salt Lake City segment of the Wasatch Fault indicate likely Modified Mercalli shaking intensity of VI (strong) at the Devils Castle headwall (Pankow et al., 2015). Such intensity values are known to be capable of triggering coherent landslides, such as rock avalanches (Keefer, 1984), while additional local topographic and rock fracturing effects on steep mountain ridges may further amplify ground shaking (Gischig et al., 2015). Other triggers for the rock avalanche are also plausible, including snowmelt or rain infiltration, permafrost thaw, and extended cold or exceptionally hot temperatures, but equally feasible is that the event had no recognizable trigger, failing simply through progressive slope fatigue (Collins and Stock, 2016). Other Rock Avalanches and Implications

Preparatory and Triggering Factors The headwall of Devils Castle is topographically complex, reﬂecting preparatory factors (sensu McColl,

We identiﬁed ﬁve additional deposits of prehistoric rock avalanches within 60 km of the Wasatch Front. Table 3 provides summary information for these

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 201–215

211

Pedersen, Moore, Quirk, Giraud, and McDonald Table 3. Details of additional rock avalanches identified near the Wasatch front.

Name Devils Castle Kelsey Peak Grandview Peak Smith and Morehouse South Fork White Pine

Latitude (ºN)

Longitude (ºW)

Scarp Elevation (m)

Runout Distance (km)

Deposit Volume (m3 )

Fahrboeschung (º)

40.575 40.445 40.844 40.770 40.722 40.574

111.615 112.223 111.748 111.090 111.168 111.688

3,280 3,065 2,650 3,050 2,635 2,825

1.8 4.3 3.2 2.7 1.3 1.2

1.2 × 106 –2.4 × 106 3.0 × 106 –6.0 × 106 8.6 × 106 –2.2 × 107 2.3 × 107 –5.9 × 107 1.4 × 106 –3.4 × 106 9.8 × 105 –2.0 × 106

14 21 10 14 15 22

five additional slides, including estimated volume and Fahrboeschung angle. These rock avalanches were selected based on (1) their proximity to the Wasatch Fault zone and other active Holocene faults, (2) the character of the deposit suggesting very rapid runout, and (3) size, with each having a volume similar to or larger than Devils Castle. Figure 4 places these landslides in the context of worldwide case history data and compares them to previously studied events. All five Wasatch Front rock avalanches are large in volume (1 × 106 m3 –2 × 107 m3 ), three have low Fahrboeschung angles for their respective size (Devils Castle, South Fork, and Grandview Peak), and two have comparably large volumes (Smith and Morehouse and Grandview Peak). As these five deposits as well as Devils Castle are located close to the Wasatch Front, regardless of how they were triggered, geological evidence shows that large rock avalanches occur and should be considered a high-magnitude, low-frequency hazard. CONCLUSIONS We present a detailed case history analysis of the Devils Castle rock avalanche, a large-magnitude landslide characterized by high mobility and fluid-like runout. The deposit is located in the Albion Basin at the head of Little Cottonwood Canyon, near the town of Alta, Utah, in a popular area for summer and winter recreation. The rock avalanche deposit consists mainly of Deseret/Gardison Limestone and is ∼1.5 km in length, with the visible portion of the deposit 1 km long and source-proximal debris obscured by superposed glacial and periglacial deposits. The calculated Fahrboeschung angle is 14º, which indicates that the slide has relatively high mobility for its volume. We hypothesize there may have been a small lake at the head of the basin that, when overrun, increased the mobility through entrainment of water. Through topographic reconstruction and differencing, we calculated a deposit volume of 1.7 million m3 , and accounting for a fragmentation bulking during failure; we estimated

212

that the source volume was 1.36 × 106 m3 . The reconstructed maximum thickness of the deposit is ∼25 m, and the average thickness is 7 m. Through cosmogenic nuclide surface exposure dating of deposited boulders, we estimate a failure age for the rock avalanche of 14.4 ± 1.0 ka. Studying the headwall, we found no obvious evidence indicating the precise source geometry. Reconstructing two plausible source areas, we found that results of numerical runout simulations agree well with mapped deposit boundaries in each case. However, the east source model better reproduced characteristics of the deposit observed in the field, including the two super-elevation points and the uppermost lateral deposit. The east source failure attains a maximum velocity of ∼62 m/s, with velocities of 25–30 m/s at the upper super-elevation point and 10–15 m/s at the lower super-elevation point. Evidence suggests that the Devils Castle rock avalanche was a single event failure, drawing on combined results of field mapping, runout modeling, and cosmogenic nuclide dating. Modeled extents match well with boundaries mapped in the field, and the modeled thickness and overall distribution of debris agree with our topographic reconstruction. Meanwhile, no field evidence of superposed deposits was discovered that might indicate multiple overlapping rock avalanche events. Exposure dating of boulders distributed across the surface of the deposit also suggests that the boulders were deposited simultaneously at ∼14 ka. The one outlier is interpreted as an old outlier likely due to isotope inheritance. Geological and geochronological evidence suggest the Albion Basin was nearly ice free at the time of the rock avalanche. Preceding glacial erosion of the cirque headwall likely progressively undermined the source area, debuttressing the wall through removal of bedrock and altering stresses over time, potentially playing a key role in conditioning the slope for future failure. Later climate cycles and environmental weathering processes, in addition to seismicity, likely generated further rock mass damage, contributing to failure surface development and ultimate collapse.

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 201–215

Devils Castle Rock Avalanche

ACKNOWLEDGMENTS The authors gratefully acknowledge constructive reviews from Eldon Gath, John Clague, and Tom Badger. Reviews of an earlier draft by Steve Bowman, Kimm Harty, Mike Hylland, and Rick Allis (Utah Geological Survey) are also appreciated. This study incorporates work supported by the National Science Foundation under grant nos. EAR-1358514, 1358554, 1358401, 1358443, and 1101100 (EarthScope National Office). Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation. Thanks to the EarthScope program for its support. REFERENCES Aaron, J. and Hungr, O., 2016, Dynamic analysis of an extraordinarily mobile rock avalanche in the Northwest Territories, Canada: Canadian Geotechnical Journal, Vol. 53, No. 6, pp. 899–908. doi:10.1139/cgj-2015-037. Aaron, J.; McDougall, S.; Moore, J. R.; Coe, J. A.; and Hungr, O., 2017, The role of initial coherence and path materials in the dynamics of three rock avalanche case histories: Geoenvironmental Disasters, Vol. 4, No. 5, 15 p. doi:10.1186/s40677017-0070-4. Ashland, F. X. and McDonald, G. N., 2008, Reconnaissance of the Grandview Peak Rock Slide, Salt Lake County, Utah— A Possible Earthquake-Induced Landslide?: Utah Geological Survey Open File Report 518, 13 p. Baker, A. A.; Calkins, F. C.; Crittenden, M. D.; and Bromfield, C. S., 1966, Geologic Map of the Brighton Quadrangle, Utah: U.S. Geological Survey Geologic Quadrangle Map GQ-534. Ballantyne, C. K.; Sandeman, G. F.; Stone, J. O.; and Wilson, P., 2014, Rock-slope failure following Late Pleistocene deglaciation on tectonically stable mountainous terrain: Quaternary Science Reviews, Vol. 86, pp. 144–157. Calkins, F. C. and Butler, B. S., 1943, Geology and Ore Deposits of the Cottonwood-American Fork Area, Utah: U.S. Geological Survey Professional Paper 201, 195 p. Cardoso, R. K., 2002, Timing and Mechanics of the White Pine Rockslide in Little Cottonwood Canyon: Unpublished M.S. Thesis, University of Utah, Salt Lake City, UT, 77 p. Castleton, J., Moore, J. R.; Aaron, J.; Christl, M.; and IvyOchs, S., 2016, Dynamics and legacy of 4.8 ka rock avalanche that dammed Zion Canyon, Utah: GSA Today, Vol. 26, No. 6, pp. 4–9. Christenson, G. E. and Ashland, F. X., 2006, Assessing the stability of landslides—Overview of lessons learned from historical landslides in Utah, in 40th Symposium on Engineering Geology and Geotechnical Engineering: Utah State University, Logan, UT, 2006, pp. 1–17. Coe, J. A.; Baum, R. R. L.; Allstadt, K. E.; Kochevar, B. F.; Schmitt, R. G.; Morgan, M. L.; White, J. L.; Stratton, B. T.; Hayshi, T. A.; and Kean, J. W., 2016, Rock-avalanche dynamics revealed by large-scale field mapping and seismic signals at a highly mobile avalanche in the West Salt Creek valley, western Colorado: Geosphere, Vol. 12, No. 2, pp. 607–631. doi: https://doi.org/10.1130/GEOS01265.1. Coe, J. A.; Bessette-Kirton, E. K.; and Geertsema, M., 2017, Increasing rock-avalanche size and mobility in Glacier Bay

National Park and Preserve, Alaska detected from 1984 to 2016 Landsat imagery: Landslides, Vol. 15, No. 3, pp. 393–407. doi:10.1007/s10346-017-0879-7. Collins, B. D. and Stock, G. M., 2016, Rockfall triggering by cyclic thermal stressing of exfoliation fractures: Nature Geoscience, Vol. 9, No. 5, pp. 395–400. doi:10.1038/ngeo2686. Crosta, G.; Hermanns, R.; Dehls, J.; Lari, S.; and Sepulveda S., 2017, Rock avalanches clusters along the northern Chile coastal scarp: Geomorphology, Vol. 289, pp. 27–43. David Evans and Associates, 2011, Albion Basin Transportation Feasibility Study: David Evans and Associates, Salt Lake City, UT, 159 p. Davidson, C., 2011, Rock Avalanches: Unpublished M.S. Thesis, University of British Columbia, Vancouver, BC, Canada, 13 p. Douglass, J.; Dorn, R. I.; and Gootee, B., 2005, A large landslide on the urban fringe of metropolitan Phoenix, Arizona: Geomorphology, Vol. 65, No. 3–4, pp. 321–336. Dufresne, A.; Davies, T. R.; and McSaveney, M. J., 2010, Inﬂuence of runout path material on emplacement of the Round Top rock avalanche, New Zealand: Earth Surface Processes Landforms, Vol. 35, No. 2, pp. 190–201. doi:1031002/esp.1900. Dunning, S. A.; Mitchell, W. A.; Rosser, N. J.; and Petley, D. N., 2007, The Hattian Bala rock avalanche and associated landslides triggered by the Kashmir earthquake of 8 October 2005: Engineering Geology, Vol. 93, No. 3–4, pp. 130–144. DuRoss, C. B., 2008, Holocene vertical displacement on the central segments of the Wasatch fault zone, Utah: Bulletin Seismological Society of America, Vol. 98, No. 6 pp. 2918–2933. doi:10.1785/0120080119. Gischig, V. S.; Eberhardt, E.; Moore, J. R.; and Hungr, O., 2015, On the seismic response of deep-seated rock slope instabilities—Insights from numerical modeling: Engineering Geology, Vol. 193, pp. 1–18. Gischig, V. S.; Moore, J. R.; Evans, K. F.; Amann, F.; and Loew, S., 2011, Thermomechanical forcing of deep rock slope deformation: 1. Conceptual study of a simpliﬁed slope: Journal Geophysical Research: Earth Surface, Vol. 116, No. 4, pp. 1–18. doi:10.1029/2011JF002006. Grämiger, L. M.; Moore, J. R.; Gischig, V. S.; Ivy-Ochs, S.; and Loew, S., 2017, Beyond debuttressing: Mechanics of paraglacial rock slope damage during repeat glacial cycles: Journal Geophysical Research: Earth Surface, Vol. 122, No. 4, pp. 1004–1036. doi:10.1002/2016JF003967. Grämiger, L. M.; Moore, J. R.; Vockenhuber, C.; Aaron, J.; Hajdas, I.; and Ivy-Ochs, S., 2016, Two early Holocene rock avalanches in the Bernese Alps (Rinderhorn, Switzerland): Geomorphology, Vol. 268, pp. 207–221. doi:10.1016/j.geomorph.2016.06.008. Heim, A., 1932, Der Bergsturz und Menschenleben: Fretz und Wasmuth Verlag, Zürich, 218 p. Hooper, W. G., 1951, Geology of the Smith and Morehouse-South Fork Area, Utah: Unpublished M.S. Thesis, University of Utah, Salt Lake City, UT, 60 p. Hungr, O. and Evans, S. G., 1996, Rock avalanche runout prediction using a dynamic model. In Senneset, K. (Editor), Proceedings of the 7th International Symposium on Landslides. CRC Press, Rotterdam, pp. 233–238. Hungr, O. and Evans, S. G., 2004, Entrainment of debris in rock avalanches: An analysis of a long run-out mechanism: Bulletin Geological Society of America, Vol. 116, No. 9–10, pp. 1240– 1252. doi:10.1130/B25362.1. Hungr, O.; Evans, S.; and Hazzard, J., 1999, Magnitude and frequency of rock falls and rock slides along the main

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 201–215

213

Pedersen, Moore, Quirk, Giraud, and McDonald transportation corridors of southwestern British Columbia: Canadian Geotechnical Journal, Vol. 36, pp. 224–238. doi:10.1139/t98-106. Hungr, O. and McDougall, S., 2009, Two numerical models for landslide dynamic analysis: Computers and Geosciences, Vol. 35, No. 5, pp. 978–992. doi:10.1016/j.cageo.2007.12.003. Ivy-Ochs, S.; Martin, S.; Campedel, P.; Hippe, K.; Alfimov, V.; Vockenhuber, C., Andreotti, E., Carugati, G., Pasqual, D., Rigo, M., and Viganò, A., 2017, Geomorphology and age of the Marocche di Dro rock avalanches (Trentino, Italy): Quaternary Science Reviews, Vol. 169, pp. 188–205. Ivy-Ochs, S., Poschinger, A. V.; Synal, H. A.; and Maisch, M., 2009, Surface exposure dating of the Flims landslide, Graubünden, Switzerland: Geomorphology, Vol. 103, No. 1, pp. 104–112. Jibson, R. W.; Harp, E.; Schulz, W.; and Keefer, D. K., 2006, Large rock avalanches triggered by the M 7.9 Denali Fault, Alaska, earthquake of 3 November 2002: Engineering Geology, Vol. 83, No. 1–3, pp. 144–160. doi:10.1016/j.enggeo.2005.06.029. Keefer, D. K., 1984, Landslides caused by earthquakes: Geological Society America Bulletin, Vol. 95, pp. 406–421. Köpfli, P.; Grämiger, L. M.; Moore, J. R.; Vockenhuber, C.; and Ivy-Ochs, S., 2018, The Oeschinensee rock avalanche, Bernese Alps, Switzerland: A co-seismic failure 2300 years ago?: Swiss Journal Geosciences, Vol. 111, pp. 205–219. Laabs, B. J.; Marchetti, D. W.; Munroe, J. S.; Refsnider, K. A.; Gosse, J. C.; Lips, E. W.; Becker, R. A.; Mickelson, D. M.; and Singer, B. S., 2011, Chronology of latest Pleistocene mountain glaciation in the western Wasatch Mountains, Utah, USA: Quaternary Research, Vol. 76, No. 2, pp. 272–284. Laabs, B. J. and Munroe, J. S., 2016, Late Pleistocene mountain glaciation in the Lake Bonneville basin. In Oviatt, G. C. (Editor), Lake Bonneville: A Scientific Update: Elsevier, Amsterdam, pp. 462–498. Levenson, Y.; Ryb, U.; and Emmanuel, S., 2017, Comparison of ﬁeld and laboratory weathering rates in carbonate rocks from an Eastern Mediterranean drainage basin. Earth Planetary Science Letters, Vol. 465, pp. 176–183. Lifton, N.; Sato, T.; and Dunai, T. J., 2014, Scaling in situ cosmogenic nuclide production rates using analytical approximations to atmospheric cosmic-ray ﬂuxes: Earth Planetary Science Letters, Vol. 386, pp. 149–160. Lipovsky, P. S.; Evans, S. G.; Clague, J. J.; Hopkinson, C.; Couture, R.; Bobrowsky, P.; Ekstrom, G.; Demuth, M. N.; Delaney, K. B.; Roberts, N. J.; Clarke, G.; and Schaeffer, A., 2008, The July 2007 rock and ice avalanches at Mount Steele, St. Elias Mountains, Yukon, Canada: Landslides, Vol. 5, No. 4, pp. 445–455. doi:10.1007/s10346-008-0133-4. Loew, S.; Gschwind, S.; Gischig, V.; Keller-Singer, A.; and Valenti, G., 2017, Monitoring and early warning of the 2012 Preonzo catastrophic rockslope failure: Landslides, Vol. 14, No. 1, pp. 141–154. doi:10.1007/s10346-016-0701-y. Madsen, D. B. and Currey, D. R., 1979, Late Quaternary glacial and vegetation changes: Quaternary Research, Vol. 12, No. 2, pp. 254–270. Marrero, S. M.; Phillips, F. M.; Borchers, B.; Lifton, N.; Aumer, R.; and Balco, G., 2016a, Cosmogenic nuclide systematics and the CRONUScalc program: Quaternary Geochronology, Vol. 31, pp. 160–187. Marrero, S. M.; Phillips, F. M.; Caffee, M. W.; and Gosse, J. C., 2016b, CRONUS-Earth cosmogenic 36Cl calibration: Quaternary Geochronology, Vol. 31, pp. 199– 219.

214

McCalpin, J. P., 2002, Post-Bonneville Paleoearthquake Chronology of the Salt Lake City Segment, Wasatch Fault Zone, from the 1999 “Megatrench” Site: Geological Survey Miscellaneous Publication 02-7, 45 p. McColl, S. T., 2012, Paraglacial rock-slope stability: Geomorphology, Vol. 153–154, pp. 1–16. doi:10.1016/j.geomorph.2012. 02.015. McColl, S. T.; Davies, T. R. H.; and McSaveney, M. J., 2010, Glacier retreat and rock-slope stability: Debunking debuttressing: Delegate Papers, Geologically Active, 11th Congress of the International Association for Engineering Geology and the Environment, Auckland, Aotearoa, September 5–10, 2010: Auckland, New Zealand, pp. 467–474 McCoy, W. D., 1977, A Reinterpretation of Certain Aspects of the Glacial History: Unpublished M.A. Thesis, University of Utah, Salt Lake City, UT, 95 p. McDougall, S. and Hungr, O., 2004, A model for the analysis of rapid landslide motion across three-dimensional terrain: Canadian Geotechnical Journal, Vol. 41, No. 6, pp. 1084–1097. doi:10.1139/t04-052. Moore, J. R.; Pankow, K. L.; Ford, S. R.; Koper, K. D.; Hale, J. M.; Aaron, J.; and Larsen, C. F., 2017, Dynamics of the Bingham Canyon Rock Avalanches (Utah, USA) resolved from topographic, seismic, and infrasound data: Journal Geophysical Research Earth Surface, Vol. 122, No. 3, pp. 615–640. doi:10.1002/2016JF004036. Munroe, J. S. and Laabs, B. J., 2017, Combining radiocarbon and cosmogenic ages to constrain the timing of the last glacialinterglacial transition in the Uinta Mountains, Utah, USA: Geology, Vol. 45, No. 2, pp. 171–174. Nicoletti, P. G. and Sorriso-Valvo, M., 1991, Geomorphic controls of the shape and mobility of rock avalanches: Geological Society America Bulletin, Vol. 103, No. 10, pp. 1365–1373. Pankow, K. W.; Arabasz, J.; Carey, R.; Christenson, G.; Groeneveld, J.; Maxfield, B.; McDonough, P. W.; Welliver, B.; and Youd, T. L., 2015, Scenario for a Magnitude 7.0 Earthquake on the Wasatch Fault—Salt Lake City Segment: Hazards and Loss Estimates: Earthquake Engineering Research Institute, Salt Lake City, UT, 57 p. Pedersen, P., 2018, Reconstructing the Devils Castle Rock Avalanche, Albion Basin, Utah: Unpublished M.S. Thesis, University of Utah, Salt Lake City, UT, 86 p. Prager, C.; Zangerl, C.; Patzelt, G.; and Brandner, R.; 2008, Age distribution of fossil landslides in the Tyrol (Austria) and its surrounding areas: Natural Hazards Earth Systems Sciences, Vol. 8, No. 2, pp. 377–407. Quirk, B. J.; Moore, J. R.; Laabs, B. J. C.; Caffee, M. W.; and Plummer, M. A., 2018, Termination II, last glacial maximum, and Lateglacial chronologies and paleoclimate from Big Cottonwood Canyon, Wasatch Mountains, Utah: GSA Bulletin, Vol. 130, No. 11–12, pp. 1889–1902. doi: 10.1130/B31967.1. Reimer, P. J.; Bard, E.; Bayliss, A.; Beck, J. W.; Blackwell, P. G.; Ramsey, C. B.; Buck, C. E.; Cheng, H.; Edwards, R. L.; Friedrich, M.; Grootes, P. M.; Guilderson, T. P.; Haflidason, H.; Hajdas, I.; Hatté, C.; Heaton, T. J.; Hoffmann, D. L.; Hogg, A. G.; Hughen, K. A.; Kaiser, K. F.; Kromer, B.; Manning, S. W.; Niu, M.; Reimer, R. W.; Richards, D. A.; Scott, E. M.; Southon, J. R.; Staff, R. A.; Turney, C. S. M.; and van der Plicht, J., 2013, IntCal13 and Marine13 radiocarbon age calibration curves 0–50,000 years cal BP: Radiocarbon, Vol. 55, No. 4, pp. 1869–1887. doi:10.2458/azu_js_rc.55.16947. Richmond, G. M., 1964, Glaciation of Little Cottonwood and Bells Canyons, Wasatch Mountains, Utah: U.S. Geological Society Professional Paper 454-D, 45 p. doi:10.1021/nn204239d.

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 201–215

Devils Castle Rock Avalanche Sanders, J. W.; Cuffey, K. M.; Moore, J. R.; MacGregor, K. R.; and Kavanaugh, J. L., 2012, Periglacial weathering and headwall erosion in cirque glacier bergschrunds: Geology, Vol. 40, No. 9, pp. 779–782. doi:10.1130/G33330.1. Shakun, J. D.; Clark, P. U.; He, F.; Marcott, S. A.; Mix, A. C.; Liu, Z.; Otto-Bliesner, B.; Schmittner, A.; and Bard, E., 2012, Global warming preceded by increasing carbon dioxide concentrations during the last deglaciation: Nature, Vol. 484, pp. 49–55. Sharma, P.; Bourgeois, M.; Elmore, D.; Granger, D.; Lipschutz, M. E.; Ma, X.; Miller, T.; Mueller, K.; Rickey, F.; Simms, P.; and Vogt, S., 2000, PRIME lab AMS performance, upgrades and research applications: Nuclear Instruments and Methods in Physics Research Section B: Beam Interactions with Materials and Atoms, Vol. 172, No. 1–4, pp. 112–123. Sherard, J. L.; Woodward, R. J.; Gizienski, S. F.; and Clevenger, W. A., 1963, Earth and Earth-Rock Dams: John Wiley and Sons, New York, 725 p.

Stock, G. M. and Uhrhammer, R. A., 2010, Catastrophic rock avalanche 3600 years BP from el Capitan, Yosemite Valley, California: Earth Surface Process Landforms, Vol. 35, No. 8, pp. 941–951. doi:10.1002/esp.1982. Utah Automated Geographic Reference Center, State Geographic Information Database, 2006, 2-Meter LiDAR Data: Electronic document, available at http://gis.utah.gov/ data/elevation-terrain-data/2-meter-lidar Willenberg, H.; Eberhardt, E.; Loew, S.; McDougall, S.; and Hungr, O., 2009, Hazard assessment and runout analysis for an unstable rock slope above an industrial site in the Riviera valley, Switzerland: Landslides, Vol. 6, No. 2, pp. 111–116. doi:10.1007/s10346-009-0146-7. Wolter, A.; Gischig, V.; Stead, D.; and Clague, J. J., 2015, Investigation of geomorphic and seismic eﬀects on the 1959 Madison Canyon, Montana, landslide using an integrated ﬁeld, engineering geomorphology mapping, and numerical modelling approach: Rock Mechanics Rock Engineering, Vol. 49, No. 6, pp. 2479–2501. doi:10.1007/s00603-015-0889-5.

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 201–215

215

Microgravity Mapping of an Inception Doline Shaft System PETER J. HUTCHINSON* ALEXANDER BALOG THG Geophysics, Ltd., 4280 Old William Penn Highway, Murrysville, PA 15668

SHAD E. HOOVER CMT Laboratories, Inc., 2701 Carolean Industrial Drive, State College, PA 16801

Key Terms: Sinkhole, Doline, Subsidence, Grout Stabilization ABSTRACT Reactivation of an inception doline shaft system through anthropogenic actions, precipitation, and possibly seismic activity induced subsidence in a hospital emergency room that was under construction in State College, PA. The convergence of Tropical Storm Lee and Hurricane Irene is interpreted to have caused the building’s brick edifice to fall and induce vertical shifts in the reinforced concrete entrance floor slab. Microgravity mapping of the existing hospital emergency room entrance; the emergency room building under construction; and the parking lot in front of the emergency room entrance documented the presence of a doline shaft system (i.e., inter-connected sinkhole). Groundwork for the construction of the new emergency room included grading and leveling of the property. Surface water runoff entered the construction site from a parking lot that sloped toward the addition and to a non-functioning stormwater inlet. The grading for the new construction exposed an open fracture for surface runoff. Subsequent channeling of surface water to the conduit provided drainage for surface runoff, but it also initiated subsidence throughout the existing structure and the addition that was under construction. Engineering rehabilitation included a limited mobility (LM) grout program to plug subsoil fracture karren drainage systems and stabilize the surface. Drilling progressed in four stages, initially focusing on areas of greatest subsidence. In total, 60 injection points were completed to a mean depth of 24 m below grade in an area measuring approximately 370 m2 . During LM grouting, 867 m3 of a sand-and-cement grout mixture were injected to stabilize the area.

*Corresponding author email: pjh@thggeophysics.com

INTRODUCTION In March of 2011, Mount Nittany Medical Center (MNMC) started construction of an emergency room adjacent to the existing emergency room (Figure 1). By April, the grading and filling had been completed, and foundation excavations for the new addition commenced. Between August 28 and 29, 2011, Hurricane Irene dumped 1.7 cm of rain on the project site. Subsequently, Tropical Storm Lee dumped an additional 13.7 cm of rain from September 5 through 9, 2011 (Figure 2). On or before September 7, 2011, portions of the front façade collapsed, and the floor to the existing emergency room showed structural damage. On September 13, subsidence to the support columns for the new emergency room was also noted (Figure 1). Discussions with the construction crew revealed that during site-grading activities, an opening was discovered in the subsurface between the two buildings (Figure 3). Surface water was then channelized to the narrow opening by the construction crew to help clear the site of ponded water. The opening was probably an extension of a fracture/void from a subsurface karren (i.e., a rugose type of rock texture) in the underlying bedrock that had been exposed by the grading activities. GEOLOGY MNMC is located within Nittany Valley, in the northwestern margin of the Valley-and-Ridge Province (Fenneman, 1938; Faill and Nickelsen, 1999; Faill, 2000; and Miles and Whitfield, 2001). The Valley-and-Ridge Province is the western-most portion of the Alleghenian foreland fold-and-thrust tectonic belt (Nickelsen, 1988; Figure 4). The foreland fold-and-thrust belt is characterized by décollementbased deformation as part of the Allegheny Orogeny (Boyer and Elliott, 1982; Geiser and Engelder, 1983). In this style of deformation, an allochthonous mass of

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 217–225

217

Hutchinson, Balog, and Hoover

Figure 1. Digital elevation model (1 m grid; North American Vertical Datum of 1988; 2006–2008 Department of Conservation and Natural Resources PAMAP) showing inferred fractures in red. Inset is from Google Earth (2018) showing the footprint of the gravity study area (white), new hospital addition (red), and existing emergency room (green).

Figure 2. Daily precipitation for State College, PA, August through September 2011.

218

foreland rocks moves as a thrust sheet on a subhorizontal fault (décollement) outward from an orogenic core towards the craton (Chapple, 1978; Faill, 1997a, 1997b, 1998). At a depth of about 7 km below grade, the basal décollement consists of the Cambrian Potsdam Formation detaching from the Cambrian Tomstown Formation (Figure 5). The shallowest décollement is the top of the Ordovician Reedsville Formation with the base of the Cambrian Potsdam Formation (Berg et al., 1980). This thrust sheet brought the Nittany Dolomite to the surface as a gentle fault-bend fold anticline (Figure 5). Due to stacking of the carbonate suites on two décollements, there is an estimated 7,000 m section of carbonate rock beneath the MNMC (Demicco and Mitchell, 1982; Gold et al., 2017). The formation of fault-bend fold anticlines further deformed the carbonates by creating a stress environment conducive to carbonate dissolution and fracturing. Within the footprint of the MNMC, periglacial clayrich silt and sand unconformably overlie the Lower Ordovician Nittany Dolomite (Parizek et al., 1971). The Nittany Dolomite is a medium- to dark-gray, thick-bedded dolomite containing chert and siliceous

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 217–225

Microgravity Mapping of a Sinkhole

Figure 3. Fracture discovered during grading operations. Note the iron staining on the wall of the fracture.

oolites with an estimated thickness in this area of 360 m. The Nittany Dolomite is characterized by having large-scale drawdown dolines (for example, State College’s Memorial Stadium is a partially exhumed doline). The Nittany Dolomite conformably overlies the Stonehenge Limestone, which is aphanitic to ﬁne grained, argillaceous and dolomitic in part, and contains abundant ﬂat pebble conglomerate lenses. The Ordovician Stonehenge Limestone is approximately 180 m thick and is only locally karst forming. Conformably underlying the Lower Ordovician Stonehenge Limestone are the Upper Cambrian Gatesburg and Warrior Formations. Both formations are karst-forming dolomites with a total thickness of 850 m. Karst Evolution Deeply buried limestones can undergo metasomatic processes (dolomitization) that rearrange and enlarge pores (Wilson, 1975; Sibley, 1982; and Hem, 1989). During the formation of dolomite, the net rock volume of limestone decreases by 13 percent, leaving voids and vugs that can induce further dissolution through in-

Figure 4. Geologic map and section (Berg et al., 1980). Note geologic colors and descriptions differ between the two images. No vertical exaggeration to the profile. Map key: Precambrian granitic gneiss (gn): Cambrian Tomstown (–Cl), Warrior (–Cw),;Gatesburg (–Cgm and –Cgl): Ordovician Stonehedge/Lark (Osls); Nittany (On); Axeman (Oa); Bellefont (Obf); Brenner (Obv); Coburn (Ocn); Reedsville (Or); Bald Eagle (Obe); Juniata (Oj): Silurian Shawangunk (Ss); Clinton Group (Sc). Profile key: Precambrian granitic gneiss (gn): Cambrian Tomstown (–Cwh); Potsdam (–Cpl); Gatesburg (–Cga): Ordovician Stonehedge/Lark, Nittany, Axeman, and Bellefonte (Osls); Brenner and Coburn (Oca); Reedsville (Orm); Bald Eagle (Obe); Juniata (Oj); Queenston (Oqa): Silurian Shawangunk and Clinton Group (Sbis).

creased permeability (Hohlt, 1948). 2CaCO3 + Mg2+ ↔ CaMg(CO3 )2 + Ca2+ .

(1)

One important aspect of converting limestone to dolomite is that the increase in porosity and permeability within the rock creates excellent reservoirs for oil and gas production (Levorsen, 1967; Longman, 1982). Current theories show that karst forms through two mechanisms, hypogene and epigene speleogenesis (Klimchouk, 2018). Klimchouk (2007, 2014) documented hypogene speleogenesis, or vertically upward migration of groundwater during the course of depositional history, as a mechanism for the creation of karst topography. Remanent magnetization of ferromagnesium minerals during Pennsylvanian–Permian deformation has been attributed to the vertically upward percolation of orogenic ﬂuids through the carbonate units within the Arbuckle Mountains (Nick

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 217–225

219

Hutchinson, Balog, and Hoover

Figure 5. Proﬁle through Nittany Mountain with indeterminate location of the upper décollement (1) and location where it is theoretically exposed (2) at the surface (from Gold et al., 2017). MNMC is the Mount Nittany Medical Center.

and Elmore, 1990) and Valley-and-Ridge Province of central Pennsylvania (Mathur et al., 2008). Two types of karst systems are recognized in the world today: epigenic and hypogenic (Klimchouk, 2007). Regional groundwater flow systems provide the systematic transport and distribution mechanisms needed to produce and maintain the disequilibrium conditions necessary for speleogenesis. Epigenic karst systems exist in unconfined groundwater conditions and are predominantly local systems, whereas hypogenic karst is associated with discharge regimes of regional flow systems. Hypogenic speleogenesis occurs within confined groundwater settings and may lose the confining groundwater conditions due to uplift. The confined cave systems can be further modified under unconfined groundwater conditions through epigenic processes. The primary criteria for identifying hypogenic caves are morphological and hydrogeological features. Hypogenic caves have network mazes, spongework mazes,

220

irregular chambers, and isolated passages or crude passage clusters. They often combine to form composite patterns and complex three-dimensional structures. Hypogenic caves have an areal coverage that is ﬁve times greater than in epigenic karst systems (Klimchouk, 2007). Hypogenic speleogenesis commonly results in more isotropic conduit permeability within highly karstiﬁed areas measuring up to several square kilometers. Hypogenic speleogenesis contributes to hydrothermal mineralization, diagenesis, and hydrocarbon transport and entrapment. Hypogenic processes have contributed to the inception dolines within the study area. Deformation in the form of a fault-bend fold anticline increased permeability and provided more opportunity for metasomatic processes. Rauch and White (1970) noted that most caves in the Nittany Valley occur within limestone epigene karst and that cave development in dolomite is extremely rare. However, dolines are common in the Nittany Dolomite, which points to a

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 217–225

Microgravity Mapping of a Sinkhole

Figure 6. A profile of an inception doline filled with periglacial soil and showing karren structures (modified from Sauro, 2012). Figure is not to scale.

deep burial environment in a hypogene environment (Figure 6). Hearn et al. (1987) noted that during the Alleghenian Orogeny, regional potassic alteration in unmineralized carbonate rocks and mineralization in carbonate and other sedimentary deposits were related to deep brines migrating towards the basin margin under a hydraulic gradient. Also, late Paleozoic orogenic fluid expulsion from the evolving Appalachian-Ouachita Orogen increased vertically upward migration of orogenic fluids (Oliver, 1986; Duane and de Wit, 1988; and Bethke and Marshak, 1990). Consequently, fluid movement of orogenic fluids through fractures and faults resulted in the development of hypogene karst. Further, Mathur et al. (2008), in their work with fluid inclusions, noted that late Paleozoic–aged, vertically upward, low-temperature hydrothermal migration deposited sulfide minerals (i.e., epithermal pyrite) in fractures and faults (note the staining on the wall of the fracture in Figure 2). Later, during the Cenozoic, it is speculated that the Chesapeake Bay bolide or meteor impact event triggered overprinting of the late Paleozoic mineralization through vertically upward migration of high-temperature fluids (Mathur et al., 2015). Consequently, several mechanisms can be attributed to the vertically upward migration of water during episodic mountain-building events. Consequently, the hypogenic paradigm of the movement of orogenic fluids, both low and high temperature, left deep-seated voids.

Microgravity Theory Microgravity measurements are not readily impacted by cultural noise; consequently, microgravity measurements can be collected in buildings and adjacent to urban development (Milsom, 1989). Microgravity has been used for many geologic purposes; however, for the environmental geophysicist, microgravity is used to determine the presence of subsurface voids, to image subsurface bedrock topography, and to find the depth of waste (Carmichael and Henry, 1977; Stewart, 1980; and Kick, 1985). Small changes in rock density produce small changes in the gravity field that can be measured by the microgravimeter (Hinze, 1990). These readings change from day to day due to tidal response and lunar pull, among other phenomena, that have an impact on Earth’s gravitational flux. Processing raw gravity data includes corrections for latitude, elevation, Bouguer gravity, tidal action, and terrain. A microgravimeter measures the acceleration due to Earth’s gravitational field in meters per second squared (m/s2 ) using an astatic spring mechanism (Carmichael and Henry, 1977). Earth’s gravitational field is roughly equivalent to a sphere, with variations for sea level and elevation (Nettleton, 1976; Woollard, 1975; and Milsom, 1989). The 1980 International Gravity Formula (Moritz, 1980) for calculating absolute gravity is: gφ = go 1 = αsin2 φ − βsin2 2φ ,

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 217–225

(2)

221

Hutchinson, Balog, and Hoover

where gφ is the theoretical acceleration due to gravity at a given latitude, and α and β are constants that depend on the amount of ﬂattening of the spheroid and upon the speed of rotation of Earth (Kaufman, 1992; Reynolds, 2011). Gravity is calculated in units of acceleration, and so the SI unit for the pull of gravity is meters per second squared (m/s2 ). The International Gravity Formula was reﬁned to the Geodetic Reference System 1967 (Woollard, 1975) and was derived (Moritz, 1980) thus: gφ (1980) − gφ (1967) = (0.8316 + 0.0782sin2 φ − .0007sin4 φ)μm/s2 .

(3)

Latitude corrections are automatically corrected in the microgravimeter by subtracting the International Gravity Formula normal datum from the observed gravity measurement: 8.12 sin 2L , (4) km where Gl is the theoretical local gradient; L is the latitude; and km is kilometers. The elevation or free-air correction normalizes the gravity data to a given datum that does not have to be sea level. Free-air correction is based upon the free-air correction of 3.0855 μm/s2 (ASTM, 2018). The normal elevation (h) adopted for this survey was 350 m above mean sea level (amsl), and elevation changes above this were corrected as: Gl =

g = g (R) × 2h/R,

(5)

where the change in gravity ( g) based upon elevation was derived from Earth’s radius (R = 6,378 km) and normal gravitational ﬁeld g (9.80 m/s2 ) when corrected for h. Bouguer corrections (b) account for the rock mass between the measuring station and sea level and are based upon: b = 2πρgs h,

(6)

where Bouguer gravity is related to density (ρ = 2.54 mg/m3 ) and known thickness (h) above sea level. Microgravity Field Survey The CG-5 microgravimeter (Scintrex, Concord, Ontario, Canada) contains automated features that signiﬁcantly reduce the possibility of reading errors. To minimize drift, a drift calibration was conducted for over 24 hours. After completion of the drift calibration, ﬁeld readings repeated to within a standard deviation of 0.05 μm/s2 . The CG-5 was utilized to collect relative measurements that were compared to the gravitational attraction at a base station with an assumed absolute value

222

Figure 7. Diﬀerential gravity map with volumetric grout (GV) overlay. The grout injected into borings is displayed as colored dots based upon the total volume injected. Gravity measurement points are black dots; the margin of the existing emergency room building is a green line; the margin of the new emergency building is a red line; and the black squares are support columns for the two buildings. A forward modeling and inversion program, based upon the post-processed microgravity values, was used to develop the depth proﬁle for section A-A’.

for gravity (Long and Kaufmann, 2013). A base station was established in the neighboring parking lot and re-occupied every 4 hours (Figure 7). Data points were collected on an approximately orthogonal three-meter grid pattern; however, during the field work, the survey was adjusted after the anomaly was acquired to emphasize the anomaly’s dimensions. Post-processing of all measurements is necessary prior to computing the differential gravity measurement for a particular point. The CG-5 applied an automatic gravitational tidal correction to all data based upon the diurnal variation in Earth’s position to the moon and sun. Further, measurement point elevations were collected with a differential global positioning system (Trimble GEO-7XH) and post-processed for accurate (∼5 cm) terrain elevation. Terrain elevation values were then applied to the background elevation of 350 m amsl for the free-air calculation.

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 217–225

Microgravity Mapping of a Sinkhole

Figure 8. Pseudo–top-of-rock map based on gravity inversion modeling. Dots represent microgravity records. Note karren structures within the anomaly. All axes are in meters.

The difference between the post-processed base station and collection point value is the differential gravity for a measurement point. The differential gravity data were used to make a plan map that illustrates the difference between what is expected and what is actually measured, since the presence of voids and lowdensity earthen materials lowers Earth’s gravitational attraction (Figure 7). Processed gravity data were integrated into a forward modeling and inversion program (IX2D-GM; Interpex, Golden, CO) to portray a two-dimensional model of the doline (Figure 7). Unfortunately, the model does not show the open fracture in the limestone or the karren structures, but it does provide a two-dimensional conceptual model of the doline (Figure 7).

a deep-seated fracture-based void (Figure 7). The deep-seated void evolved through dissolution from structural deformation and from vertically upward migration of cold and hot ﬂuids, probably saline in nature. The microgravity survey identiﬁed the void as an elongate anomaly, approximately 30 m long and 10 m wide (370 m2 ), which was interpreted to be a fractureinduced doline. Although the microgravity survey was too coarse to see individual pinnacles, karren features were evident in the microgravity model (Figure 8). This fracture-induced doline is oriented north 40o east and is likely based upon deep-seated en echelon fractures. Surface topography supports the presence of an en echelon conjugate fracture system (Figure 1).

DISCUSSION

The State College, PA, area and speciﬁcally the MNMC are underlain by several thousand feet of carbonates. During construction of the new emergency room, a fracture in the subsurface was

The discharge of surface runoﬀ into the open fracture stimulated subsidence of the periglacial soil into

CONCLUSION

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 217–225

223

Hutchinson, Balog, and Hoover

exposed. Excessive precipitation in late August and early September of 2011 was channeled to the fracture. Migration of runoff vertically downward through a fracture also mobilized the overburden in the area of the new emergency room to migrate vertically downward, creating a subsidence event for the former building and new building. Over 100 microgravity readings were collected, processed, and compared to the processed base station reading to create a differential value. Mapping the differential microgravity readings identified the anomaly as an incipient doline. An incipient doline is generated by the convergence of water within a pre-existing structure, such as a fracture. To correct the ongoing subsidence, borings were placed in locations thought to help prevent subsidence. LM grouting through 60 injection points, completed to a mean depth of 24 m below grade over the 370 m2 area, stabilized the subsidence. However, since LM grouting occurred in four temporal stages, it was difficult to determine where the grout was placed three-dimensionally, since many of the injection points were connected through fractures. Ultimately, approximately 867 m3 of a sand-and-cement grout mixture were injected into the subsurface to stabilize the area. The grouting program stabilized the surface, and thus construction continued until the building was completed. The new emergency room to the hospital is now in full use. ACKNOWLEDGMENTS The authors thank Heather L. Krivos, GIT, and Kate S. McKinley, PG, at THG Geophysics, Ltd., for their help with the field work. Further, we thank Whitney E. Greenawalt, PE, CMT Laboratories, Inc., for providing technical assistance after the field work was completed. We would also like to thank the three reviewers for their help: Mustafa Saribudak, David W. Abbott, and David Bieber. REFERENCES ASTM, 2018, Standard Guide for Using the Gravity Method for Subsurface Investigation, Vol. 04.09 Soil and Rock (II): ASTM International, Conshohocken, PA, 10 p. Berg, T. M.; Edmunds, W. E.; Geyer, A. R.; Glover, A. D.; Hoskins, D. M.; Maclachlan, D. B.; Root, S. I.; Sevon, W. D.; and Socolow, A. A., compilers, 1980, Geologic Map of Pennsylvania, 2nd ed.: Pennsylvania Geological Survey 4th Series, Map 1, 3 sheets, scale 1:250,000. Bethke, C. M. and Marshak, S., 1990, Brine migrations across North America—The plate tectonics of groundwater: Annual Review of Earth and Planetary Sciences, Vol. 18, No. 1, pp. 287–315. Boyer, S. E. and Elliott, D., 1982, Thrust systems: AAPG Bulletin, Vol. 66, No. 9, pp. 1196–1230.

224

Carmichael, R. S. and Henry, G., 1977, Gravity exploration for groundwater and bedrock topography in glaciated areas: Geophysics, Vol. 42, No. 4, pp. 850–859. Chapple, W. M., 1978, Mechanics of thin-skinned fold-and-thrust belts: Geological Society of America Bulletin, Vol. 89, No. 8, pp. 1189–1198. Demicco, R. V. and Mitchell, R. W., 1982, Facies of the Great American Carbonate Bank in the Central Appalachians, in Lyttle, P. T. (Editor), Central Appalachian Geology: Northeast and Southeast Sections Field Trip Guidebook: Geological Society of America, Boulder, CO, pp. 171–266. Duane, M. J. and de Wit, M. J., 1988, Pb-Zn ore deposits of the northern Caledonides: Products of continental-scale ﬂuid mixing and tectonic expulsion during continental collision: Geology, Vol. 16, No. 11, pp. 999–1002. Faill, R. T., 1997a, A geologic history of the north-central Appalachians; Part 1, Orogenesis from the Mesoproterozoic through the Taconic Orogeny: American Journal of Science, Vol. 297, No. 6, pp. 551–619. Faill, R. T., 1997b, A geologic history of the north-central Appalachians; Part 2, The Appalachian Basin from the Silurian through the Carboniferous: American Journal of Science, Vol. 297, No. 7, pp. 729–761. Faill, R. T., 1998, A geologic history of the north-central Appalachians; Part 3, The Alleghany Orogeny: American Journal of Science, Vol. 298, No. 2, pp. 131–179. Faill, R. T., 2000, Reverse Sequence of Alleghany Fold-and-Thrust Tectonics, North-Central Appalachians: Pennsylvania Geological Survey 4th Series, 34 p. Faill, R. T. and Nickelsen, R. P., 1999, Chapter 19: Appalachian Mountain section of the Ridge and Valley Province, in Shultz, C. H. (Editor), The Geology of Pennsylvania: Pennsylvania Geological Survey, Murrysville, PA, p. 888. Fenneman, N. M., 1938, Physiography of Eastern United States: McGraw-Hill, New York, NY. Geiser, P. and Engelder, T., 1983, The distribution of layer parallel shortening fabrics in the Appalachian foreland of New York and Pennsylvania: Evidence for two non-coaxial phases of the Alleghanian Orogeny. In Hatcher, R. D. (Editor), Contributions to the Tectonics and Geophysics of Mountain Chains: Memoir 158, Geological Society of America, Boulder, CO, pp. 161–176. Gold, D. P.; Miller, C. E., Jr.; and Engelder, T., 2017, Smallscale fault-bend fold in the Appalachian Valley and Ridge Ordovician—Salona to Linden Hall section. In Anthony, R. (Editor), Recent Geologic Studies and Initiatives in Central Pennsylvania; 82nd Annual Field Conference of Pennsylvania Geologists, October 5–7, 2017; State College, PA, pp. 41–46. Hearn, P. P.; Sutter, J. F.; and Belkin, H. E., 1987, Evidence for late-Paleozoic brine migration in Cambrian carbonate rocks of the central and southern Appalachians: Implications for Mississippi Valley–type sulﬁde mineralization: Geochimica et Cosmochimica Acta, Vol. 51, No. 5, pp. 1323–1334. Hem, J. D., 1989, Study and Interpretation of the Chemical Characteristics of Natural Water: U.S. Geological Survey WaterSupply Paper 2254, 263 p. Hinze, W. J., 1990, 4. The role of gravity and magnetic methods in engineering and environmental studies. In Ward, S. H. (Editor), Geotechnical and Environmental Geophysics: Society of Exploration Geophysicists, Tulsa, OK, pp. 75–126. Hohlt, R. B., 1948, The nature and origin of limestone porosity: Quarterly Colorado School of Mines, Vol. 43, No. 4, p. 51. Kaufman, A. A., 1992, Geophysical Field Theory and Method: Part A, Gravitational, Electric, and Magnetic Fields: Academic Press, San Diego, CA, 581 p.

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 217–225

Microgravity Mapping of a Sinkhole Kick, J. F., 1985, Depth to bedrock using gravimetry: The Leading Edge, Vol. 4, No. 4, pp. 38–42. Klimchouk, A. B., 2007, Hypogene Speleogenesis: Hydrogeological and Morphogenetic Perspective: Special Paper 1, National Cave and Karst Research Institute, Carlsbad, NM, 106 p. Klimchouk, A. B., 2014, The methodological strength of the hydrogeological approach to distinguishing hypogene speleogenesis. In Klimchouk, A. B.; Sasowsky, I.; Mylroie, J.; Engle, S. A.; and Engle, A. S. (Editors), Hypogene Cave Morphologies: Karst Waters Institute, Leesburg, VA, pp. 4–12. Klimchouk, A. B., 2018, Advances in understanding hypogene karst. In Staﬀord, K. W. and Veni, G. (Editors), Hypogene Karst of Texas: Monograph 3, Texas Speleological Survey, Austin, TX, pp. 3–15. Levorsen, A. I., 1967, Geology of Petroleum: W. H. Freeman Co., New York, NY, 724 p. Long, L. T. and Kaufmann, R. D., 2013, Acquisition and Analysis of Terrestrial Gravity Data: Cambridge University Press, Cambridge, U.K., 171 p. Longman, M. W., 1982, Carbonate diagenesis as a control on stratigraphic traps. In Proceedings of the 1981 AAPG Education Conference in Calgary, Canada: American Association of Petroleum Geologist Department of Education, Course Note Series No. 21, Tulsa OK, 159 p. Mathur, R.; Mutti, L.; Barra, F.; Gold, D.; Smith, R. C.; Doden, A.; Detrie, T.; Ruiz, J.; and McWilliams, A., 2008, Fluid inclusion and Re-Os isotopic evidence for hot Cenozoic mineralization in the central Pennsylvanian Valley and Ridge Province: Mineralogy and Petrology, Vol. 93, No. 3–4, pp. 309–324. Mathur, R.; Gold, D. P.; Ellsworth, C. J.; Doden, A.; Wilson, M.; Ruiz, J.; Shcheetz, B. E.; and Herman, G. C., 2015, Re-Os isotope evidence of Early Tertiary crustal faulting and sulﬁde-mineralization in Pennsylvania with probable ties to the Chesapeake Bay bolide impact in Maryland, USA. In Herman, G. C. and Ferguson, S. M. (Editors), Neotectonics of the New York Recess; Geological Association of New Jersey XXXII Annual Conference and Field Trip: Geological Association of New Jersey, Trenton, NJ, pp. 68–79. Miles, C. E. and Whitfield, T. G., compilers, 2001, Bedrock Geology of Pennsylvania, Edition 1.0 [digital map]: Pennsylvania Bureau of Topographic and Geologic Survey, Department of Conservation and Natural Resources, Harrisburg, PA, scale 1:250,000. Milsom, J., 1989, Field Geophysics: Open University Press, New York, NY, 182 p.

Moritz, H., 1980, Geodetic Reference System 1980: Journal of Geodesy, Vol. 74, No. 1, pp. 128–133. Nettleton, L. L., 1976, Gravity and Magnetics in Oil Prospecting: McGraw-Hill, New York, NY, 464 p. Nick, K. E. and Elmore, R. D., 1990, Paleomagnetism of the Cambrian Royer Dolomite and Pennsylvanian Collings Ranch Conglomerate, southern Oklahoma: An early Paleozoic magnetization and nonpervasive remagnetization by weathering: Geological Society of America Bulletin, Vol. 102, No. 11, pp. 1517–1525. Nickelsen, R. P., 1988, Structural evolution of folded thrusts and duplexes on a ﬁrst-order anticlinorium in the Valley and Ridge Province of Pennsylvania. In Mitra, G. and Wojtal, S. (Editors), Geometries and Mechanics of Thrusting with Special Reference to the Appalachians: Special Paper 222, Geological Society of America, Boulder, CO, pp. 89–106. Oliver, J., 1986, Fluids expelled tectonically from orogenic belts: Their role in hydrocarbon migration and other geologic phenomena: Geology, Vol. 14, No. 2, pp. 99–102. Parizek, R. R.; White, W. B.; and Langmuir, D., 1971, Hydrogeology and Geochemistry of Folded and Faulted Rocks of the Central Appalachian Type and Related Land Use Problems: Circular 82, The Pennsylvania State University College of Earth and Mineral Sciences, University Park, PA, 210 p. Rauch, H. W., and White, W. B., 1970, Lithologic controls on the development of solution porosity in carbonate aquifers: Water Resources Research, Vol. 6, No. 4, pp. 1175–1192. Reynolds, J. M., 2011, An Introduction to Applied and Environmental Geophysics: Wiley-Blackwell, New York, NY, 796 p. Sauro, U., 2012, Closed depressions in karst areas. In Culver, D. C. and White, W. B. (Editors), Encyclopedia of Caves: Academic Press, Amsterdam, pp. 140–155. Sibley, D. F., 1982, The origin of common dolomite fabrics: Clues from the Pliocene: Journal of Sedimentary Research, Vol. 52, No. 4, pp. 1087–1100. Stewart, M. T., 1980, Gravity survey of a deep buried valley: Ground Water, Vol. 18, No. 1, pp. 24–30. Wilson, J. L., 1975, Carbonate Facies in Geologic History: Springer, Berlin, 472 p. Woollard, G. P., 1975, Regional changes in gravity and their relation to crustal parameters. Bureau Gravimetrique International Bulletin d’Information, Vol. 36, pp. 106–110.

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 217–225

225

Sisal Fiber-Polymer–Treated Sand Mechanical Properties in Triaxial Test LILIN WU WEI QIAN JIN LIU* ZEZHUO SONG School of Earth Sciences and Engineering, Hohai University, Nanjing, 210098, China

DEBI PRASANNA KANUNGO CSIR–Central Building Research Institute (CBRI), Roorkee 247667, India

YUXIA BAI FAN BU School of Earth Sciences and Engineering, Hohai University, Nanjing, 210098, China

Key Terms: Mechanical Properties, Soil Stabilization, Polymer, Failure Mode, Triaxial Test ABSTRACT Natural sisal fiber is an environment-friendly and efficient material for soil reinforcement. Many studies have reported that the shear strength of soil has been improved by the addition of fiber. However, the mechanical properties of sand can be more effectively improved by the incorporation of water-based polymer and sisal fiber. An extensive laboratory testing program was conducted to determine the effect of water-based polyurethane and sisal fiber reinforcement on sand. Laboratory tests included sieve analysis, X-ray diffraction, conventional triaxial compression, and scanning electron microscopy (SEM) tests. The effects of polymer content (PC), fiber content (FC), fiber length (FL), and sample dry density (ρ) are thoroughly investigated. The results indicate that the increases of PC, FC, and ρ all improve the mechanical properties of sand. For FL, this improvement in shear strength was maintained to FLs of up to 18 mm. Beyond 18 mm, the shear strength decreased with further increase in FL. The mixing of polymer and fiber changes the failure mode from shear faulting to ductile flow. This indicates that the ductility of sand is improved. From the SEM images we found that sisal fibers, binding with colloidal materials formed by polymer, fill the sand voids and join the sand particles. This demonstrates that mixing of fiber and polymer can enhance the bonding of sand particles.

*Corresponding author email: jiuliu920@163.com

INTRODUCTION Sand is widely used in engineering projects. However, unconsolidated sand causes the differential settlement of buildings, foundation sliding, liquefaction failure, etc. (Shogaki and Kaneda, 2013; Keramatikerman and Chegenizadeh, 2017). To improve the mechanical properties of sand, much research has been conducted (Van et al., 2015). Although some methodologies are effective, they may have irretrievable influences on the environment. Liu et al. (2012) focus on the sand stabilization using environment-friendly material. Sand stabilization is generally categorized into physical-mechanical, chemical, and biological approaches (Hataf et al., 2018). Among these, the physical-mechanical approach is the oldest, dating from thousands of years ago when ancients used a mixture of straw and sand to enhance the stability of buildings. This practice continues in recent years with geosynthetics (Hamidi and Hooresfand, 2013). Unlike the traditional geosynthetics-reinforced soil, fiber-reinforced soil is a kind of geosynthetic that is formed by fully mixing fiber silk or fiber mesh with the soil material in certain proportions. As a result of the advantages of good dispersion and easy mixing, fiber-reinforced soil is usually regarded as homogeneous isotropic material (Tang et al., 2007). As an effective soil stabilization/modification technology, the improvement of physical properties of short fiberreinforced soil has been extensively studied (Yetimoglu and Salbas, 2003; Jiang et al., 2010; Davarci et al., 2014; and Gumuser and Senoi, 2016). Sand is generally considered to have no cohesive strength due to the fact that the intergranular surface force between its particles can be neglected when compared to its weight. Thus, traditional chemical stabilization is the

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 227–242

227

Wu, Qian, Liu, Song, Kanungo, Bai, and Bu

most common approach and uses Portland cement, lime, fly ash, asphalt, and other chemicals (Pedersen and Hansen, 2003; Tingle et al., 2007). Many studies have shown that these traditional stabilizers improve the engineering properties of sand significantly. However, these stabilizers also increase the stiffness and brittleness of sand, causing sudden instability after deformation (Marri et al., 2001; Schnaid et al., 2001; and Consoli et al., 2011). To improve the ductility of cement-reinforced sand, Park (2011) and Kutanaei and Choobbasti (2016) have added fibers to reduce the post-peak softening behavior. Most chemical stabilizations have irreversible effects on the environment (Liu et al., 2011). The cement surface reflects radiant heat and noise, aggravating the thermal effect and noise pollution of the city. Cementing also reduces the chances of microbial life in the sand, thus destroying surface ecology and reducing the chance of organic matter replenishment. As alternatives, organic polymer materials such as resins, xanthan gum, polyacrylamide, and polyurethane can be used as new stabilizers; these materials confer good elasticity and high strength and impart little impact on the environment (Naeini and Ghorbanalizadeh, 2010; Mohsin and Attia, 2015). Replacement of traditional chemical stabilizers with polymers has become an important research topic in recent years. Some studies (Inbar et al., 2015; Choi et al., 2016) have shown that the polymer treatment improves the mechanical properties of sand significantly. The introduction of polymer makes the sand body coherent. Therefore, it is interesting to study whether the fiber polymer mixture can further enhance the stabilization effect. In this study, we investigate the stabilization effects of the combined use of sisal fiber and water-based polyurethane. We measure the shear strength, energy absorption, failure mode, and sand microstructure under triaxial testing to understand the stabilization process involved. TESTING PROGRAM Forty-five unconsolidated and un-drained triaxial shear tests were performed. Polyurethane content (PC), sisal fiber content (FC), dry density (ρ). and sisal fiber length (FL) were varied within the testing program.

Table 1. Physical properties of the sand. Parameter

Description

Origin Color Void ratio Dry density (g/cm3 ) Specific gravity (g/cm3 ) Coefficient of uniformity (Cu ) Coefficient of curvature (Cc ) Clay content Mineral composition

River-bed Light-yellow 0.57–0.91 1.39–1.69 2.65 3.33 1.30 0 Quartz, K-feldspar, and plagioclase

1a shows the X-ray diﬀraction diagram of the sand, from which it can be seen that the sand is composed of quartz, K-feldspar, and plagioclase, and the sand is of the quartz-rich type. The gradation curve of the sand is shown in Figure 1b, and its physical properties are listed in Table 1. The base sand particles have an effective grain size (d10 ) of about 0.12 mm, constrained grain size (d60 ) of about 0.30 mm, and average particle size (d50 ) of about 0.25 mm. Polyurethane Organic Polymer The light-yellow transparent and soluble polyurethane organic polymer presented in Figure 2a was used as a cementing agent in the tests. The structural formula is shown in formula 1, and its properties are listed in Table 2. The polymer has a main component of polyurethane resin that contains excessive isocyanate (–NCO). Isocyanate reacts with water to form unstable carbamate (–NHCOOH), which is then decomposed into carbon dioxide (CO2 ) and amine (–NH2 ). In the presence of excess isocyanate, the resulting amine reacts with isocyanate to form urea (–NHCONH–) and denotes as formulas 2 and 3, as follows: O = C = N(R1 − NH − CO − R2 − O − CO − NH)n R1 − N = C = O; (1) O = C = N − R − N = C = O + 2H2 O → HO − CO − NH − R − NH − CO − OH → H2 N − R − NH2 + 2CO2 ; (2) Table 2. Properties of the polyurethane organic polymer.

Materials Sand The sand from the bed of Yangzi River (Nanjing area, China) was used as samples after they were airdried and sieved at 2 mm in the laboratory. Figure

228

Parameter Speciﬁc gravity (g/cm3 ) Mass fraction (%) Viscosity (MPa/s) pH Solidiﬁcation time (s)

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 227–242

Value 1.15 75 ± 1 650–700 7 30–1,800

Sisal Fiber-Polymer–Treated Sand Mechanical Properties

Figure 1. (a) The X-ray diﬀraction diagram of the sand; (b) Gradation curve of the sand.

Table 3. Physical characteristics of the sisal fiber. Parameter

Value 3

Speciﬁc gravity (g/cm ) Average tensile strength (MPa) Elasticity modulus (MPa) Elongation at break (%)

0.96 15.3 456.4–559.8 3.02–3.04

(n + 1) H2 N − R − NH2 + nO = C = N − R − N = C = O → H2 N ( R − NH − CO − NH )2n R − NH2 , (3) where R1 is and R2 is poly-oxypropylene diol or poly-oxyethylene glycol. Sisal Fiber White monofilament sisal fibers with a diameter of 0.25 mm and an elliptical cross section were randomly distributed in the sample. Table 3 lists the physical characteristics of the sisal fiber shown in Figure 2b. In order to avoid the variations in material proper-

ties, sisal fibers were selected from the same production site and batch in Guangxi, China. Sisal fiber has high strength, good elasticity, low elongation, and a large elastic modulus. The ratio of sisal fiber weight to dry sand weight is considered as fiber content (FC) and is denoted by formula 4, as follows: (4) FC = W f /Ws × 100%, where Wf is sisal fiber weight (in grams) and Ws is dry sand weight (in grams). Experimental Methods Unconsolidated-Undrained Triaxial Tests In sample preparations, the desired content of sisal fiber was mixed with a requisite amount of dry sand by hand for uniform distribution of fiber. About 10 percent water (to dry sand weight) was poured into the desired content of polymer solution and stirred continuously to obtain a uniform polymer dilution. Polymer dilution was then added to the fiber-sand mixture and stirred continuously to obtain the final mixture (Figure 2c). Finally, the samples were prepared with the static

Figure 2. Photos of (a) polyurethane organic polymer; (b) sisal ﬁber; (c) ﬁber-polymer-sand mixture; (d) computerized triaxial tester.

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 227–242

229

Wu, Qian, Liu, Song, Kanungo, Bai, and Bu Table 4. Description of variables in present study.

Variable

No. of Levels

Type of soil

Type of polymer

Type of ﬁber Polymer content

1 5

Fiber content

Fiber length Dry density Water content Sample size

5 5 1 1

Curing condition

Description of Samples Poorly graded sand from the bed of Yangzi River Light-yellow transparent polyurethane organic polymer White monoﬁlament sisal ﬁbers 0%, 1%, 2%, 3%, 4% dry weight of sand 0.0%, 0.2%, 0.4%, 0.6%, 0.8% dry weight of sand 6, 12, 18, 24, 30 mm 1.40, 1.45, 1.50, 1.55, 1.60 g/cm3 10% weight of sand 39.1-mm diameter and 80.0-mm height, compacted in four layers Cured for 48 hours in incubator chamber at 20°C ± 2°C

compaction method based on ASTM D2850. The final mixture was divided into quarters; each of these mixtures was poured into a split mold with 39.1-mm diameter and 80.0-mm height and compacted by constant power until the requisite height was reached. For samples with different dry densities, they have different weights and the same volume. Samples of different dry densities can be obtained by pressing mixtures of different weights to a specified height. Each sample was kept in an incubator chamber at 20°C ± 2°C for 48 hours. Triaxial shear tests under unconsolidated and undrained conditions were performed. The variables considered in the triaxial shear tests are listed in Table 4. The samples were tested in a computerized system (shown in Figure 2d) at confinements of 100, 200, 300, and 400 kPa. The samples in the triaxial cell were compressed to a required confining pressure, and the shear rate of the sample remained at 0.8mm/min. Cell pressure and stress-strain are measured by a data acquisition system during the shear process.

Scanning Electron Microscope (SEM) The mechanisms of polyurethane-, ﬁber-, and polyurethane composite-treated sand were studied by SEM. The samples (FC = 0, 0.4 percent; PC = 4 percent; FL = 18 mm; ρ = 1.50 g/cm3 ) after unconsolidated-undrained triaxial shear tests were divided into 1,000-mm3 cubes. Before testing, gold coating was applied to improve the conductivity after dehydration treatment. The microscopic images of polyurethane- and sisal ﬁber-treated sand were used to study its microstructure.

230

RESULTS AND DISCUSSION All samples underwent excessive elastic deformation, plastic deformation, and ultimate failure in the shear process. The whole structure of the specimen was gradually destroyed, and the relative displacements of particles occurred when shear stress reached its peak. As the shear process proceeds, the particles in the damaged specimen are rearranged under certain confining pressure, which makes the reduced shear stress rise again. At this time, the specimen has been destroyed. Thus, the first peak in the stress-strain curve is considered as the failure point of the sample. The damaged stress circles under different confinements were plotted in τ-σ stress plane and the shear strength parameters were obtained. Table 5 summarizes the variables used, the peak deviatoric stresses, energy absorption, friction angles, and cohesion. The formula of deviatoric stress (q) is denoted within formula 5: q = σ1 − σ3 ,

(5)

where σ1 (kPa) is the maximum principal stress and σ3 (kPa) is the minimum principal stress. Stress-Strain Characteristics and Peak Strength Figure 3 shows the relation curves between deviatoric stress and axial strain for different PC, FC, FL, and ρ values under a confining pressure (CP) of 100 kPa. When there is a peak value, take the peak value on the curve as the failure point; when there is no peak value, take the deviatoric stress corresponding to 15 percent axial strain as the failure point. The peak stress and failure point of unreinforced sand were obvious in the stress-strain curves. It can be seen from Figure 3a that the deviatoric stress increased with the increase in polyurethane content without introduction of fiber. Figure 3b shows that for a given fiber content and CP, the deviatoric stress improved as polyurethane content increased and the axial strain at the yield point increased slightly. The polyurethane has a great effect on the shear strength of sand. When polyurethane content ranged from 0 percent to 1 percent, the peak deviatoric stresses increased from 343.18 kPa to 509.10 kPa (FC = 0 percent) and 586.74 kPa to 864.59 kPa (FC = 0.4 percent), respectively. In contrast, when polyurethane content ranged from 1 percent to 4 percent, the peak deviatoric stress increased from 509.10 kPa to 749.79 kPa (FC = 0 percent) and from 864.59 kPa to 1,206.2 kPa (FC = 0.4 percent), respectively. This difference indicates that the enhancement efficiency of polyurethane was obvious at low polyurethane content. Figure 3c indicates that the shear strength improved with the increase in fiber content (PC = 0 percent).

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 227–242

Sisal Fiber-Polymer–Treated Sand Mechanical Properties

Figure 3. Stress-strain curves of the samples with (a) FC = 0 percent, FL = 0 mm, ρ = 1.50 g/cm3 , and different PC; (b) FC = 0.4 percent, FL = 18 mm, ρ = 1.50 g/cm3 , and different PC; (c) PC = 0 percent, FL = 18 mm, ρ = 1.50 g/cm3 , and different FC; (d) PC = 4 percent, FL = 18 mm, ρ = 1.50 g/cm3 , and different FC; (e) PC = 0 percent, FC = 0.4 percent, ρ = 1.50 g/cm3 , and different FL; (f) PC = 4 percent, FC = 0.4 percent, ρ = 1.50 g/cm3 , and different FL; (g) PC = 0 percent, FC = 0.4 percent, FL = 18 mm, and different ρ; (h) PC = 4 percent, FC = 0.4 percent, FL = 18 mm, and different ρ.

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 227–242

231

Wu, Qian, Liu, Song, Kanungo, Bai, and Bu Table 5. Summary of triaxial tests results in present study. Deviatoric Stress (kPa)/Energy Absorption (kJ/m3 ) Sample No. T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 T13 T14 T15 T16 T17 T18 T19 T20 T21 T22 T23 T24 T25 T26 T27 T28 T29 T30 T31 T32 T33 T34 T35 T36 T37 T38 T39 T40 T41 T42 T43 T44 T45

PC (%)

FC (%)

ρ (g/cm3 )

FL (mm)

100 kPa

200 kPa

300 kPa

400 kPa

Fraction Angle (°)

Cohesion (kPa)

0 1 2 3 4 0 1 2 3 4 0 1 2 3 4 0 0 2 2 4 4 0 0 0 0 2 2 2 2 4 4 4 4 0 0 0 0 2 2 2 2 4 4 4 4

0.0 0.0 0.0 0.0 0.0 0.2 0.2 0.2 0.2 0.2 0.4 0.4 0.4 0.4 0.4 0.6 0.8 0.6 0.8 0.6 0.8 0.4 0.4 0.4 0.4 0.4 0.4 0.4 0.4 0.4 0.4 0.4 0.4 0.4 0.4 0.4 0.4 0.4 0.4 0.4 0.4 0.4 0.4 0.4 0.4

1.50 1.50 1.50 1.50 1.50 1.50 1.50 1.50 1.50 1.50 1.50 1.50 1.50 1.50 1.50 1.50 1.50 1.50 1.50 1.50 1.50 1.40 1.45 1.55 1.60 1.40 1.45 1.55 1.60 1.40 1.45 1.55 1.60 1.50 1.50 1.50 1.50 1.50 1.50 1.50 1.50 1.50 1.50 1.50 1.50

0 0 0 0 0 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 6 12 24 30 6 12 24 30 6 12 24 30

343.18/6.50 509.09/16.32 575.35/32.62 679.58/42.03 749.78/53.86 444.8/12.64 687.77/30.61 726.03/42.49 824.74/56.46 965.13/70.16 586.74/26.43 864.58/46.82 966.63/71.82 1,081.8/87.75 1,206.2/107.85 642.94/28.52 743.85/38.09 1,109.66/75.43 1,266.30/93.14 1,450.12/118.84 1,527.87/144.91 508.47/17.27 556.29/23.47 645.44/29.74 682.06/31.55 802.18/52.56 877.78/62.00 1,079.41/78.37 1,158.73/86.34 994.60/75.34 1,073.32/88.85 1,332.94/108.66 1,417.23/127.95 375.87/16.22 470.57/19.15 539.86/24.32 450.09/18.88 730.43/48.11 873.05/58.34 924.46/54.90 881.38/62.69 944.23/75.81 1,059.30/85.70 1,159.90/99.66 1,086.21/94.75

642.94/18.89 727.12/26.51 868.25/47.93 895.03/53.11 938.71/61.02 817.9/19.24 888.67/43.51 1,061.15/56.11 1,128.85/78.85 1,225.61/92.68 997/41.31 1,076.69/52.87 1,158.18/76.72 1,306.02/93.66 1,452.03/126.20 1,196.95/39.17 1,338.29/45.48 1,420.39/88.61 1,603.09/110.05 1,758.34/143.09 1,886.63/175.51 825.30/28.67 925.98/34.86 1,093.73/44.39 1,206.81/46.71 1,001.57/52.28 1,128.28/57.61 1,204.23/81.68 1,423.75/89.62 1,241.55/89.15 1,332.98/97.01 1,565.62/128.86 1,663.13/136.29 661.97/20.69 762.25/31.99 848.12/32.83 734.62/30.06 956.92/53.57 1,104.54/60.66 1,085.97/61.02 1,046.25/60.85 1,103.65/90.85 1,304.64/100.92 1,407.46/114.29 1,361.44/110.43

857.73/27.73 953.43/34.45 1,050.96/57.67 1,089.94/65.81 1,253.66/82.16 1,198.03/25.62 1,204.27/66.76 1,269.11/70.35 1,451/90.46 1,509.26/111.68 1,386.92/55.97 1,392.13/62.12 1,489.71/86.41 1,645.15/101.33 1,732.03/143.30 1,554.76/52.22 1,801.15/65.57 1,681.59/102.55 1,865.02/123.34 1,994.59/155.08 2,147.71/184.83 1,196.00/37.71 1,304.39/44.57 1,462.24/55.81 1,521.11/56.51 1,237.85/58.79 1,366.47/67.01 1,500.37/86.84 1,599.53/94.68 1,476.53/96.68 1,548.08/107.86 1,806.81/145.09 1,895.79/152.46 920.97/21.48 1,073.32/42.12 1,224.77/46.47 1,075.63/42.49 1,217.22/61.95 1,383.05/64.27 1,356.12/63.06 1,322.26/62.93 1,347.50/98.85 1,550.35/108.51 1,695.97/118.83 1,652.64/115.66

1,118.13/32.45 1,129.76/38.54 1,202.1/66.35 1,335.69/80.18 1,390.22/95.27 1,400.05/29.21 1,401.38/72.21 1,459.16/84.75 1,518.69/105.10 1,731.07/132.45 1,652.93/62.56 1,651.7/68.35 1,673.67/94.93 1,808.98/115.67 2,014.03/170.97 2,000.76/57.62 2,340.69/80.40 2,059.16/105.65 2,348.26/138.46 2,295.56/188.87 2,383.73/201.67 1,523.06/50.22 1,604.39/52.87 1,816.48/64.29 1,914.18/67.52 1,561.03/72.78 1,610.27/77.27 1,827.73/91.77 1,942.41/99.71 1,605.00/119.35 1,745.03/128.42 2,051.74/158.69 2,134.52/169.37 1,205.88/22.57 1,379.95/53.62 1,477.36/60.95 1,349.65/54.56 1,403.36/66.32 1,590.40/69.36 1,608.35/70.81 1,561.56/70.12 1,498.80/103.45 1,794.78/119.29 1,914.91/127.92 1,847.60/123.47

35.15 30.74 30.72 31.11 30.05 38.41 33.53 33.29 33.64 31.32 40.05 34.96 33.55 34.04 33.09 43.64 46.43 37.54 39.73 35.54 36.05 39.11 39.66 41.33 41.98 33.95 33.31 34.36 34.07 30.70 31.85 33.04 32.95 35.37 36.91 37.98 37.11 32.22 33.30 32.59 32.54 32.98 33.43 34.96 34.32

21.74 87.32 114.32 128.56 143.35 35.07 115.00 140.41 163.33 187.93 59.07 149.81 188.99 219.05 255.22 50.45 48.08 194.08 207.57 303.74 322.35 37.86 50.67 63.77 71.57 137.86 171.77 199.30 237.33 230.00 240.77 295.60 321.11 26.80 46.19 44.16 27.42 139.63 169.75 180.69 170.16 172.84 219.01 230.42 222.05

For a given polyurethane content and confining pressure, the shear strength improved and the axial strain at yield point increased with the increase in fiber content (Figure 3d). Under a certain CP, the sand grains and fibers in the samples are subject to greater binding force. In the process of shear deformation of the samples, sisal fibers can effectively form a spatial network structure with the increase in FC, which makes the sand particles interlock, restricts the dis-

232

placement of the sand particles, and increases the integrity, strength, and stability of the samples, and polyurethane reinforced this function. With the addition of polyurethane and sisal ﬁber, the residual strength of the samples disappeared and the axial strain of the failure point of the samples increased gradually. At the same time, the shape of the stressstrain curve gradually transited from a strain softening type to a strain hardening type.

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 227–242

Sisal Fiber-Polymer–Treated Sand Mechanical Properties

The change in fiber length has a significant effect on the deviatoric stress of samples. Figure 3e and f show that for a specific fiber content, polyurethane content, and CP, the shear strength can be improved by adding randomly distributed sisal fibers. This improvement in shear strength was maintained to a FL of up to 18 mm. The peak deviatoric stress reached the maximum of 586.74 kPa (PC = 0 percent) and 1,206.2 kPa (PC = 4 percent) when the FL was equal to 18 mm. Beyond 18 mm, the deviatoric stress decreased with further increase in fiber length. When the fiber length continued to increase, the shear strength of the sample decreased as a result of the size effect. The shear strength increased with the addition of polyurethane content under diverse fiber lengths. Figure 3g and h show that the shear strength increased uniformly with the increase in sample dry density under the same polyurethane content, fiber content, and CP. The increase in dry density decreased the porosity and improved the force between sand-sand and sand-fiber. Figure 4 plots the variation of shear stress versus axial strain of unreinforced sand (FC = 0 percent, PC = 0 percent), fiber-reinforced sand (FC = 0.4 percent, PC = 2 percent), polymer-treated sand (FC = 0 percent, PC = 2 percent), and fiber-polymer composite– treated sand (FC = 0.4 percent, PC = 2 percent) under different conditions of CP. It can be seen that the shear stress of all the samples increased as the CP increased. This is a result of interfacial friction and bite force between sand-sand and sand-fiber enhanced by the increase of CP during the shear process. However, the shape of stress-strain curves varied greatly between the samples treated using different methods. During the shearing process, the shear stress of unreinforced sand reached its peak with the axial strain of about 2 percent. Subsequently, the shear stress remained basically unchanged, and the sample has an obvious shear plane (Figure 4a). The addition of sisal fibers improved the shear strength of the specimens and increased the strain related to the failure point, and the stress-strain curve was a typical strain hardening. The shear stress of the sample increased with increase in the CP. When the axial strain is small, the influence of CP on strength is not obvious. This phenomenon indicates that in the case of small axial strain, sisal fiber has not yet fully stretched in the sample, which results in no effect of sisal fiber. With the increase in axial strain, the fiber extended further until it exceeded the initial state, thus giving full play to the frictional effect between the sand grains and sisal fibers. As a result, the influence of CP on deviatoric stress increased with the increase in axial deformation (Figure 4b). Compared with the unreinforced sand, the shear strength was improved by the addition of polymer (Figure 4c), and the failure mode was changed from shear plane to barreling. The tran-

sition of failure modes is attributed to the fact that the polymer can enhance the bond between sand particles. Among the three treatment methods mentioned above, the composite treatment method of fiber and polymer has the greatest degree of increase in deviatoric stress, and the axial strain of the elastic deformation stage increased with the addition of sisal fiber and polymer (Figure 4d). The obvious rebound phenomenon of the fiber-polymer composite samples after shear tests indicated that the fiber-polymer composite–treated sand has strong ductility. Treatment methods were evaluated by factor analysis of the relationships among those variables in Table 5. Figure 5 shows the correlation coefficient between the variables and shear strength under different conditions of CP. When the CP was 100 kPa, the correlation between shear strength and PC was the highest, followed by that between shear strength and FC. With the increase in CP, the correlation between PC and shear strength decreases gradually, while the correlations between FC, FL, and shear strength increase gradually. This indicates that when CP increases, the most significant factor affecting shear strength changes from PC to FC. Among these factors, ρ factor has the lowest significance and the least influence on shear strength. Under different CPs, PC and FC can be used as the main indices affecting shear strength owing to the fact that the eigenvalues of PC and FC factors are more than 1.0 acceptable values and the cumulative explained variance contribution rate is more than 70 percent. Energy Absorption In the process of axial compression, the sample must absorb energy continuously to overcome the friction between internal sand particles, so the deformation and failure of the sample essentially comprise the process of absorbing energy from the outside. Energy absorption can indicate the amount of energy required for a certain deformation of the sample. In the shearing process, the work done by the triaxial shear apparatus to the sample is the energy absorbed by the sample. In the present study, the axial strain at failure point was used to calculate the energy absorption. The energy absorption (E) is considered to be the area between the stress-strain curve and the strain axis of the sample, denoted as formula 6: ε (6) E = (σ1 − 2vσ3 )dε, 0

where E (kJ/m3 ) is energy absorption capacity, ε (percent) is axial strain at failure point, σ1 (kPa) is the maximum principal stress, σ3 (kPa) is the minimum principal stress, and v is Poisson’s ratio.

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 227–242

233

Wu, Qian, Liu, Song, Kanungo, Bai, and Bu

Figure 4. Stress-strain curves of the samples with diﬀerent CP: (a) FC = 0 percent, PC = 0 percent; (b) FC = 0.4 percent, PC = 0 percent; (c) FC = 0 percent, PC = 2 percent; (d) FC = 0.4 percent, PC = 2 percent.

Figure 5. The correlation between the variables and shear strength.

234

The variation curves of the energy absorption capacities of different PC, FC, ρ, and FL values under a CP of 100 kPa are shown in Figure 6. The polymer content played an important role in the energy absorption of specimens with and without fiber reinforcement. Figure 6a shows the effect of polymer contents on energy absorption capacity of different fiber content samples. As shown, energy absorption capacity increased with polymer content. Polymer enhanced the energy required for sample failure by enhancing the cohesive force between the materials inside the sample. In Figure 6a, the lowest energy absorption capacity (6.50 kJ/m3 ) is recorded for unreinforced sand, and the maximum energy absorption capacity is 107.85 kJ/m3 (FC = 0.4 percent, PC = 4 percent), which is 16.6 times more than that of unreinforced sand. The addition of fibers also played an important role in the energy absorption of the sample.

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 227–242

Sisal Fiber-Polymer–Treated Sand Mechanical Properties

Figure 6. Variations of energy absorption capacity with (a) polymer content; (b) ﬁber content; (c) dry density; (d) ﬁber length.

It is observed from Figure 6b that energy absorption increased gradually with the increase in fiber content. For the fiber-reinforced samples, the energy absorption increased by 7.84–10.22 kJ/m3 for every 0.2 percent increase in fiber content. By contrast, when the samples were treated with fiber-polymer composite, the energy absorption increased by 14.03–20.77 kJ/m3 (PC = 2 percent) and 18.05–27.19 kJ/m3 (PC = 4 percent) for each 0.2 percent increase in fiber content. The reason for this phenomenon is that the polymer effectively fills the pores in the sample, strengthens the adhesion between materials, and improves the anchoring effect of fiber in the sample. Therefore, the increase in frictional force and adhesive force to be overcome in the process of sample deformation increased the energy required for sample failure. The variation in fiber length also affects the energy absorption of samples. The curve of energy absorption versus fiber length is shown in Figure 6c. It is observed that for a given fiber content, energy absorption increased slowly with an increase in

fiber length; this trend ends when the fiber length is 18 mm. The maximum energy absorption capacities were recorded as 26.43, 71.82, and 107.85 kJ/m3 , corresponding to the polymer content of 0 percent, 2 percent, and 4 percent, respectively. Beyond 18-mm fiber length, a further increase in fiber length led to a decrease in energy absorption. The addition of a certain amount of fibers enhanced the overall stability of the sample as well as the deformation of the sample under the same axial strain needed to overcome greater friction. When the fiber length increased to the optimum length, the appropriate amount of fibers formed three-dimensional network structures to interlock sand grains, strengthened the effect of fiber reinforcement, and maximized the energy required for sample failure. Beyond the optimal length, a further increase in fiber length resulted in a decrease in the number of fibers for a given fiber content and disappearance of effective three-dimensional reticular structures. This is the reason for the decrease in energy absorption

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 227–242

235

Wu, Qian, Liu, Song, Kanungo, Bai, and Bu

Figure 7. Eﬀect of CP on (a) energy absorption capacity and (b) energy absorption ratio.

of the sample required for failure. With the increase in dry density, the sand grains in the compacted samples made the fibers bend and deform and increased the roughness of the fiber surface. In the process of deformation, the fibers need to overcome more forces to be pulled out, which leads to an increase in energy absorption. The energy absorption of PC = 0 percent, 2 percent, and 4 percent samples increased by 13.29, 28.16, and 41.61 kJ/m3 , respectively, with the increase in dry density, and this difference indicates that the effect of dry density on energy absorption of samples increased with the addition of polymer (Figure 6d). Figure 7a shows the variation of energy absorption versus CP of unreinforced sand (FC = 0 percent, PC = 0 percent), fiber-reinforced sand (FC = 0.4 percent, PC = 2 percent), polymer-treated sand (FC = 0 percent, PC = 2 percent), and fiber-polymer composite sand (FC = 0.4 percent, PC = 2 percent). These samples have a density of 1.5 g/cm3 and a fiber length of 18 mm. It is observed that the energy absorption of all samples increased with an increase in the CP, but the energy absorption capacity of unreinforced samples was much lower than that of the treated samples. For the unreinforced sand, the energy absorption increased from 6.50 kJ/m3 to 32.45 kJ/m3 with the CP ranging from 100 kPa to 400 kPa. The maximum value of energy absorption observed in the sample treated by fiber-polymer under 400 kPa confinement is 170.97 kJ/m3 , which is 138.52 kJ/m3 higher than that of the unreinforced sand. As a result of the reinforcement effect, the energy required to destroy the samples treated with different materials is clearly indicated. The present study introduced a dimensionless quantity, a parameter—energy absorption ratio (Re )— that is calculated as the ratio of the energy absorp-

236

tion of treated sand to the energy absorption of unreinforced sand and can be mathematically deﬁned as formula 7: Re = Et /Eu ,

(7)

where Et (kJ/m3 ) is the energy absorption of treated sand and Eu (kJ/m3 ) is the energy absorption of unreinforced sand. The energy absorption ratio of sand is presented in Figure 7b. The energy absorption ratio of fiberreinforced sand remained unchanged with the increase in CP, and its value is about 4. The energy absorption ratio of pure polymer-treated sand decreased from 6.13 to 3.39 with the CP ranging from 100 kPa to 400 kPa. Under high CP (300–400 kPa), the energy absorption ratio of polymer-treated sand gradually decreased to lower than that of fiber-reinforced sand. This indicates that the reinforcement effect of fibers is higher than that of polymers under high CP. Increased CP enhances the external restraint stress of a sample. Samples have shown obvious compressive hardening, which strengthens the friction between fibers and sand particles. In contrast, polymers coat and bound sand grains to form an effective curing film by reacting with water. This makes the energy absorption rate of pure polymer-treated sand lower than that of the fiber-reinforced sand under high CP. An increase in CP also results in a decrease in the energy absorption ratio of fiber-polymer composite sand from 11.04 to 5.27. This demonstrates that the reinforcement effect of the fiber-polymer treatment enhances with CP. In general, the energy absorption ratio of fiber-polymer composite sand is the highest. This indicates that the reinforcement effect of fiber-polymer composite is the best.

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 227–242

Sisal Fiber-Polymer–Treated Sand Mechanical Properties

Figure 8. Variation of cohesion with (a) polymer content; (b) ﬁber content; (c) dry density; (d) ﬁber length.

Shear Strength Parameters The failure Mohr circles of specimens under different CPs are drawn on the τ-σ plane. The drawing of the strength envelope inevitably is arbitrary. In this article, we use the least-squares method to fit the failure stress point by the “p-q method.” PC, FC, ρ, and FL all have important effects on the shear strength, such as cohesion (c) and friction angle (ϕ). In Figure 8, we show cohesion and friction angle variations against these four parameters. Figure 8a shows the cohesion of polymer-treated sand with fiber mixing contents of 0.0 percent, 0.2 percent, and 0.4 percent. The value of cohesion increased significantly with the addition of polymer. For polymer contents greater than 1%, the increase rate becomes small. The cohesions at 1 percent polymer content are 87.32,

115.01, and 149.81 kPa, respectively. This is 4.01, 3.28, and 2.53 times that of untreated sand. As a result of the polymer cementation between the sand particles, the cohesive force increases as capillary pressure forms between sand particles containing certain amounts of water. With the increase in polymer content, the binding force between particles increases, but the cohesive force remains the same. This makes the reinforcement changes slow for polymer contents of greater than 1 percent. The cohesion value of fiber-only–reinforced sand changes slightly with the increase in fiber content. The cohesion value of fiber-polymer composite– reinforced sand, however, shows an obvious increase with the increase in fiber content. Without polymer, fiber had a limited effect on the cohesive force. The cohesive force was mainly determined by the capillary pressure in the pores (Figure 8b). After the addition of

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 227–242

237

Wu, Qian, Liu, Song, Kanungo, Bai, and Bu

the polymer, the rough-faced sisal fiber was anchored in the sample, strengthening the interparticle bonding force and thus further enhancing the integrity of the sample. As observed in Figure 8c, an incremental change in sample dry density leads to an incremental change in the value of cohesion. With dry density ranging from 1.40 g/cm3 to 1.60 g/cm3 , the cohesion of samples without polymer increased from 17.26 kPa to 31.55 kPa, and the value of cohesion of the samples added with polymer increased from 52.56 kPa (PC = 2 percent) and 86.34 kPa (PC = 4 percent) to 80.72 kPa and 127.95 kPa, respectively; these values were higher than those of the samples without polymer. The addition of fibers of different lengths leads to an increase in the value of cohesion compared with that of unreinforced sand. Among them, the cohesion values for 18-mm fiber length reinforced samples are the largest, at 26.43 kPa, 71.82 kPa, and 107.85 kPa, respectively. Beyond 18-mm fiber length, a further increment in fiber length leads to a decrease in the cohesion of the samples (Figure 8d). When the fiber length continued to increase beyond its optimal length, the effectiveness of the three-dimensional structure is decreased, and the interlocking of sand-fiber is also reduced. For 0.8 percent content of 18-mm fiber and 4 percent polymer content–treated sample, the cohesion reached the maximum 157.66 kPa, which is 24.26 times more than that of the unreinforced sand. This indicates that the mechanical properties of sand treated with suitable fiber length and content and polymer content have been significantly improved. The curves of friction angles varying with different variables are shown in Figure 9. Figure 9a compares the friction angle of polymer-treated sand, mixed with a fiber content of 0.0 percent, 0.2 percent, and 0.4 percent. The addition of 1 percent of the polymer reduced the friction angle by about 5°. As the polymer content continued to increase, the friction angle tended to be stable. In the sample formed by the mixture of sand and polymer, the mesh film formed by the polymer wraps the sand particles, which increases the roundness of the particles, to a certain extent. Moreover, the polymer fills the intergranular pores and connects adjacent particles to fix the relative position of the sand particles, resulting in the limitation of mutual dislocation of the sand particles in the process of sample deformation. This is the reason for weakening the interparticle occlusal friction and thus slightly reducing the internal friction angle of the sample. With or without the addition of polymer, an increase in fiber content will lead to an increase in the internal friction angle of the sample (Figure 9b). The increase in internal friction angle caused by the increase in the number of fibers is attributed to a large number of fibers forming a three-dimensional mesh structure, so that the sand

238

particles interlock in the deformation process of the sample to render the fiber fully stretched. As a result, the occlusal friction effect is improved. The friction angle increased by 1°–2° with the increment in density from 1.4 g/cm3 to 1.6 g/cm3 (Figure 9c). With an increment in dry density, the distance between particles becomes closer, and the contact points of sand particles per unit area are increased, which improves the shear strength properties of sand. For any particular polymer content and fiber content, it is observed from Figure 9d that with increasing fiber length, the friction angle improves, but the internal friction angle starts to decrease when the fiber length exceeds 18 mm. The optimum length of fibers can form an effective network structure to interlock the sand particles and improve the mechanical properties of sand. The internal friction angle of sandy soil can be improved obviously by adding fibers, and the friction angle can be increased by about 12°. The addition of polymer slightly reduced the internal friction angle of sandy soil by about 5°. This indicates that the mechanical properties of sand treated with suitable fiber length and content and polymer content have been significantly improved. Mode of Failure In the present study, the failure modes of three methods (sisal fiber reinforcement, polymer treatment, and composite fiber-polymer treatment) of treated sand and the untreated sand were studied in detail. For the convenience of comparison, the fiber length and sample dry density were kept constant at 18 mm and 1.5 g/cm3 . Figure 10 shows the experimental photographs of unreinforced sand (FC = 0 percent, PC = 0 percent), fiber-reinforced sand (FC = 0.4 percent, PC = 2 percent), pure polymer-treated sand (FC = 0 percent, PC = 2 percent), and fiber-polymer composite– treated sand (FC = 0.4 percent, PC = 2 percent) under 100 kPa, respectively. It also includes the failure modes and the fracture surface structure diagrams of samples. From Figure 10, it can be seen clearly that the untreated sand experienced an obvious shear-faulting with an apparent peak in the stress-strain curve (Figure 10a) during the shearing process. As the fiber content increased to 0.4 percent, the failure mode of the sample transforms from shear-faulting to ductilefaulting, and the after-peak softening behavior disappears in the stress-strain curve (Figure 10b). The specimens of unreinforced sand and fiber-reinforced sand disintegrate to a certain extent after unloading and removing of rubber film. This indicates that the cohesion of these two types of specimens is mainly due to the capillary pressure between unsaturated sand particles and the CP restricting the specimens. In contrast, the failure modes of pure polymer-treated sand

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 227–242

Sisal Fiber-Polymer–Treated Sand Mechanical Properties

Figure 9. Variation of friction angle with (a) polymer content; (b) ﬁber content; (c) dry density; (d) ﬁber length.

(Figure 10c) and composite fiber-polymer–treated sand (Figure 10d) are typical ductile flow, and the failure modes of composite fiber-polymer–treated sand are more obvious. After unloading and removing the rubber film, the two types of specimens did not undergo obvious damage and rebound to a certain extent, and the composite treated specimens had a higher rebound rate. This indicates that the polymer effectively binds sand particles and enhances the ductility of the sample. It can be expected that the specimens undergo a change from shear-faulting to ductile flow with the increase of fiber and polymer content. Microstructure The microstructures of the samples after traditional triaxial testing were observed by SEM to study the interactions among fiber, polymer, and sand particles.

The microscopic images of pure polymer-treated sand and composite fiber-polymer–treated sand, respectively, are shown in Figure 11. Figure 11a clearly shows the microscopic images of 2 percent polymertreated sand, and the sand particles are covered with a large number of thin polyurethane cured membranes, which fill the voids between the sand particles and provide the bonds between them. A series of interlaced and interconnected cured membranes further enhances the cementation between the sand particles. SEM images of composite treated samples with 0.4 percent fiber content and 2 percent polymer content are clearly shown in Figure 11b. It can be clearly observed that the voids are filled with fibers and polymers, and the fibers are well anchored in the adhesive formed by polyurethane. These fibers act as bridges, and polymers act as connections that connect spaced sand grains together to form a stable structure. The

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 227–242

239

Wu, Qian, Liu, Song, Kanungo, Bai, and Bu

Figure 10. The failure modes of samples with (a) FC = 0 percent, PC = 0 percent; (b) FC = 0.4 percent, PC = 0 percent; (c) FC = 0 percent, PC = 2 percent; (d) FC = 0.4 percent, PC = 2 percent.

effective three-dimensional cross-linking network structures are largely attributed to this solid-phase bridging effect, which enhances the cementation between sand particles. The network structure formed by fibers and polymers effectively transfers stress through bridging, provides mechanical support for the sample, and improves the mechanical properties of the sample. Application of Polyurethane When sisal fiber–polyurethane are used to improve different mechanical properties under various engineering environmental conditions, the optimal content should be adjusted to meet the different engineering re-

240

quirements in the practical application. The age degradation of polyurethane is usually affected by temperature variations, water content, and strong ultraviolet rays. The durability of the impact of sisal fiber– polyurethane on the strength characteristics of sand with the conditions of temperature of −10°C to 50°C and ultraviolet rays of ࣘ250 mw/m2 is about 10 years. The optimum amount and cost of sand stabilizers are presented in Table 6 and are compared with other solutions. The sisal fiber unit price is $0.60/kg, and the cost is $0.60/m3 . According to Table 6, the cost of sisal fiber–polyurethane is lower than that of the other options. If the scale is large enough, there is still room for cost reduction. When polyurethane reacts with water

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 227–242

Sisal Fiber-Polymer–Treated Sand Mechanical Properties

Figure 11. SEM photos of samples: (a) pure polymer-treated sand; (b) ﬁber-polymer–treated sand. Table 6. Costs of soil stabilizers. Type

Polyurethane

Polyacrylamide

Xanthan Gum

Synthetic Resins

Carboxymethyl Cellulose

Unit price ($/kg) Cost ($/m3 )

0.7 30

3.0 225

2.9 87

1.1 33

to form colloids, the reaction rate is as high as 98 percent, and there will be no secondary reaction after the reaction is completed. There is no adverse eﬀect on the environment. Thus, sisal ﬁber–polyurethane should be considered as a new sand stabilizer to reinforce sand in construction.

CONCLUSIONS In present study, the following conclusions can be made about the engineering properties and mechanical behavior of polymer-fiber–reinforced sand: 1. The shear strength of the samples was improved with increased fiber content, polymer content, and dry density. An optimal fiber length of up to 18 mm achieved the best overall stability of the internal structure. The energy required for specimen failure was increased with the addition of fiber and polymer, and the specimen’s resistance to deformation was enhanced. 2. The addition of fibers has no significant effect on the cohesion of sand but can significantly increase the internal friction angle. In contrast, polymer treatment alone can significantly improve the cohesion of the sample. As expected, the mechanical properties of the samples treated with fiber-polymer composites have been significantly enhanced.

3. With the increase in polymers and fibers, the failure modes of the specimens mainly change from shearfaulting to ductile flow. With the increase in polymer content, the sample does not disintegrate and rebound after unloading, which indicates that the polymer can increase the bonding of sand particles, thus enhancing the integrity of the sample, and the mixed use of fiber and polymer further strengthens this effect. 4. SEM images of sand treated with curing agent clearly show that the surface and pores of sand particles are encapsulated and filled by polyurethane curing film, which enhances the cementation between sand particles and occupies a large amount of pore space. SEM images of fiber-polymer–treated sand indicate that fibers anchor in the polymer between sand particles, transfer the stress through a bridging effect (further strengthening the cementation between sand particles), and more effectively improve the physical properties of sand.

ACKNOWLEDGMENTS This research was ﬁnancially supported by the Water Conservancy Science and Technology Project of Jiangsu Province, China (grant 2017010) and the National Natural Science Foundation of China (grant 41472241).

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 227–242

241

Wu, Qian, Liu, Song, Kanungo, Bai, and Bu

REFERENCES Choi, S. G.; Wang, K.; and Chu, J., 2016, Properties of biocemented, fiber reinforced sand: Construction Building Materials, Vol. 120, pp. 623–629. Consoli, N. C.; Rosa, D. A.; and Cruz, R. C., 2011, Water content, porosity and cement content as parameters controlling strength of artificially cemented silty soil: Engineering Geology, Vol. 122, No. 3, pp. 328–333. Davarci, B.; Ornek, M.; and Turedi, Y., 2014, Model studies of multi-edge footings on geogrid-reinforced sand: European Journal Environmental Civil Engineering, Vol. 18, No. 2, pp. 190–205. Gumuser, C. and Senoi, A., 2016, Effect of fly ash and different lengths of polypropylene fibers content on the soft soils: International Journal Civil Engineering, Vol. 12, pp. 167–178. Hamidi, A. and Hooresfand, M., 2013, Effect of fiber reinforcement on triaxial shear behavior of cement treated sand: Geotextiles Geomembranes, Vol. 36, pp. 1–9. Hataf, N.; Ghadir, P.; and Ranjbar, N., 2018, Investigation of soil stabilization using chitosan biopolymer: Journal Cleaner Production, Vol. 170, pp. 1493–1500. Inbar, A.; Ben-Hur, M.; and Sternberg, M., 2015, Using polyacrylamide to mitigate post-fire soil erosion: Geoderma, Vol. 239, pp. 107–114. Jiang, H.; Cai, Y.; and Liu, J., 2010, Engineering properties of soils reinforced by short discrete polypropylene fiber: Journal Materials Civil Engineering, Vol. 22, No. 12, pp. 1315–1322. Keramatikerman, M. and Chegenizadeh, A., 2017, Effect of particle shape on monotonic liquefaction: Natural and crushed sand: Experimental Mechanics, Vol. 57, No. 8, pp. 1341–1348. Kutanaei, S. S. and Choobbasti, A. J., 2016, Triaxial behavior of fiber-reinforced cemented sand: Journal Adhesion Science Technology, Vol. 30, No. 6, pp. 579–593. Liu, J.; Shi, B.; Jiang, H.; Huang, H.; Wang, G.; and Kamai, T., 2011, Research on the stabilization treatment of clay slope topsoil by organic polymer soil stabilizer: Engineering Geology, Vol. 117, No. 1, pp. 114–120. Liu, J.; Shi, B.; Lu, Y.; Jiang, H.; Huang, H.; Wang, G.; and Toshitaka, K., 2012, Effectiveness of a new organic polymer

242

sand-fixing agent on sand fixation: Environmental Earth Sciences, Vol. 65, No. 3, pp. 589–595. Marri, A.; Wanatowski, D.; and Yu, H. S., 2001, Drained behaviour of cemented sand in high pressure triaxial compression tests: Geomechanics Geoengineering, Vol. 7, No. 3, pp. 159–174. Mohsin, M. A. and Attia, N. F., 2015, Inverse emulsion polymerization for the synthesis of high molecular weight polyacrylamide and its application as sand stabilizer: International Journal Polymer Science, Vol. 2015, pp. 1–10. Naeini, S. A. and Ghorbanalizadeh, M., 2010, Effect of wet and dry conditions on strength of silty sand soils stabilized with epoxy resin polymer: Journal Applied Sciences, Vol. 10, No. 22, pp. 2839–2846. Park, S. S., 2011, Unconfined compressive strength and ductility of fiber-reinforced cemented sand: Construction Building Materials, Vol. 25, No. 2, pp. 1134–1138. Pedersen, F. M. and Hansen, P. J., 2003, Effects of high pH on the growth and survival of six marine heterotrophic protists: Marine Ecology Progress, Vol. 260, No. 8, pp. 33–41. Schnaid, F.; Prietto, P. D. M.; and Consoli, N. C., 2001, Characterization of cemented sand in triaxial compression: Journal Geotechnical Geoenvironmental, Vol. 127, No. 10, pp. 404–411. Shogaki, T. and Kaneda, K., 2013, Feasible method, utilizing density changes, for estimating in situ, dynamic strength and deformation properties of sand samples: Soils Foundations, Vol. 53, No. 1, pp. 64–76. Tang, C. S.; Shi, B.; Gao, W.; Chen, F.; and Cai, Y., 2007, Strength and mechanical behavior of short polypropylene fiber reinforced and cement stabilized clayey soil: Geotextiles Geomembranes, Vol. 25, No. 3, pp. 194–202. Tingle, J. S.; Newman, J. K.; Larson, S. L.; Weiss, C. A.; and Rushing, J. F., 2007, Stabilization mechanisms of nontraditional additives: Transportation Research Record, Vol. 2, No. 1, pp. 59–67. Van, D. B. J. H.; Van Gelder, A.; and Mastbergen, D. R., 2015, The importance of breaching as a mechanism of subaqueous slope failure in fine sand: Sedimentology, Vol. 49, No. 1, pp. 81–95. Yetimoglu, T. and Salbas, O., 2003, A study on shear strength of sands reinforced with randomly distributed discrete fibers: Geotextiles Geomembranes, Vol. 21, No. 2, pp. 103–110.

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 227–242

Full-Scale Test and Numerical Simulation of Guided Flexible Protection System under a Blasting Load XIN QI HU XU* ZHIXIANG YU KEQIN SUN SHICHUN ZHAO School of Civil Engineering, Southwest Jiaotong University, Chengdu, 610031, China

Key Terms: Blasting Load, Guided Flexible Protection System, Full-Scale Test, Simulation, Rockfall Hazard ABSTRACT Both active and passive flexible protection methods are effective against rockfalls, but they can result in a secondary hazard due to cumulate rocks inside the structure. To solve this problem, guided flexible protection systems are receiving increased attention in the engineering community. In this study, a full-scale test of a guided flexible protection system was carried out, where the bottom of the mesh was anchored under a blasting load, which can be considered as an extreme loading event related to rockfall hazards. The fluid-solid coupling method was employed in a finite element model to simulate the entire process from the blast to the accumulation of rocks at the bottom of the slope. Based on the experimental and numerical results, a two-stage process was revealed, the internal force and the dissipated energy of each component were compared and analyzed, and the load-transferring path within the system was obtained. The internal forces of the support ropes reached their maximum values in the intercept stage. The posts experienced two peak values, the first of which, in the guide stage, was twice that in the intercept stage. The brake rings were the main energy-dissipating components, and the energy dissipation in the intercept stage was much greater than that in the guide stage. Furthermore, the interaction in terms of collision and friction between the rocks, the slope, and the system was not insignificant, particularly in the guide stage, which can account for more than 40 percent of the consumed energy of the rockfall. INTRODUCTION Rockfalls are a common geological hazard in mountainous areas and pose a serious threat to the oper-

*Corresponding author email: xuhu@swjtu.edu.cn

ational safety of roads and railways (Gentilini et al., 2012). In engineering practice, traditional rockfall protection can be identified as either active or passive systems (Lambert and Nicot, 2011; Bertrand et al., 2012). Active protection systems prevent the potential rocks from falling by pinning metal meshes onto the unstable cliff, whereas passive protection systems, which support dynamic loads directly, stop falling rocks due to their deformable capacity. For both types of systems, intercepted rockfalls are contained inside the structure, such that some additional internal forces and unfavorable effects can occur (Liu et al., 2020). Manual external force is needed to remove the accumulated rocks, but in reality, due to the low frequency of patrol, it is often impossible to remove the accumulated rocks in a timely fashion. When too many rocks accumulate and the weight exceeds the structural bearing capacity of the system, the rocks will penetrate the protective barrier, leading to structural failure and secondary hazards (Bertolo et al., 2009; Yu et al., 2019) (Figure 1). To solve this problem, guided flexible protection systems are receiving increased attention in the engineering community (Figure 2). These types of systems essentially represent a hybrid type mix between a barrier and a drape design, and they consist of two parts: the upper intercept part and the lower guide part. The superstructure is composed of an intercept net, wire rope cables, steel posts, and numerous energy-dissipating devices. The steel posts support the net, forming an opening for intercepting rockfalls. The substructure consists of a trailing net, support cables, and energydissipating devices. The trailing net is held a certain distance from the slope to ensure that the rocks enter the intended collection area at the bottom of the trailing net. The superstructure intercepts the rockfall, and the substructure changes the direction of the rockfall (Figure 3). Guided flexible protection systems can capture the fallen rocks in the intercept net and then control their trajectory by compressing the rebound height, leading them to a prospective accumulation position along the

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 243–256

243

Qi, Xu, Yu, Sun, and Zhao

Figure 2. A typical guided ﬂexible protection system.

Figure 1. Accumulation of rocks in (a) a passive ﬂexible barrier and (b) an active draped mesh.

slope. Additionally, a large structural elongation is not required, so this method can be applied on steep slopes next to roads and railways. The trailing net is pinned at the bottom so that all rocks can be stopped at the base of the slope to avoid the invasion of the road, which also reduces the cost of maintenance and rock removal. In general, for high and steep slopes where rockfalls occur frequently, guided flexible protection systems are economical and effective methods for controlling rockfall. Guided flexible protection systems in engineering are divided into two types: hybrid flexible protection and Long-span Pocket-type Rock-net (LPR for short). Glover et al. (2012) carried out seven impact tests with hybrid flexible protection systems. Energy and net displacement were recorded by a high-speed cam-

244

era, and the energy consumption of the system was evaluated. Arndt et al. (2009) studied the mechanical behaviors and durability of the net and support cables by performing a full-scale impact test using a hybrid flexible protection system. Giacomini et al. (2012) conducted an in situ impact test of a hybrid flexible protection system to analyze the rockfall velocity and energy distribution of the system. At present, most studies of guided flexible protection systems are based on full-scale tests, which are expensive and make it difficult to measure the force of each component; thus, scholars have studied structures via numerical techniques to solve these problems. Badger et al. (2008) combined component performance tests, rolling rock tests, and numerical simulations to study the mechanical performance of hybrid flexible protection systems. Dhakal et al. (2013) calibrated and verified a finite element model of the LPR structure and analyzed the performance of the energy-dissipating device installed in system and the effects of rock size and impact point on the displacement of the net under the same impact energy. Boetticher et al. (2011) established a finite element model of the net in a hybrid flexible protection system and explored the effects of design parameters on the energy consumption of the system. A similar rockfall attenuator system has been used in the railway industry in Canada for decades, and series of tests were conducted by Wyllie (2014), Wyllie and Shevlin (2015), Shevlin et al. (2016), and Glover et al. (2016), and the protective effect of the system was investigated by capturing the rocks’ accelerations and rotations during contact with the netting. Through numerical calculations, Dhakal et al. (2012) revealed that the energy-dissipating device plays an important role in energy dissipation of LPR systems and studied the

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 243–256

Guided Flexible Protection System during Blasting

Figure 3. Interception and guidance process.

control of the global displacement of the LPR system. Del Coz Díaz et al. (2010) studied the performance of a new type of loop energy dissipator that is applied to flexible protection by comparing numerical and experimental results. Tajima et al. (2009) investigated energydissipating devices by modeling the energy distributions and mechanical performances of LPR systems with and without such devices. Existing research and applications show the protective energy level of hybrid flexible protection systems is greater than that of LPR systems, and LPR is not suitable for field applications due to the large distances associated with mountains. Thus, this paper mainly focuses on hybrid flexible protection. The actual methods of dangerous rock mitigation include slope cutting, rock fixation, drilling, and blasting. Among them, blasting is widely used because of its strong adaptability and economic advantages (Bhandari, 1997). Rockfall from a mountain is not a single rock but a large number of scattered rocks, called a rockfall group. In most cases of existing studies, a single rock is adopted for numerical modeling of the barrier system, with little consideration of the impact of a rock group and even no consideration of the effects of a blasting load. Thus, a new simulation approach that includes both rock groups produced by the blast and the flexible barrier system should be investigated. In addition, full-scale impact tests on guided flexible protection systems are very costly, and no tests have been performed under blasting loads. Thus, it is necessary to carry out a full-scale impact test under blasting conditions. The force characteristics of the guided protective net under a blasting load can be obtained intuitively, and they can provide a basis for the numerical simulation. This paper first describes an in situ full-scale test carried out using a blasting load and then presents the experimental results. Then, the numerical strategies are

presented based on the finite element model, including the state equation, fluid-solid coupling, many-body collision, and the establishment of special elements. The subsequent section describes the elements and material properties and then introduces the entire procedure to establish the model. Next, there is a discussion on the interaction between the rocks and the guided flexible protection system, and the mechanical characteristics of the net, support cable, and steel post are analyzed and described. Then, the energy distributions of the components and the two stages are demonstrated. Finally, the load-transferring path is illustrated, which can be used to provide guidance for engineering designs. EXPERIMENTAL STUDY The test site was located in Fuling District, Chongqing, China, and the vertical height of hillside covered in broken rocks was 45 m. Explosives were set at a vertical height of 14 m, and there were a total of 8 holes with depths of 4 m. Each hole was filled with 1.5 kg of trinitrotoluene (TNT) explosives. Only one blast was conducted in this test, and explosives in all holes were ignited at the same time. The guided flexible protection system was 10 m in width and 17 m in height, and the middle of the structure was 1.8 m from the slope, as shown in Figure 4. The detailed parameters of the components are shown in Table 1. The steel wire-ring net consisted of connected wire rings, each of which was made of single ultra-high-strength steel wire with a cross-sectional diameter of 3 mm and was manufactured with 7 windings as well as 300 mm internal diameter (R7/3/300). All the brake rings used in the test were GS-8002 type, which were performed according to the Standard for Flexible Safety Net for Protection of Slope Along the Line (China Railway Publish House, 2004).

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 243–256

245

Qi, Xu, Yu, Sun, and Zhao

Figure 4. Composition of the experimental model.

Experimental Process and Results Analysis After ignition, many rocks were separated from the mountain, and a rockfall group formed (Figure 5a). When the rockfall group impacted the system, the deformation of the contacted net increased (Figure 5b), the steel posts were deﬂected, and the brake rings in the superstructure became involved. The velocity of the rocks sharply decreased. After 0.5 seconds, rocks rolled down along the trailing net under gravity, and

Table 1. System components. Component Steel wire-ring net Upslope anchor cable Lateral anchor cable Upper support cable Longitudinal cable Number of brake rings Steel post Lower support cable

246

Size (mm) R7/3/300 Ø18 Ø18 Ø22 Ø20 12 H150×150×3 Ø20

the deformation of the trailing net increased (Figure 5b and c). Finally, the rockfall group formed a pile at the bottom of the slope (Figure 5d). In this experiment, the guided flexible protection system first suppressed the horizontal movements of rocks produced by blasting and then controlled the trajectories of rocks so that the rocks entered the collection area according to a predetermined trajectory. Five analysis points on the wire-ring net were selected for observing the deformation as well as the internal force distribution of the system, which are referred to as points 1–5 in the following section. Points 1 and 2 were located near the end of the posts at the upper corner of the net, and points 3, 4, and 5 were located along the middle column of the net, at the top, middle, and bottom, respectively, as shown in Figure 4. The deformation at points 1 and 2 was significantly greater than that of other locations after the experiment. The post feet were inwardly bent due to the medial tensions from the upper support cables and the longitudinal cables (Figure 6a). After the experiment, the elongation of brake ring 8 reached approximately

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 243–256

Guided Flexible Protection System during Blasting

Figure 5. Experimental process: (a) detonating; (b) impacting; (c) rolling down; and (d) accumulating at the bottom.

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 243â&#x20AC;&#x201C;256

247

Qi, Xu, Yu, Sun, and Zhao

Figure 6. Experimental results: (a) bending deformation of the post foot; (b) deformation of the brake rings; and (c) accumulation of rocks constrained in the mesh.

0.60 m, the elongation of brake ring 10 was approximately 0.30 m, and brake ring 12 did not experience elongation (Figure 6b). Finally, all blasted rocks were contained and accumulated at the bottom of the trailing net, a great pile of debris pulled the net downwards, and the wire rings in the net indicated tensile behavior. The height of the accumulation area was 1.5 m, and the depth perpendicular to the slope was 3.3 m (Figure 6c). Moreover, the volume of the rocks was approximately 16 m3 , and the weight was about 40 tons. NUMERICAL CALCULATION METHOD To establish an accurate numerical method, various key features should be considered in addition to the impact contact relationship and the elastoplastic material properties. These parameters include the formation of the blasting load, the large deformation of the net, the sliding relationships among the components, and the failure criteria of the components. Then, a three-dimensional finite element model that can predict the impact results may be developed (Yu et al., 2018). The blasting simulation was realized based on a fluid-solid coupling algorithm. Explosives and air are considered as fluids; the blasting load is transmitted among these fluids and rocks in terms of a group of equilibrium equations of mechanics through coincident nodes. Regarding the fluid material, the relationship between volume and internal energy under pressure can be simulated by the state equations (Jiang and

248

Zhou, 2012). The TNT explosives were described by a constitutive model in the form of the Jones-WilkensLee (JWL) equation of state: ω ω −R1V P =A 1− e e−R2V +B 1− VR1 VR2 ωE , + (1) V where P is the detonation pressure, V is the relative volume, and E is the internal energy of the unit volume. The physical meanings of the other parameters and values in this paper are shown in Table 2. The null material model and the linear polynomial state equation were used to model the air: P = C0 + C1 μ + C2 μ2 + C3 μ3 + (C4 + C5 μ + C6 μ2 )E, (2)

Table 2. Explosive material parameters. Symbol P E0 V0 A B R1 R2 ω

Deﬁnition

Value

Detonation pressure, Pa Initial internal energy, J/m3 Initial relative volume Material constant, Pa Material constant, Pa Material constant Material constant Material constant

1.85 × 1010 7 × 109 1 3.712 × 1011 3.230 × 109 4.15 0.95 0.3

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 243–256

Guided Flexible Protection System during Blasting Table 3. Element and material properties. Model Bedrock Ground Explosive Air Ring Brake ring Cable Post

Element Solid Shell ALE multiple material solid ALE multiple material solid Beam Beam Beam Beam

Material Ideal elastic plasticity Elasticity High explosive Null Multistage linear plasticity Multistage linear plasticity Discrete beam Ideal elastic plasticity

Density (kg/m3 )

Elastic Modulus

Yield Stress (Pa)

2,500 2,500 1,600 1.225 7,900 7,900 7,900 7,900

5.5 × 10 2.0 × 1010 — — 1.2 × 1011 1.5 × 1011 1.2 × 1011 2.6 × 1011

1.2E8 — — — 1.15 × 109 1.0 — 3.45 × 108

following equations.

and 1 μ= − 1, V

(3)

where the meanings of P, V, and E are the same as those above. Furthermore, C0 = C1 = C2 = C3 = C6 = 0, C4 = C5 = 0.4, and the initial relative volume of the model is taken as 1. In the fluid-solid interaction process, air and explosives are described by a compressible Newtonian that should follow the laws of mass, momentum, and energy conservation (Kim and Guyer, 2014). In addition, the equilibrium between the air and the rock is controlled by Newton’s second law. The penalty function was used to address collisions between different parts that involved the rocks, the structure, the hillside, and the ground in the computing model (Wyllie, 2014). This method realizes the global automatic contact among all different parts, even among the separated elements included in the same part. Normal impact force and tangential frictional force can be obtained by defining proper penalty stiffness and friction coefficients, so that the energy dissipated by collision can also be derived. Combined with the above numerical calculation methods, the finite element model, which consists of different element types, material models, elasticity values, ideal elastic plasticity values, and other factors (Table 3), was established based on LS-DYNA (Qi et al., 2018).

F =

mi ai .

(4)

i=1

Ek =

n 1 i=1

mi v i 2 .

(5)

mi v i

i=1

. (6) m In fact, blasting produces rocks with varying masses and velocities that are related to their mass and distance from the explosive point. However, the total mass of the rock group m = ni mi = 9,815 kg. When the velocity of the center of mass (hereafter referred to as velocity) and the total kinetic energy of rocks increase to their maximum values (0.07 seconds), the impact energy is approximately 2,600 kJ. There is no horizontal impact force acting on the net because the rocks have not yet contacted the net during the blasting process. The contact between the fastest rock and the net at 0.09 seconds marks the beginning of the interactive process in the system (Figure 7a). For a simpler study of the mechanical properties of the components and transfer mechanism of the system, the process (Figure 7b and c) can be divided into the intercept stage and the guide stage, during which the development of deformation, internal force, and dissipating energy of components are depicted in Figure 8 and Figure 9. vc =

SIMULATION ANALYSIS

Intercept Stage (0.09–0.68 Seconds)

From 0 seconds to 0.09 seconds, the terms n, m, mi, vi , and αi are the number of rocks produced by blasting, the total mass, and the mass, velocity, and acceleration of a single rock, respectively. Therefore, the sums of the horizontal impact force, the total kinetic energy, and the horizontal impact velocity of the center of mass of the rockfall group are represented by the

From 0.09 seconds to 0.14 seconds, the lightweight rocks close to the explosive points reach the net more quickly than others, and the initial contact occurs at 0.09 seconds (Figure 8a). Then, all the rocks arrive at the net successively until time reaches 0.14 seconds, leading to evident increasing contact area and out-of-plan deformation of the net (Figure 8b). From

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 243–256

249

Qi, Xu, Yu, Sun, and Zhao

Figure 7. Rockfall distribution: (a) beginning of the interactive process; (b) beginning of the guide stage; and (c) side view of the accumulation of rocks.

0.14 seconds to 0.48 seconds, the strains of rings in the intercept net reach their maximums, with almost no strain present in the trailing net, and the strain peaks at points 1 and 2 are three times that at point 3 (Figure 9a). The system reaches the maximum limit state of horizontal impact, with a peak horizontal impact load up to 580 kN (Figure 9b), when the horizontal velocity and the total kinetic energy of the rocks decrease to zero and 617 kJ, respectively (Figure 9c and d). All brake rings begin to elongate. Furthermore, the elongations of brake ring 1 and brake ring 4 on the lateral upslope anchor cables are greater than those of brake ring 2 and brake ring 3 on the internal upslope anchor cables (Figure 9e). From 0.48 seconds to 0.68 seconds, the motion of the rocks reverses, and the rocks separate from the intercept net (Figure 8c), resulting in a rapidly decreasing impact force. The intercept stage, which is from the moment of initial contact between the rocks and the net to the moment when a small and stable impact force forms, takes 0.59 seconds. Guide Stage (0.68–3.50 Seconds) From 0.68 seconds to 1.76 seconds, the intercepted rocks roll down the trailing net (Figure 8d and e). During this phase, the strain at point 4 increases rapidly, and the deformation of the intercept net is driven by the trailing net. After 1.76 seconds, when the rocks begin to enter the intended collection area, all the brake rings on the support cables elongate again (Figure 9f). Finally, the accumulation area of rocks forms at the

250

end of the net with a width of 5 m (Figure 8f). This type of accumulation is easy to maintain and remove. In addition, the post for which the initial and ﬁnal vertical angles with the slope are 88° and 81° is bent inward by 7°. For the entire impact process, the intercept stage takes 0.59 seconds, while the guide stage takes 2.82 seconds. By comparative analysis, experimental phenomena such as the deformation of the net, the elongation of the brake rings (Table 4), the bending of the steel posts, and the accumulation of the rocks are in good agreement with the numerical results, indicating the validity of the adopted numerical approach. In the following sections, based on this ﬁnite element model, the mechanical characteristics as well as the energy consumption are analyzed. Table 4. Comparison of elongation of the brake rings between tests and simulations (unit: m). Brake ring

Test

Simulation

1 2 3 4 5 6 7 8 9 10 11 12

0.70 0.40 0.40 0.70 0.70 0.70 0.60 0.60 0.30 0.30 0 0

0.84 0.57 0.51 0.82 0.81 0.87 0.73 0.71 0.50 0.51 0.15 0.15

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 243–256

Guided Flexible Protection System during Blasting

Figure 8. Deformation of the net at key moment (unit: m): (a) initial contact; (b) full contact; (c) limit state of horizontal impact; (d) rocks rolling down along the net; (e) rocks beginning to enter the collection area; (f) rocks have accumulated in the collection area completely and are contained.

MECHANICAL ANALYSIS OF THE COMPONENTS Tension Force of the Support Cable For the upper support cables, before 0.09 seconds, there is no tensional force. From 0.09 seconds to 0.68 seconds, the tensional force increases to peak 1 (250 kN) and then decreases to a lower value. From 0.68 seconds to 3.50 seconds, the axial force increases again and reaches peak 2 (180 kN) before ﬁnally stabilizing at approximately 15 kN (Figure 10a). From 0.09 seconds to 0.68 seconds, the lower support cables are stressed because of the deformation in the intercept net. Moreover, all axial forces reach peak 1 at 0.48 seconds, and the maximum value is 125 kN. Af-

ter 0.68 seconds, the axial forces are near zero until 1.70 seconds, and then the axial force of the lower support cable 1 increases to peak 2 (100 kN). However, the peaks for lower support cables 2 and 3 are not obvious (Figure 10b). Overall, the axial forces of the upper support cables are two times those of the lower support cables. In the intercept stage, rocks act directly on the superstructure, and the forces of all support cables reach peak 1. In the guide stage, rocks act directly on the substructure, and the forces of upper support cables and lower support cable 1 reach peak 2. For all support cables, the peaks in the intercept stage are greater than those in the guide stage, and the internal forces of the support cables also reﬂect the two-stage behavior.

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 243–256

251

Qi, Xu, Yu, Sun, and Zhao

Figure 9. Time-history curves: (a) strain of wire-ring net; (b) horizontal impact force; (c) horizontal velocity of the center of the rock mass; (d) total kinetic energy of the rocks; (e) elongation of the brake rings on the upslope anchor cables; and (f) elongation of the brake rings on the support cables.

Axial Force of the Post The post is subjected to actions Ts1 and Ts2 from the upper support cables and Tu1 and Tu2 from the upslope anchor cables (Figure 11a). The components of Ts1 in the −x and y directions impose pressure on the 252

post, and the component in the −z direction imposes a downward force. The right support cable connected to the hillside is close to the level of the post, and the pressure associated with its force Ts2 acts in the x and y directions. The upper anchor cables apply an upwardinclined force to the post, and the components of Tu1

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 243–256

Guided Flexible Protection System during Blasting

Figure 10. Internal force acting on the support cables: (a) upper support cable; (b) lower support cables.

in the −x and y directions and those of Tu2 in the x and y directions all apply pressure to the post. Fx , Fy , and Fz , represent the horizontal force, the axial pressure, and the vertical force experienced by the post, respectively. Thus, the forces of the post in the x, y, and z directions can be summarized as follows: ⎧ ⎪ ⎨Fx = Tu1x + Ts1x − Ts2x − Tu2x Fy = Tu1y + Ts1y − Ts2y − Tu2y . ⎪ ⎩F = T +T −T −T z u1z s2z s1z u2z

(7)

The axial forces of the two posts reach peak 1 (−100 kN) in the intercept stage and peak 2 (−200 kN) in the guide stage, and peak 2 is 2 times larger than peak 1 (Figure 11b). As the only compressed member in the system, the post is involved in the whole process, and its force evolution also follows the two-stage pattern.

Energy Distribution An initial kinetic energy equal to 2,600 kJ is produced by the blast. Although the effect of gravity on rocks is considered until the final residual energy is zero, the total energy dissipated during the entire process is 3,777 kJ, which includes the collision and friction among the rocks, the slope, the ground, and the structure. As illustrated in Figure 12, the structure consumes energy equal to 2,292 kJ by means of different approaches, including plastic deformation, movements, and sliding of components (Xu et al., 2018). In particular, net and brake rings dissipate 910 kJ and 957 kJ, respectively, taking up 39.7 percent and 41.8 percent. Additionally, the energy dissipated by the brake rings on the support cables is much greater than that dissipated by the brake rings on the anchor cables. The energy dissipated by the other components, such as posts, cables, and shackles, accounts for 18.5 percent. During the impacting contact, the loss of energy exchange between contact pairs can be treated as the energy dissipated by the collision due to penetration

Figure 11. Internal forces of the posts: (a) internal force analysis of post; (b) axial force-time curves for the posts.

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 243–256

253

Qi, Xu, Yu, Sun, and Zhao

Figure 12. Statistics of energy consumption (unit: kJ).

and friction. Figure 12 also depicts the energy consumption between each contact pair, which reaches a value of 1,506 kJ, accounting for 39.8 percent of the total energy. Based on the two-stage pattern, as the intercept stage transfers into the guidance stage, energy dissipated by the structure decreases while the energy dissipated by collision and friction eﬀects grows. Figure 12 shows that the structure consumes 1,651 kJ and 641 kJ successively in the two stages, and the energy dissipated by collision and friction increases from 536 kJ to 970 kJ. In general, the energy dissipated in the intercept stage and guidance stage account for 57.9 percent and 42.1 percent, respectively. This shows that collision and friction during the impact could take up a large portion of the dissipated total energy, and the guidance stage also plays an important role in consuming energy. Path of Force Transfer The path of force transfer was obtained from the operational order of each component (Figure 13). From 0.09 seconds to 0.68 seconds, the blasted rocks directly contacted the intercept net, leading to the instantaneous deformation of the intercept net. This deformation applied force to the trailing net, and the force was passed to the upper and lower support cables. Next, the brake rings on the support cables began to elongate. The deformation of the support cables transferred the force to the steel posts, and then the steel posts transferred the force to the anchor cables, resulting in the

254

initial elongation of the brake rings on the anchor cables. The force was then transferred from the anchor cables to the foundation. From 0.68 seconds to 3.50 seconds, as the rocks rolled along the trailing net, the deformation of the intercept net decreased, whereas that of the trailing net and the internal force of the lower support cable increased. At the same time, the deformation of the trailing net transferred the force to the upper support cables. Finally, the force was passed to the foundation through the posts and the anchor cables. CONCLUSIONS This paper presents the working mechanism for a guided flexible protection system through a full-scale impact test under a blasting load. Moreover, a finite element model of mountain blasting was established based on fluid-solid coupling. Through analysis of the deformation and displacement of the structure, the interaction between the rocks and the structure was simulated. Then, by combining the full-scale test and finite element simulation, a mechanical analysis was performed, and the following conclusions were reached: 1. Based on the actual rockfall hazard in a natural environment, the action of multiple rocks on the system should be taken into account for structural design, to consider the effects of spatial distribution of the load, different impact angles, and interaction among rocks. This loading mode can be realized by

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 243–256

Guided Flexible Protection System during Blasting

Figure 13. Path of force transfer.

means of a blast in a field test. In addition, a guided flexible protection system designed under a blasting load has a higher capacity to absorb a natural rockfall, since bursting rocks contain additional kinetic energy from the blast. 2. The interaction in terms of collision and friction between the rocks, the slope, and the system is significant, particularly in the guide stage. It can account for more than 40 percent of the energy dissipation of the total energy of the rockfall, the rest of which is consumed by the guided flexible protection system. 3. The internal forces of the support cables and steel posts all present a two-stage pattern for the interaction between the rocks and the structure. The upper and lower support cables both reach two peaks in the intercept and guide stages, and the internal force peak in the intercept stage is greater than that in the guide stage. Among the support cable components, the internal force of the upper support cable has a peak is twice that of the lower support cable. The axial force of the steel post reaches peak values in the intercept and guide stages, and the peak in the guide stage is larger. 4. The main dissipative components in the system are the brake rings, which account for 51 percent of

the total energy consumption. In contrast, the steel posts dissipate almost no energy. The energy consumption of the system in the intercept stage is much greater than that in the guide stage. ACKNOWLEDGMENTS The ﬁnancial support provided by the National Key R&D Program of China (grant no. 2018YFC1505405), the National Natural Science Foundation of China (grant no. 51678504), the Fundamental Research Funds for the Central Universities (grant no. 2682017CX006), and the fund of the Department of Science and Technology of Sichuan Province (grant no. 2018JY0029 and 2019YJ0221). OST Slope Protection Engineering, Co. (Sichuan, China), is also acknowledged for allowing use of the data presented in the paper. REFERENCES Arndt, B.; Ortiz, T.; and Turner, A. K., 2009, Colorado’s FullScale Field Testing of Rockfall Attenuator Systems: Transportation Research Circular E-C141, Washington, DC, 108 p. Badger, T. C.; Duffy, J. D.; and Sassudelli, F., 2008, Hybrid barrier systems for rockfall protection. In Volkwein, A. (Editor),

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 243–256

255

Qi, Xu, Yu, Sun, and Zhao Interdisciplinary Workshop on Rockfall Protection: IABSE, Morschach, Switzerland, pp. 10–12. Bertolo, P.; Oggeri, C.; and Peila, D., 2009, Full-scale testing of draped nets for rock fall protection: Canadian Geotechnical Journal, Vol. 46, No. 3, pp. 306–317. Bertrand, D.; Trad, A.; Limam, A.; and Silvani, C., 2012, Fullscale dynamic analysis of an innovative rockfall fence under impact using the discrete element method: From the local scale to the structure scale: Rock Mechanics and Rock Engineering, Vol. 45, No. 5, pp. 885–900. Bhandari, S., 1997, Engineering Rock Blasting Operations: CRC Press Taylor and Francis Group, Rotterdam, Netherlands, 872 p. Boetticher, A. V.; Glover, J.; Volkwein, A.; and Denk, M., 2011, Modelling flexible wire netting applied to rock fall attenuating systems. In Proceedings of Interdisciplinary Workshop on Rockfall Protection, Geological Service of the Austrian Torrent and Avalanche Control, Innsbruck-Igls, Austria, pp. 71– 72. China Railway Publish House, 2004, The flexible safety net for protection of slope along the line (TB/T 3089-2004). Professional Standard of the Peoples Republic of China, Beijing (Editor). (in Chinese) Del Coz Díaz, J. J.; García Nieto, P. J.; Castro-Fresno, D.; and Rodríguez-Hernández, J., 2010, Nonlinear explicit analysis and study of the behaviour of a new ring-type brake energy dissipator by FEM and experimental comparison: Applied Mathematics and Computation, Vol. 216, No. 5, pp. 1571–1582. Dhakal, S.; Bhandary, N. P.; Yatabe, R.; and Kinoshita, N., 2012, Numerical and analytical investigation towards performance enhancement of a newly developed rockfall protective cable-net structure: Natural Hazards and Earth System Science, Vol. 12, No. 4, pp. 1135–1149. Dhakal, S.; Bhandary, N. P.; Yatabe, R.; and Kinoshita, N., 2013, Finite element modelling and parametric analyses of a long-span pocket-type rockfall interceptive cable-net structure. In Margottini, C.; Canuti, P.; and Sassa, K. (Editors), Landslide Science and Practice: Risk Assessment, Management and Mitigation, Vol. 6: Springer, Berlin, pp. 597–606. Gentilini, C.; Govoni, L.; de Miranda, S.; Gottardi, G.; and Ubertini, F., 2012, Three-dimensional numerical modelling of falling rock protection barriers: Computers and Geotechnics, Vol. 44, pp. 58–72. Giacomini, A.; Thoeni, K.; Lambert, C.; Booth, S.; and Sloan, S. W., 2012, Experimental study on rockfall drapery systems for open pit highwalls: International Journal of Rock Mechanics and Mining Sciences, Vol. 56, pp. 171–181. Glover, J.; Denk, M.; Bourrier, F.; Gerber, W.; and Volkwein, A., 2012, Kinetic energy dissipation effects of rock fall attenuating systems. In Koboltschnig, G.; Hubl, J.; Braun, J.

256

(Editors), Proceedings of the 12th Congress Interpraevent, Grenoble, France, pp. 138–139. Glover, J.; Wyllie, D. C.; and Bucher, R., 2016, Attenuator systems—An old method to deviate rocks but a new testing method for developing a design concept. In Dight P.M. (Editor), Proceedings of the First Asia Pacific Slope Stability in Mining Conference, Australian Centre for Geomechanics, Perth, pp. 435–442. Jiang, N. and Zhou, C., 2012, Blasting vibration safety criterion for a tunnel liner structure: Tunnelling and Underground Space Technology, Vol. 32, pp. 52–57. Kim, A. and Guyer, R. A., 2014, Nonlinear Elasticity and Hysteresis: Fluid-Solid Coupling in Porous Media: Markono Print Media Pte, Ltd., Singapore, pp. 227. Lambert, S. and Nicot, F., 2011, Rockfall Engineering: ISTE, Ltd., and John Wiley & Sons, Inc., Boca Raton, FL, 270 p. Liu, C.; Yu, Z.; and Zhao, S. C., 2020, Quantifying the impact of a debris avalanche against a flexible barrier by coupled DEMFEM analyses: Landslides, Vol. 17, No. 1, pp. 33–47. Qi, X.; Yu, Z. X.; Zhao, L.; Xu, H.; and Zhao, S. C., 2018, A new numerical modelling approach for flexible rockfall protection barriers based on failure modes: Advanced Steel Construction, Vol. 14, No. 3, pp. 479–495. Shevlin, T.; Wyllie, D. C.; and Glover, J., 2016, Attenuators for controlling rockfall: First results of a state-of-the-art full-scale testing program. In Proceedings of the 67th Highway Geology Symposium, Colorado, United States, pp. 567–584. Tajima, T.; Maegawa, K.; Iwasaki, M.; Shinohara, K.; and Kawakami, K., 2009, Evaluation of pocket-type rock net by full scale tests: IABSE Symposium Report, Vol. 96, No. 4, pp. 10–19. Wyllie, D. C., 2014, Rock Fall Engineering: CRC Press, Boca Raton, FL, 235 p. Wyllie, D. C. and Shevlin, T., 2015, Attenuators for controlling rock fall: Do we know how they work? Can we specify what they should do? In Proceedings of the 66th Highway Geology Symposium, Stubridge Massachusetts, United States, pp. 414– 430. Xu, H.; Gentilini, C.; Yu, Z., Qi, X.; and Zhao, S., 2018, An energy allocation based design approach for flexible rockfall protection barriers: Engineering Structures, Vol. 173, pp. 831–852. Yu, Z. X.; Qiao Y. K.; and Zhao, L., 2018, A simple analytical method for evaluation of flexible rockfall barrier part 1: Working mechanism and analytical solution: Advanced Steel Construction, Vol. 14, No. 2, pp. 115–141. Yu, Z. X.; Zhao, L.; Liu, Y. P.; Zhao, S. C.; Xu, H.; and Chan, S. L., 2019, Studies on flexible rockfall barriers for failure modes, mechanisms and design strategies: A case study of western China: Landslides, Vol. 16, No. 2, pp. 347–362.

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 243–256

Characteristics, Controlling Factors, and Formation of Shallow Buried Karst in Eastern China: A Case Study in the Wuxi Metro Areas, Jiangsu Province SHULAN GUO CHANGHONG YAN* LIANGCHEN YU YANG LIU School of Earth Science and Engineering, Nanjing University, Nanjing 210023, China

YINKANG ZHOU School of Civil Engineering, Anhui University of Technology, Maanshan 243032, China

XIAOZHONG SHI Wuxi Municipal Design Institute Co., Ltd, Wuxi 214000, China

Key Terms: Eastern China, Shallow Buried Karst, Controlling Factors of Karst Development, Formation Process, Treatment ABSTRACT Karst-related geo-disasters often occur in karst regions during underground construction. In recent years, a number of shallow buried karst features have been discovered in eastern China. Survey boreholes show that karst caves exist at depths of 26–30 m and 33–36 m, which are typical of shallow buried karst. Using crosshole seismic computed tomographic techniques, 36 geological anomalies consisting of 26 mud-filled caves, three empty caves, and seven fissure zones/loose lens bodies were detected. Most of the karst caves were oblate or oval. After determining the scale of and the connection between the caves, the study area was divided into two areas to provide advice on the treatment of caves during construction. An analysis of the development characteristics of the karst suggests that karst is controlled by several key factors, including topography, structural features, groundwater conditions, and human activities. Following an analysis of these factors and field studies of the process of formation of the shallow buried caves, solutions are proposed to prevent karst-related geo-hazards. The research results can be used in the design of foundations for buildings that are located in regions that include shallow buried karst in eastern China.

*Corresponding author email: yanchh@nju.edu.cn

INTRODUCTION Buried karst is often overlooked because it is not obvious on the surface. In recent years, rapid urbanization in eastern China has led to the extensive development and utilization of underground spaces (You et al., 2018). Karst-related geo-disasters have frequently occurred during underground engineering in limestone regions, such as the karst collapse that occurred during underground construction in Xuzhou, Jiangsu Province, and the water inrush caused by karst during the construction of the Shangyuanmen Station of the Nanjing Metro (Xing et al., 2014; Cui et al., 2015); these events pose serious threats to people’s lives and property. Safety of construction in buried karst areas is, therefore, an urgent problem that must be addressed (Wang et al., 2011) and has attracted the attention of relevant departments and researchers. Some studies have been published on the characteristics and controlling factors of karst (Wei and Shi, 2017; Zhang et al., 2018). Other research has targeted more speciﬁc topics; for example, the geological structure and lithological features of carbonate rocks have been analyzed to clarify karst development (Pan, 1985; Zhu and Dong, 2006; and Li et al., 2015), geological models for karst collapse have been generated (Abdeltawab, 2013; Gong et al., 2018; and Liu et al., 2018), and proposals have been made for improving the detection of karst (Zhou and Beck, 2005; Cai et al., 2011; Tang et al., 2011; Duan et al., 2013; and Zajc et al., 2014). Studies have also been carried out on risk analysis (Zhu and Dong, 2006; Shao et al., 2015; and Huang et al., 2017) and on the mitigation of georelated hazards (Gutiérrez et al., 2014; Duan et al., 2017).

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 257–269

257

Guo, Yan, Yu, Liu, Zhou, and Shi

It is generally accepted that a full understanding of the characteristics of karst and karst detection and formation are vital to avoid the risk of the sudden collapse of tunnels (Alija et al., 2013; Taheri et al., 2015; and Santo et al., 2017). Recently, during urban construction in eastern China, many karst caves have been found between 20 and 40 m below the surface, which are typical locations of shallow buried karst; however, there are few studies on the development characteristics and development mechanisms of shallow buried karst in eastern China. Preliminary investigations indicate the development of shallow buried karst in one particular underground section of the proposed Wuxi Metro Line 4. Combined with the karst in this area, the development characteristics, controlling factors, and formation process of shallow buried karst are discussed and proposals are put forward to prevent karstrelated geo-hazards. The work presented in this article provides a new contribution to similar engineering construction projects in the area with shallow buried karst of eastern China. BACKGROUND Study Area The study area of the proposed metro line is located in the Binhu District of Wuxi city, Jiangsu Province, eastern China (Figure 1); the area is located 500 m southeast of Chanshan Mountain, 1.1 km north of the Liangxi River, and 1.8 km west of the Beiijing– Hangzhou Canal. The topography of the study area is high in the north and low in the south with a relief of 5.0–9.8 m. Buildings and underground pipelines are dense in the study area. Geological Setting Wuxi city is located in the transition area between the second uplift zone of the Neocathaysian tectonic system and the complex tectonic zones of the Qinling Mountains. The bedrock of the study site consists of Carboniferous limestone, Devonian sandstone, and Quaternary loose sediments. The limestone in the study area lies in the Middle Carboniferous (C2 h)– Upper Carboniferous (C3 c) strata. The location of the strata is shown in Figure 2a, and the special terrain is indicated by “V” in Figure 2b (two basins sandwiched between three mountains). Field investigations on the northwest side of Chanshan Mountain have shown that there are two faults located at 31º34 12 N and 120º15 8 E; one is a large fault with a dip angle of 75º toward the southwest, which is a reverse fault associated with intensive multiple parallel cleavages (Figure 3), whereas the other is a small fault with a dip angle of 71º toward the northwest.

258

The study area lies in an alluvial plain of the Yangtze River delta and along the shore of Taihu Lake, China; its stratigraphy and lithology are shown in Figure 4 (borehole Z103 was drilled to a depth of 52.34 m). The upper strata are Q4 and Q3 sediments comprising gray, gray-yellow clay, silty clay, and stony soil; the underlying bedrock is the gray-pink moderately weathered limestone of the Carboniferous. The undulating limestone surface indicates that a steep terrain existed before it was covered by Quaternary sediments (Figure 5); it has a depth of about 27–52 m with a dip angle of 10º–20º toward the southeast. Hydrogeological Setting Field investigations indicate that there are no existing rivers and ponds in the study area, but the surface water system is very close to the Beijing–Hangzhou Grand Canal and the Liangxi River. Simultaneously, field investigation indicated that the groundwater flows along strata in a southeast direction. During the survey period, the groundwater level of the study area ranged from a depth of 14.41 to 19.38 m and was affected mainly by rainfall and the water level of the nearby river and, according to water quality analyses (Jiang et al., 2013), often contains CO2 at corrosive levels (10.34 mg/L). According to the groundwater conditions, the aquifers in the study area consist of the unconfined water of Quaternary strata and karst water and fissure water within the Carboniferous strata (Table 1). Quaternary The Quaternary strata consist of alluvial and flood deposits. Texturally, they are composed of filling, sandy clay, and silts; their thickness ranges from 0 to 10.2 m with an average of 6.34 m. The groundwater level of this aquifer is controlled mainly by topography and short-term meteorological conditions and is characterized by considerable changes. The groundwater level is highest from June to September, which is the season with the greatest rainfall, and lowest during the dry season from December to March. In extreme cases, the groundwater level can reach the surface, or there may be no groundwater. Karst Water The karst water is enriching in the limestone of C2 h and is a confined aquifer, which is concentrated mainly in the northwest tensional fault fracture zone and in the syncline tectonic basin that formed by vertical displacement during tectonic movement. It receives mainly the recharge of the adjacent karst water, a small

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 257–269

Characteristics, Controlling Factors, and Formation of Shallow Buried Karst in Eastern China

Figure 1. Aerial view of the study area (satellite map from http://www.map.google.com).

amount of Quaternary aquifer infiltration, and fissure water overflow. Fissure Water Fissures have developed where the surface is exposed, and they receive the infiltration recharge of at-

mospheric precipitation and the lateral leakage replenishment of surface water. The runoff conditions are affected by the terrain slope and the underlying geological structures and associated fissures. Our experiments show that the sulfate ion concentration is about 300 mg/L in the bedrock fissure water. This level of sulfate accelerates the dissolution of limestone.

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 257–269

259

Guo, Yan, Yu, Liu, Zhou, and Shi Table 1. Hydrogeological parameters of the aquifers.

Aquifer Quaternary Karst water Fissure water

Bottom Depth (m)

Type

Groundwater Level (depth [m])

0.90–4.68 28.65–40.12 38.10–54.25

Unconfined Confined Unconfined

0.80–2.54 25.38–30.41 37.10–44.53

CHARACTERISTICS OF KARST

Figure 2. Tectonic map (a) and topography (b) of the study area.

Initially, 18 boreholes were drilled to a depth of 52 m (Figure 1). The limestone strata were exposed in eight boreholes, and karst caves were exposed in five boreholes with a cavern ratio of 62.5%. The roof rock of the caves lies at depth of 27–47 m. The mean diameter of the caves is 2.61 m, and the maximum is 12.10 m (in Z18). Most of the caves contain different types of filling (Table 2); some are completely filled with hard plastic clay; some are partially filled with soft, plastic, fluid-like silty clay with mud; and others contain silty clay with gravel. Only a small proportion of the caves are empty (filled with water). In order to ascertain the development of karst in detail, we used cross-hole seismic computed tomography (CT), which our study team has been using and improving on. We arranged 33 survey lines to allow observations between one borehole and the neighboring boreholes (Figure 6). The layout is shown in Figure 7. We used a one-hole excitation/two-hole receiving system to effectively reduce test time. In the triggering borehole, an electric spark transmitter was used as the seismic source with a trigger interval of 0.5 m. A string of 12 geophones with a frequency of 100 Hz was placed in the receiving borehole. When the electric spark device was activated, the probe was discharged in the form of gasification and an explosion, which was transmitted outward in the form of elastic waves. The optical signal of the elastic waves was converted to

Figure 3. Field tomogram.

260

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 257–269

Characteristics, Controlling Factors, and Formation of Shallow Buried Karst in Eastern China Table 2. Statistics on karst caves in boreholes in the study area. Borehole No. Z16

Z17

Z18

Depth (m)

Dimension (m)

32.03–32.73 33.93–34.43 34.83–36.03 26.75–27.75 29.25–29.85 32.25–32.65 30.45–42.55

0.7 0.5 1.2 1 0.6 0.4 12.1

Z20 Z22 Z103

44.35–45.05 36.18–40.88 32.05–32.25 44.51–47.40

0.7 4.7 1.2 2.89

Z109

35.50–39.50 27.00–31.50 38.40–40.40

4 4.5 2

Filling Type Empty cave Empty cave Empty cave Silty clay Hard plastic clay Empty cave Silty clay with gravel or with mud Silty clay with mud Clay Silty clay with gravel Weathered debris and clay Clay Silty clay Silty clay with gravel

an electrical signal by the geophones. Seismic records were obtained with the Terraloc Pro 2 seismograph. We then used MATLAB software to analyze the wave velocity and create the CT image. As shown in Figure 8, profile 103-105 is used as an example to display the process of interpreting the field seismic data; the CT image reflects the wave velocity of the elastic longitudinal wave in the section. There are 36 geological anomalies in the wave velocity profile of the cross-hole seismic CT data; these are interpreted as 26 mud-filled caves (including three pairs of beaded caves), three empty caves, and seven fissure zones or loose soil bodies (Figure 9). As illustrated in Figure 9a, good connectivity exists between boreholes Z103 and Z109.

Shape of Karst Caves In this area, most karst caves are oblate or irregularly elliptical in the cross section, and their shape depends on the occurrence and integrity of the rock strata. And when the strata inclination is low and the integrity of the rock is good, the caves tend toward horizontal development; they always have a stratified distribution and have nearly always horizontal oblate shapes. When the inclination is steep and the rock is broken, the caves tend toward vertical development and have nearly always vertical beaded oblate shapes, which are almost always connected. In addition, the type of filling has a large influence on the volume of the cave; when the filling is predominantly muddy clay and the structure is loose, the diameter of the caves is usually 2–5 m; when the filling is soil with gravel, the diameter of the caves is usually in the range of 4–12 m.

Vertical Zonal Distribution The detection results indicate that karst develops along layers and fissures with particular vertical zonation. Two general observations can be made from Figure 10: (1) Karst development in the study area is characterized by an obvious distribution of strong fractures and weak zones in the vertical direction. The depth of the karst is approximately 46 m. The karst is strongly developed at depths of 27–30 and 33–36 m; the number and the size of the caves are small at depths of 24–27, 36–42, and 42–48 m. Karst development is weak in the other layers and dissolution phenomena are uncommon. (2) Most of the caves are fully filled or partially filled. The most developed sections of the caves are between depths of 27–30 and 33–36 m, and the filling rate of the caves is 65%–100%; at depths of 24– 27, 36–42, and 42–48 m, the filling rate is mostly around 70%. Plane Distribution Caves from the section displayed in Figure 9 were separately projected onto the horizontal planes of two separately designated areas within the study site (Areas 1 and 2 in Figure 11). As indicated earlier, 36 geological anomalies (Area 1: 13; Area 2: 23) were detected. The 13 anomalies in Area 1 were interpreted as relatively small, isolated caves with poor connections between them (mean area of each cave was 1.54 m2 ; maximum 3.04 m2 ). In contrast, the 23 anomalies from Area 2 were interpreted as relatively large caves with good connectivity (mean area 3.08 m2 ; maximum 6.89 m2 ). There were 13 geological anomalies densely distributed in the vicinity of boreholes Z103 and Z109 (within Area 2), representing 20.0% and 25.7%, respectively, of the total number of anomalies in both areas. ANALYSIS OF THE DEVELOPMENT MECHANISM OF SHALLOW BURIED KARST Crustal movement in this area during the early Cenozoic was characterized by uplift. The weathering of rock mass was dominated by physical weathering, which was affected mainly by tectonic movement and climate change (Wei and Shi, 2017). Coarse-grained soil, such as gravelly soil, is deposited at the foot of the slope around the mountain, and its permeability is good. Surface water and groundwater are very active. Corrosive CO2 in the groundwater is constantly

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 257–269

261

Guo, Yan, Yu, Liu, Zhou, and Shi

Figure 4. Stratigraphy of the study area.

replenished, and, consequently, the erosion effect of the groundwater is strong. Karst easily develops in shallow limestone to form ancient karst, which varies in size and has stratification. In the middle and late Cenozoic, crustal movement generally declined. The sediments are mainly fine-grained soil of very poor permeability. Con-

262

sequently, surface water has little effect on karst development. However, due to the close proximity to the mountain, rainfall infiltrates along rock fissures on the sides of the mountain, groundwater forms, the hydraulic gradient is large, and runoff is fast; these factors promote karst development in the limestone of the study area mainly in the

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 257–269

Characteristics, Controlling Factors, and Formation of Shallow Buried Karst in Eastern China

Figure 5. Sketch of engineering geological conditions.

vertical direction, and beaded karst caves are easily formed. We have divided the caves in the study area into four types based on their shape, position, and type of filling: Type 1: The caves are fully filled with dense material and are of oblate shape. They have developed below the interface of the soil and rock, and the filling is mainly silty clay with gravel, which has the same properties as the overlying soil layer. Generally, the thick-

Figure 6. Layout of the survey line.

ness of the roof layer of the cave is small, and some sections are connected to the Quaternary stratum. These caves were formed soon after the limestone had become exposed, and the Q3 layer had been compacted. They can be regarded as paleo-karst, and their further development is unlikely. Type 2: The caves are fully filled with two types of loose fillings—silty clay with mud and silty clay with gravel—which have developed at the interface of soil and rock. The caves have developed from the erosion of paleo-karst. They have a particular shape: the upper parts are horizontally oval, and the lower parts are oblate, as shown in Figure 12. Additionally, each cave has good connectivity with other caves, and further development seems to occur readily. Type 3: The caves are semi-filled mainly with silty clay with mud, which is mixed with parts of the overlying soil layer. They are horizontally elliptical below the interface of the soil and rock. It is obvious that these caves were formed after Q3 was deposited and that they are increasing in scale over time to form what is considered new karst. Type 4: The caves are empty and oblate in shape. They have developed in the middle of the limestone and are of small scale. Cleavage and fissures develop in

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 257–269

263

Guo, Yan, Yu, Liu, Zhou, and Shi

Figure 7. Field apparatus and ﬁeld observation system.

the rock mass around them. They are typical of newly developed karst and only slowly increase in size. From both our preliminary analysis of the development mechanism of shallow buried karst and our anal-

ysis of the geological setting of the study area, we conclude that the factors controlling karst development here include topography, structural features, rainfall, groundwater, and illegal water extraction (Figure 13).

Figure 8. Process of the interpretation of the ﬁeld seismic data.

264

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 257–269

Characteristics, Controlling Factors, and Formation of Shallow Buried Karst in Eastern China

Figure 9. Cross sections of karst caves as interpreted from computed tomographic survey data.

Topography The study site is located in the southeast wing of an anticline (southeast of Huishan Mountain and north of the Liangxi River). Because the integrity of the rock is very poor at the core of the anticline, groundwater naturally converges on the wings of the anticline, forming a water-rich zone. The topography then governs the flow of water to the study site and to the Liangxi River. However, because of the poor permeability of the Quaternary layers, groundwater preferentially flows through the fissures of the rocks. Structural Features A fault fracture zone near the study area provides storage space for groundwater within the limestone.

When groundwater is in close contact with a structural plane, dissolution is enhanced, which causes the karst to form ﬁssures and caves to gradually form and expand. These processes cause karst collapse over time (Liu and Zheng, 2002). Faults not only increase the permeability of a karst aquifer but also enhance the movement of groundwater. Rainfall, Groundwater, and Illegal Water Extraction Groundwater conditions are among the most important factors responsible for the development of karst. Annual precipitation in the area is highly seasonal, with May to September being the period of maximum rainfall in Wuxi. The marked alternation of rainy and dry seasons is the main reason for the large ﬂuctuation

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 257–269

265

Guo, Yan, Yu, Liu, Zhou, and Shi

Figure 10. Vertical zonation of karst caves.

in the groundwater level, which corresponds to the vertical distribution of karst; the water depth ranges from 4 to 18 m, as long-term monitoring data have shown (Figure 14) (Qin et al., 2013). In addition, in recent years, illegal extraction of groundwater has become widespread, especially since the study area is located in the center of a densely populated city; the water level is sometimes even below the bedrock surface, leading to further perturbations in groundwater levels. Soluble limestone is distributed throughout the vertical circulation zone of the groundwater. Significant fluctuations in the groundwater level, in particular, will increase the flow rate through the groundwater system and can cause more collapses in the overlying soil (or even scouring out the infill materials) (Kang, 1988). The practice of returning farmland to forest on Chanshan Mountain (due to climate change initiatives) leads to a greater density in vegetation, which effectively reduces the surface runoff from the mountain. That is, most of the rainfall infiltrates

into the mountain and then collects at its foot, but the groundwater cannot seep through because of the impervious clay layer at the surface of the foot of the mountain. Water can therefore ﬂow only along layers of strata and through structural faults and collect in the central depression. During the migration process, corrosive CO2 and SO4 2− in the water will erode the limestone and structural fracture zone, forming karst caves. DISCUSSION The analysis of the shallow buried karst in the Wuxi metro areas indicated the following characteristics of the karst caves, which are similar to other shallow buried karst in eastern China: (1) They are densely developed and of small scale, and the roof rock is usually thin and broken. Although there are a few largescale caves, the overlying rock has mostly collapsed or is incomplete, and the caves are mostly ﬁlled by

Figure 11. Distribution of geological anomalies in aerial view for Areas 1 and 2.

266

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 257–269

Characteristics, Controlling Factors, and Formation of Shallow Buried Karst in Eastern China

Figure 12. Sketch map of the special shape of the caves in the study area.

Quaternary loose sediments. (2) Some of the karst caves consist of paleo-karst that is stable, while others have formed only since the Neogene because of the existence of ﬁssures in the rock. (3) The caves are inﬂuenced by topography, surface water, shallow ground-

water, fissures, and Quaternary sediments. (4) They often develop in the interface between soil and limestone or in limestone that is close to the Quaternary loose sediments, and they contain abundant water. (5) They often have good connectivity with the Quaternary system. (6) The karst develops strongly in the vertical direction, and beaded caves are common. (7) Importantly, the strength of the rock above the caves is often weak, and the stability of the rock mass is poor. According to the above analysis, we know that some caves may continue to grow and become larger because of the continuous erosion of water and further dissolution of the limestone, resulting in the enlargement of the void, in particular, filled caves described earlier as cave types 2 and 3. Other caves (type 4) are more stable and enlarge only slowly. In order to prevent karst-related geo-hazards, engineering solutions focus mainly on the use of grout to fill cavities and stabilize karstic ground (Cooper and Saunders, 2002). However, it would be impractical and wasteful if all parts of the study site were to be treated equally. Therefore, our proposals for underground construction at the site are tailored to each of the two areas already

Figure 13. Sketch of the process of karst development.

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 257–269

267

Guo, Yan, Yu, Liu, Zhou, and Shi

Figure 14. Changing groundwater levels (January 2014–December 2017) in Wuxi city.

identified (Areas 1 and 2), which were defined in terms of the connections between and the scale of the caves (see Figure 10). Our recommendations are as follows: (1) Area 1 (small isolated caves): The scale of caves in this area is small, they have developed in isolation, and there is little connectivity between them. A gunite filling method or a poured concrete plate are recommended for the floor of the tunnels when the geotechnical characteristics of the site are good, and the injection of grout or mortar through the micro-piles is advised when the geotechnical characteristics of the site are poor. (2) Area 2 (intensively developed and well-connected caves): These caves and their connectivity, as well as the associated water permeability, pose the greatest problems to the construction of tunnel sections. In order to allow safe construction through these caves, it is recommended that highcohesion products be injected into the caves or that cave filling be conducted (jet grouting). The treatment of the other caves in this area should use the methods suggested for Area 1. CONCLUSIONS The determination of the spatial distribution, size, and shape of the karst caves is important to conduct safe subway tunnel construction and help prevent karst collapse. The development characteristics of karst in eastern China were obtained. The following conclusions are drawn from this study: (1) The study site is located in a low-lying recharge area in which active groundwater circulation and

268

abundant corrosive CO2 provide ideal geological conditions for the formation of karst. The field cross-hole seismic experiments included 33 survey lines spanning distances of 10.0 m–28.5 m. The results showed that most karst caves were oblate or oval in section and that karst development occurred in vertical zones, the most developed being at depths between 26–30 m and 32–36 m. (2) The study area was divided into two areas based on the number and scale of the caves. The data indicated that karst caves in Area 1 were poorly connected and that their mean area was 1.54 m2 , whereas in Area 2, the caves were well connected and had a mean area of 3.08 m2 . (3) The karst caves of the study area were divided into four types according to their shape, position, and the type of filling. The controlling factors of karst include topography, structural features, rainfall, and groundwater. Moreover, human activities, in particular, the illegal extraction of groundwater, play an important role in accelerating karst formation. In addition, due to the development of karst over different periods, cave collapse also occurred during different periods. There is a good correlation between the occurrence of karst caves and Quaternary sediments, and soil caves are sometimes formed. (4) For the purposes of addressing engineering problems posed by karst geology, the karst terrain in the study area has been divided into Areas 1 and 2; each area was defined according to the scale of and connection between the underground cavities. On the basis of our findings, we have proposed cement grouting methods for the different zones.

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 257–269

Characteristics, Controlling Factors, and Formation of Shallow Buried Karst in Eastern China

ACKNOWLEDGMENTS This work was supported by the Natural Science Foundation of Jiangsu Province, China (grant number BE2015675). Fieldwork was supported by the Wuxi Municipal Design Institute. The authors acknowledge Xiaozhong Shi and Jiaqing Wang for on-site scheduling and graduate students, including Yang Liu, Yinkang Zhou, Zheng Zhang, Yang You, Yipeng Shi, Liangchen Yu, Lei Huang, Haixiang Wu, Zhiwei Jin, Chengliang Li, and Zhuangzhuang Hou, for undertaking the ﬁeld tests. REFERENCES Abdeltawab, S., 2013, Karst limestone foundation geotechnical problems, detection and treatment: Case studies from Egypt and Saudi Arabia: International Journal Scientific Engineering Research, Vol. 4, No. 5, pp. 376–386. Alija, S.; Torrijo, F. J.; and Quinta-Ferreira, M., 2013, Geological engineering problems associated with tunnel construction in karst rock masses: The case of Gavarres tunnel (Spain): Engineering Geology, Vol. 157, No. 5, pp. 103–111. Cai, J. J.; Yan, C. H.; Wang, N.; Yong, S.; Zheng, J.; and Tang, Z. G., 2011, Application of high density electrical method in karst exploration of subway: Journal Engineering Geology, Vol. 19, No. 6, pp. 935–940. (in Chinese) Cooper, A. and Saunders, J., 2002, Road and bridge construction across gypsum karst in England: Engineering Geology, Vol. 65, No. 2, pp. 217–223. Cui, Q. L.; Wu, H. N.; Shen, S. L.; Xu, Y. S.; and Ye, G. L., 2015, Chinese karst geology and measures to prevent geohazards during shield tunnelling in karst region with caves: Natural Hazards, Vol. 77, No 1, pp. 129–152. (in Chinese) Duan, C. L.; Yan, C. H.; Xu, B. T.; Wu, H. R.; and Zou, M. Y., 2013, Cross-hole seismic CT technology for cavern exploration in subway engineering construction application: Geological Review, Vol. 59, No. 6, pp. 1242–1248. (in Chinese) Duan, C. L.; Yan, C. H.; Xu, B. T.; and Zhou, Y. K., 2017, Crosshole seismic CT data ﬁeld experiments and interpretation for karst caves in deep foundations: Engineering Geology, Vol. 228, pp. 180–196. Gong, X.; Tóth, J.; Yang, X.; Yuan, B.; and Feng, D., 2018, A numerical model in predicting the initial karst development in porous limestone: Environmental Earth Sciences, Vol. 77:295, No. 7, pp 1–9. Gutiérrez, F.; Parise, M.; Waele, J. D.; and Jourde, H., 2014, A review on natural and human-induced geohazards and impacts in karst: Earth-Science Reviews, Vol. 138, pp. 61–88. Huang, F.; Zhao, L. H.; Ling, T. H.; and Yang, X. L., 2017, Rock mass collapse mechanism of concealed karst cave beneath deep tunnel: International Journal Rock Mechanics Mining Sciences, Vol. 91, pp. 133–138. Jiang, R. D.; Ge, L. B.; Zheng, J. Z.; and Shao, Y. Z., 2013, Analysis of rainfall water quality in Wuxi area in 2012: Journal Jiangsu Water Conservancy, Vol. 9, pp. 36–38. (in Chinese) Kang, Y. R., 1988, Forming condition of land collapse in karst regions: Carsologica Sinica, Vol. 7, No. 1, pp. 11–20. (in Chinese) Li, S. C.; Zhou, Z. Q.; Ye, Z. H.; Li, L. P.; Zhang, Q. Q.; and Xu, Z. H., 2015, Comprehensive geophysical prediction and treatment measures of karst caves in deep buried tunnel: Journal Applied Geophysics, Vol. 116, pp. 247–257.

Liu, L.; Shi, Z.; Peng, M.; Liu, C.; Tao, F.; and Liu, C., 2018, Numerical modeling for karst cavity sonar detection beneath bored cast in situ pile using 3D staggered grid finite difference method: Tunnelling Underground Space Technology, Vol. 82, pp. 50–65. Liu, J. M. and Zheng, W. Z., 2002, Neotectonic fault control on karst sinking in laiwu iron ore field. Contributions Geology Mineral Resources Research, Vol. 17, No. 44, pp. 282–285. (in Chinese) Pan, F. Y., 1985, karst landforms in southern Yixing: Carsologica Sinica, Vol. 4, pp. 75–81. (in Chinese) Qin, J. G.; Hong, G. X.; Zhang, T.; Sun, L.; Zhang, Y. J.; Shen, S. Z.; and Wu, C. M., 2013, Trend, characteristics and forecast analysis of interannual rainfall in WuXi Railway Station: Journal China Hydrology, Vol. 33, No. 4. pp. 92–96. (in Chinese) Santo, A.; Budetta, P.; Forte, G.; Marino, E.; and Pignalosa, A., 2017, Karst collapse susceptibility assessment: A case study on the Amalfi Coast (southern Italy): Geomorphology, Vol. 285, pp. 247–259. Shao, Y.; Yan, C. H.; and Ma, Q. P., 2015, Analysis of karst geology problems of Nanjing Metro Line 3: Sichuan Building Science, Vol. 41, No. 4. pp. 88–92. (in Chinese) Taheri, K.; Gutiérrez, F.; Mohseni, H.; Raeisi, E.; and Taheri, M., 2015, Sinkhole susceptibility mapping using the analytical hierarchy process (AHP) and magnitude–frequency relationships: A case study in Hamadan province, Iran: Geomorphology, Vol. 234, pp. 64–79. Tang, Z. G.; Yan, C. H.; Xu, B. T.; and Yong, S., 2011, Application of geological radar in karst exploration of Nanjing Metro: Exploration Science Technology, Vol. 6, pp. 49–51. (in Chinese) Wang, N.; Yan, C. H.; Shao, Y.; and Zheng, J., 2011, Karst development law of Nanjing Metro Line 3 and its impact on engineering construction: Jiangsu Architecture, Vol. 6, pp. 77–80. (in Chinese) Wei, Z. F. and Shi, P. Y., 2017, Discussion on controlling factors of karst development in Yaolin Town, Tonglu County, Zhejiang Province: Journal Green Science Technology, Vol. 12, pp. 249– 251. (in Chinese) Xing, X.; Zhou, D.; and Luo, Y., 2014, Karst collapse in Xuzhou, Jiangsu Province and its prevention and control measures: Chinese Journal Geological Hazard Control, Vol. 25, No. 4, pp. 51–58. (in Chinese) You, Y.; Yan, C. H.; Xu, B. T.; Liu, S.; and Che, C. H., 2018, Optimization of dewatering schemes for a deep foundation pit near the Yangtze River, China: Journal Rock Mechanics Geotechnical Engineering, Vol. 10, No. 3, pp. 555–566. Zajc, M.; Pogačnik, Ž.; and Gosar, A., 2014, Ground penetrating radar and structural geological mapping investigation of karst and tectonic features in flyschoid rocks as geological hazard for exploitation: International Journal Rock Mechanics Mining Sciences, Vol. 67, pp. 78–87. Zhang, K.; Tannant, D. D.; Zheng, W.; Chen, S.; and Tan, X., 2018, Prediction of karst for tunnelling using fuzzy assessment combined with geological investigations: Tunnelling Underground Space Technology, Vol. 80, pp. 64–77. Zhou, W. F. and Beck, B. F., 2005, Roadway construction in karst areas: Management of stormwater runoff and sinkhole risk assessment: Environmental Geology, Vol. 47, No. 8, pp. 1138–1149. Zhu, G. Q. and Dong, C. W., 2006, Basic characteristics of karst caves in Zhejiang Province and several problems in tourism development: Ecological Environment Tourism Development, Vol. 24, No. 2, pp. 109–113. (in Chinese)

Environmental & Engineering Geoscience, Vol. XXVI, No. 2, May 2020, pp. 257–269

269

EDITORIAL BOARD Watts, Chester “Skip” F. Radford University, Chair Hasan, Syed University of Missouri, Kansas City Nandi, Arpita East Tennessee State University

Oommen, Thomas Michigan Technological University Sasowsky, Ira D. University of Akron

SUBSCRIPTIONS:

Environmental & Engineering Geoscience May 2020 VOLUME XXVI, NUMBER 2

SUBMISSION OF MANUSCRIPTS

Single copies are $75.00 each. Requests for single copies should be sent to AEG, 201 East Main St., Suite 1405, Lexington, KY 40507.

THIS PUBLICATION IS PRINTED ON ACID-FREE PAPER EDITORS

Brian G. Katz Environmental Consultant Tallahassee, FL 32309 eegeditorbkatz@gmail.com

Cover photo View of the natural bridge in Natural Bridge State Park, Rockbridge County, Virginia. Photo courtesy of Brian Bruckno. See article on page 141.

Volume XXVI, Number 2, May 2020

Abdul Shakoor Department of Geology Kent State University Kent, OH 44242 330-672-2968 ashakoor@kent.edu

ENVIRONMENTAL & ENGINEERING GEOSCIENCE