EEG Journal Vol. XXVII, No. 1 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, 3053 Nationwide Parkway, Brunswick, OH 44212. 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, 3053 Nationwide Parkway, Brunswick, OH 44212. 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. SUBSCRIPTIONS: Member subscriptions: AEG members automatically receive digital access to the journal as part of their AEG membership dues. Members may order print subscriptions for $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. 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. Single copies are $75.00 each. Requests for single copies should be sent to AEG, 3053 Nationwide Parkway, Brunswick, OH 44212. © 2021 by the Association of Environmental and Engineering Geologists

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

EDITORS

Eric Peterson Department of Geography, Geology, and the Environment Illinois State University Normal, IL 61790 309-438-5669 ewpeter@ilstu.edu

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

ASSOCIATE EDITORS Ackerman, Frances Ramboll Americas Engineering Solutions, Inc. Bruckno, Brian Virginia Department of Transportation Clague, John J. Simon Fraser University, Canada Fryar, Alan University of Kentucky Hauser, Ernest Wright State University Keaton, Jeff AMEC Americas Pope, Isaac Book Review Editor Santi, Paul Colorado School of Mines Dee, Seth University of Nevada, Reno

Schuster, Bob Shlemon, Roy R.J. Shlemon & Associates, Inc. Stock, Greg National Park Service Ulusay, Resat Hacettepe University, Turkey West, Terry Purdue University Schuster, Robert Gardner, George Massachusetts Department of Environmental Protection May, David USACE-ERDC-CHL Bastola, Hridaya Lehigh University Berglund, James Montana Bureau of Mines and Geology

Environmental & Engineering Geoscience February 2021 VOLUME XXVII, NUMBER 1 Special Issue on Debris Flows, Part 1 Paul M. Santi and Lauren N. Schaefer, Guest Editors

Submitting a Manuscript Environmental & Engineering Geoscience (E&EG), is a quarterly journal devoted to the publication of original papers that are of potential interest to hydrogeologists, environmental and engineering geologists, and geological engineers working in site selection, feasibility studies, investigations, design or construction of civil engineering projects or in waste management, groundwater, and related environmental fields. All papers are peer reviewed. The editors invite contributions concerning all aspects of environmental and engineering geology and related disciplines. Recent abstracts can be viewed under “Archive” at the web site, “http://eeg.geoscienceworld.org”. Articles that report on research, case histories and new methods, and book reviews are welcome. Discussion papers, which are critiques of printed articles and are technical in nature, may be published with replies from the original author(s). Discussion papers and replies should be concise. To submit a manuscript go to 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. Manuscripts that do not follow the Style Guide and the Instructions for Authors will be returned. 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 The 9 January 2018 debris flow disaster in Santa Barbara County, California. Deep rills and gullies formed on steep hillslopes burned by the Thomas Fire during an extreme short-duration high-intensity storm the morning of 9 January 2018 (top left and right). Starting at 3:45PST, debris flow surge fronts as high as 10 m traversed down thirteen confined canyons, scouring and entraining trees, soil, and alluvium to bedrock (middle left). Boulder-laden surge fronts debouched from confined canyons onto urbanized alluvial fans, damaging or destroying 558 building structures (middle right and lower left), seven bridges, and numerous drainage structures. Debris flows exceeded the capacity of five out of nine debris retention basins. The largest single debris flow volume of >119,000 cm was mostly contained in the Santa Monica debris retention basin (lower right). Photos courtesy of Jeremy Lancaster. See article on page 3.

Volume XXVII, Number 1, February 2021

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.

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

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, 3053 Nationwide Parkway, Brunswick, OH 44212 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 27, Number 1, February 2021 Special Issue on Debris Flows, Part 1 Paul M. Santi and Lauren N. Schaefer, Guest Editors Table of Contents 1

Introduction to Special Issue on Debris Flows Part 1 Paul M. Santi, Lauren N. Schaefer

Observations and Analyses of the 9 January 2018 Debris-Flow Disaster, Santa Barbara County, California Jeremy T. Lancaster, Brian J. Swanson, Stefani G. Lukashov, Nina S. Oakley, Jacob B. Lee, Eleanor R. Spangler, Janis L. Hernandez, Brian P. E. Olson, Mike J. DeFrisco, Donald N. Lindsay, Yonni J. Schwartz, Solomon E. McCrea, Peter D. Roffers, Christopher M. Tran

Alluvial Fan Alteration Due to Debris-Flow Deposition, Incision, and Channel Migration at Forest Falls, California Kerry Cato, Brett Goforth

Time Since Burning and Rainfall Characteristics Impact Post-Fire Debris-Flow Initiation and Magnitude Luke A. McGuire, Francis K. Rengers, Nina Oakley, Jason W. Kean, Dennis M. Staley, Hui Tang, Marian de Orla-Barile, Ann M. Youberg

Forecasting and Seismic Detection of Proglacial Debris Flows at Mount Rainier National Park, Washington, USA Scott R. Beason, Nicholas T. Legg, Taylor R. Kenyon, Robert P. Jost

Water and Sediment Supply Requirements for Post-Wildfire Debris Flows in the Western United States Paul M. Santi, Blaire MacAulay

Measurements of Velocity Profiles in Natural Debris Flows: A View behind the Muddy Curtain Georg Nagl, Johannes Hübl, Roland Kaitna

Combining Instrumental Monitoring and High-Resolution Topography for Estimating Sediment Yield in a Debris-Flow Catchment Velio Coviello, Joshua I. Theule, Stefano Crema, Massimo Arattano, Francesco Comiti, Marco Cavalli, Ana Lucı́a, Pierpaolo Macconi, Lorenzo Marchi

113

Using High Sample Rate Lidar to Measure Debris-Flow Velocity and Surface Geometry Francis K. Rengers, Thomas D. Rapstine, Michael Olsen, Kate E. Allstadt, Richard M. Iverson, Ben Leshchinsky, Maciej Obryk, Joel B. Smith

127

Experimental Investigation on the Impact Dynamics of Saturated Granular Flows on Rigid Barriers Nicoletta Sanvitale, Elisabeth Bowman, Miguel Angel Cabrera

139

The Effects of Particle Segregation on Debris Flow Fluidity Over a Rigid Bed Norifumi Hotta,Tomoyuki Iwata, Takuro Suzuki, Yuichi Sakai

Introduction to Special Issue on Debris Flows Part 1 PAUL M. SANTI* LAUREN N. SCHAEFER Department of Geology and Geological Engineering, Colorado School of Mines, 1500 Illinois Street, Golden, CO 80401

INTRODUCTION We are excited to offer the first of two special issues of Environmental and Engineering Geoscience (E&EG) focusing on the topic of debris flows. These papers were all originally presented at the Seventh International Conference on Debris-Flow Hazards Mitigation (DFHM7), held in Golden, CO (USA), from June 10 to 13, 2019. The conference proceedings were published as Association of Environmental & Engineering Geologists (AEG) Special Publication 28, available at https://mountainscholar.org/handle/11124/173038, and included 134 papers from 17 countries. As guest editors, we selected a subset of these short papers that we thought would be of great interest, in expanded form, to E&EG readers and hope that you will find them valuable and informative. Of the 134 papers presented at the DFHM conference in 2019, a huge range of topics and research tools was represented. Since these conferences are held every 4 years, they provide a snapshot of important and prominent research at the time. The list below provides an idea of the variety of research presented in different categories (note that categories are not exclusive, and many papers fell into more than one group).

Modeling/Numerical Simulation Light Detection and Ranging (LiDAR) Particle Size/Grain Size Monitoring Flume Studies Peak Discharge/Flow Rate Runout and Runout Modeling Rainfall Intensity and Thresholds Seismic Monitoring and Eﬀects Avulsion Early Warning/Prediction Satellite/Remote Sensing Climate Change Instrumentation Susceptibility Mapping

*Corresponding author email: psanti@mines.edu

In addition, nearly half of the studies were case histories, and a strong majority addressed issues of mitigation, management, and flow magnitude (volume). The papers chosen for this special issue represent the spectrum of this research, including numerical simulation and modeling, instrumentation and highresolution topography and LiDAR monitoring, and high-temporal-resolution and high-spatial-resolution flume data, as well as data generated from scrambling up steep, debris-choked canyons. Much of the research focuses on thresholds, feedbacks, forecasting, and process quantification. Of the 10 papers in this first volume, six represent field-based studies, three are flume studies, and one is an overview of one of the most deadly, costly, and important debris-flow events in U.S. history (the January 2018 events in Santa Barbara County, CA, following the 2017 Thomas Fire, which killed 23 and caused over $500 million in damage and recovery costs; see Figure 1). We offer our thanks to the authors of these special issue papers and are very grateful to the many reviewers who contributed their time and expertise to improving the papers:

29% 22% 22% 21% 18% 18% 16% 16% 13% 13% 13% 12% 11% 10% 10%

Figure 1. Example of damage from January 2018 Montecito debris ﬂows in Santa Barbara County, CA (photo credit: Francis Rengers, U.S. Geological Survey).

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 1–2

Santi and Schaefer

Cláudia Abancó Eric Bilderback Holly Brunkal Chen Cheng Francesco Comiti Kahlil Frederick Cui Liang Gao Nico Gray Marcel Hürlimann Jason Kean Luke McGuire S. Poudyal Francis Rengers Luca Sarno Violchen Sepúlveda Andy Take Ann Youberg

Universitat Politècnica de Catalunya–Barcelona TECH Colorado State University Western Colorado University Institute of Mountain Hazards and Environment (China) Libera Università di Bolzano Institute of Mountain Hazards and Environment (China) Hong Kong University of Science and Technology University of Manchester Universitat Politècnica de Catalunya U.S. Geological Survey University of Arizona Hong Kong University of Science and Technology U.S. Geological Survey University of Salerno Servicio Nacional de Geología y Minería (Chile) Queens University Arizona Geological Survey

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 1–2

Observations and Analyses of the 9 January 2018 Debris-Flow Disaster, Santa Barbara County, California JEREMY T. LANCASTER* California Geological Survey, 801 K Street, Sacramento, CA 95820

BRIAN J. SWANSON California Geological Survey, 320 West 4th Street, Suite 850, Los Angeles, CA 90013

STEFANI G. LUKASHOV California Geological Survey, 801 K Street, Sacramento, CA 95820

NINA S. OAKLEY Center for Western Weather and Water Extremes, Scripps Institution of Oceanography, 9500 Gilman Drive, La Jolla, CA 92093

JACOB B. LEE California Geological Survey, 6105 Airport Road, Redding, CA 96002

ELEANOR R. SPANGLER California Geological Survey, 801 K Street, Sacramento, CA 95820

JANIS L. HERNANDEZ BRIAN P. E. OLSON MIKE J. DEFRISCO California Geological Survey, 320 West 4th Street, Suite 850, Los Angeles, CA 90013

DONALD N. LINDSAY California Geological Survey, 6105 Airport Road, Redding, CA 96002

YONNI J. SCHWARTZ U.S. Forest Service, 1190 East Ojai Avenue, Ojai, CA 93023

SOLOMON E. MCCREA PETER D. ROFFERS CHRISTOPHER M. TRAN California Geological Survey, 801 K Street, Sacramento, CA 95820

Key Terms: Post-Fire, Debris Flows, Alluvial Fan, Narrow Cold Frontal Rain Band, Rainfall Intensity, Inundation, Montecito, Thomas Fire, Loss Estimate

*Corresponding author email: jeremy.lancaster@conservation.ca.gov

ABSTRACT The post–Thomas Fire debris flows of 9 January 2018 killed 23 people, damaged 558 structures, and caused severe damage to infrastructure in Montecito and Carpinteria, CA. U.S. Highway 101 was closed for 13 days, significantly impacting transportation and commerce in the region. A narrow cold frontal rain band generated extreme rainfall rates within the western burn area, triggering runoff-driven debris flows that inundated 5.6 km2

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 3–27

Lancaster, Swanson, Lukashov, Oakley, Lee, Spangler, Hernandez, Olson, DeFrisco, Lindsay, Schwartz, McCrea, Roﬀers, and Tran

of coastal land in eastern Santa Barbara County. Collectively, this series of debris flows is comparable in magnitude to the largest documented post-fire debris flows in the state and cost over a billion dollars in debris removal and damages to homes and infrastructure. This study summarizes observations and analyses on the extent and magnitude of inundation areas, debris-flow velocity and volume, and sources of debris-flow material on the south flank of the Santa Ynez Mountains. Additionally, we describe the atmospheric conditions that generated intense rainfall and use precipitation data to compare debris-flow source areas with spatially associated peak 15 minute rainfall amounts. We then couple the physical characterization of the event with a compilation of debris-flow damages to summarize economic impacts.

INTRODUCTION Late autumn and winter wildfires in California create a challenging situation for emergency managers because storms may impact a burn area while emergency response to wildfire is still in progress. The Thomas Fire created this challenge in Ventura and Santa Barbara counties, burning 281,893 acres, with full containment declared on 12 January 2018. A Presidential Disaster Declaration was issued on 8 December 2017. On 6 January 2018, the National Weather Service issued a flash-flood watch for the burn area with anticipated 1 hour rainfall rates of 0.5 to 1.0 in./hr (12.7 to 25.4 mm/hr) (Laber, 2018), just below the U.S. Geological Survey’s (USGS) minimum Santa Barbara region debris-flow triggering rate of 1.2 in./hr (30.5 mm/hr) at the 15 minute duration (Staley, 2018). As the storm moved onshore during the morning of 9 January, intense rainfall associated with a narrow cold frontal rain band (NCFR) passed through eastern Santa Barbara County and western Ventura County, triggering debris flows and sedimentladen flows on steep burned slopes within the Thomas Fire perimeter (Figure 1). The most destructive flows originated on the south flank of the Santa Ynez Mountains and rapidly coalesced into the major canyons that drain southward into the communities of Montecito, Summerland, and Carpinteria, CA. The flows overwhelmed debris-retention basins, clogged culverts and bridge crossings, and overtopped existing channel banks, causing widespread inundation onto fan areas of the Montecito and Carpinteria piedmonts (Figure 2). The flows killed 23 people, damaged or destroyed 558 structures, inundated and closed U.S. Highway 101 (U.S. 101) for 13 days, damaged infrastructure, and disrupted all utility services in the area, including damage to a high-pressure gas line at San Ysidro Creek that

resulted in a large explosion and fire. Emergency response personnel conducted 800 rescues on the morning of the event (SBC, 2018a, 2018b). Following the debris-flow event, the County of Santa Barbara requested assistance from the State of California, Governor’s Office of Emergency Services, to assess impacted areas in the communities of Montecito and Carpinteria, including support for evacuation planning efforts. This study summarizes observations and analyses conducted by the California Geological Survey (CGS) and affiliates to document the large-magnitude debrisflow event in the coastal plain of eastern Santa Barbara County. We incorporate the results of previous assessments published in CAL FIRE (2018a), Kean et al. (2019), and Lukashov et al. (2019), and new data to provide a summary report of debris-flow inundation, runout, velocity, volume, and source area observations; analyses of debris-flow source distribution and triggering rainfall; and documentation of direct and indirect damages and preliminary cost estimates associated with direct damages. Background In California, the most dangerous consequences of alluvial-fan flooding are often underestimated in the land-use planning process in favor of developing prime real estate. Riverine flood hazard paradigms have been applied to western steeplands that issue onto alluvial fans built by successive debris-laden flood and debrisflow deposits, forming the sole regulatory framework used in the planning and mitigation process while creating residual risk (AFTF, 2010). Wildfires create the opportunity for one of those consequences—post-fire debris flows (PFDFs), which are a common threat to southern California communities due to urbanization of alluvial-fan floodplains (USGS, 2005; Lancaster et al., 2015). Examples of PFDF impacts in California including the 1934 New Year’s Day event in the La Crescenta Valley, Los Angeles County (Chawner, 1934), the 1981 Mill Creek disaster in the San Gabriel Mountains (Shuirman and Slosson, 1992), the 25 December 2003 Christmas Day event in San Bernardino, San Bernardino County (Oakley et al., 2017), and the 12 July 2008 Oak Creek event in Inyo County (Wagner et al., 2012). While storm totals are primary drivers for flooding and hillslope processes in unburned settings, PFDFs are often triggered by rainfall durations of 30 minutes or less (Moody and Martin, 2001; Kean et al., 2011), with little lag time between peak 15 minute rainfall and the passage of the debris flows (Kean et al., 2011). This short-duration and high-intensity rainfall causes impacts to denuded hillslopes and soils, including the

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 3–27

The 9 January Debris Flow, Santa Barbara County, California

Figure 1. Overview map of the study area between Montecito Creek in the west and Rincon Creek in the east, showing inundation from the 9 January 2018 debris ﬂows in purple, the Thomas Fire perimeter in black, and the names of the canyons draining to the coastal communities of Montecito, Summerland, and Carpinteria, CA. The county line is deﬁned by the black dashed line along Rincon Creek on the east side of the map. Thomas Fire burn area is shown on inset map.

dislocation of soil particles, initiation of fine-scale slope failures, concentration of runoff, and the development of rills that transition into gullies downslope. Ultimately, this process leads to the formation of surge fronts that build into massive walls of mud and water (Kean et al., 2011, 2013; Staley et al., 2013). With the addition of cobbles and boulders, these flows take on two phases—a fluidized matrix of ash, water, and

fine-grained sediment, and a solid phase of cobbles and boulders; flows grow in size by entraining additional soil, rock, and vegetative matter from channel beds and banks. Where surge fronts interact with drainage infrastructure and roads, they can avulse into areas of human occupancy, tearing homes off their foundations and carrying cars and unanchored anthropogenic features downstream. PHYSICAL SETTING Topography, Climate, and Geology

Figure 2. Damaged home along Buena Vista Creek sitting among a debris-flow boulder field and abundant woody debris. Mud line on building is asymmetric, indicating surge-front depths were greater on the left side of the structure (near the top of the handrail on upper floor deck) than on the right side (photo: Brian Swanson, CGS; all photos taken by USGS and CGS unless otherwise noted).

The study area is in the western part of the Thomas Fire burn area, consisting of large canyons on the south ﬂank of the east-west–trending Santa Ynez Mountains, which drain across the adjacent Santa Barbara Coastal Plain into the Santa Barbara Channel. These canyons include, from west to east, Cold Spring, Hot Springs, Oak, San Ysidro, Buena Vista, Romero, Toro, Arroyo Paredon, Santa Monica, Franklin, Carpinteria, Gobernador, and Rincon creeks. Rincon Creek generally deﬁnes the Santa Barbara–Ventura County boundary (Figure 1). Elevations at the crest of the Santa Ynez Mountains range

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 3–27

Lancaster, Swanson, Lukashov, Oakley, Lee, Spangler, Hernandez, Olson, DeFrisco, Lindsay, Schwartz, McCrea, Roﬀers, and Tran

Figure 3. Regional geologic map of Santa Barbara coastal area. Geologic map units underlying the source areas include Eocene-age Juncal Formation (Tj), Matilija Formation (Tma), Cozy Dell Formation (Tcd), Coldwater Formation (Tcw); late Miocene–age Sespe Formation (Ts), Vaqueros Formation (Tv), Modelo Formation (Tm); Pliocene to Pleistocene–age undifferentiated sandstone (Qss); Quaternary surficial deposits, including Pleistocene- to Holocene-age deposits, including alluvial fan (Qof) and valley alluvium (Qoa); and Holocene-age deposits, including alluvial-fan (Qf) and alluvial-valley deposits (Qa). Dark-red line depicts the fire perimeter boundary from the Thomas Fire. Faintblack lines are geologic unit contacts, and dark-black lines depict faults, solid where accurately located, dashed where inferred, dotted where concealed.

from about 1,067 m (3,500 ft) above Montecito up to 1,463 m (4,800 ft) near the headwaters of Rincon Creek. The area has a typical Mediterranean climate with warm dry summers and cool wet winters, where precipitation occurs almost entirely as rain. Mean annual precipitation ranges from 29.7 in. (75.4 cm) at Juncal Dam in the Santa Ynez Mountains (2,240 ft above sea level [682.7 m]) to 17.7 in. (45 cm) in Santa Barbara (5 ft above sea level [1.5 m]) (WRCC, 2017). This difference in mean annual precipitation demonstrates the strong topographic influence on precipitation in the area. Mean annual precipitation for Montecito is 19.8 in. (50.3 cm) (SBCFCD, 2020). The Santa Ynez Mountains are underlain by a sequence of steeply tilted, east-west–trending Cretaceous to Miocene sedimentary strata (Figure 3; Minor et al., 2009). Formations include Cretaceous unnamed sandstone, mudstone, and conglomerate, which are overlain to the south in succession by the Eocene marine fine-grained Juncal Formation, sandstone-dominated Matilija Formation, fine-grained Cozy Dell Shale, and sandstone-dominated Coldwater Formation, and Oligocene to early Miocene nonmarine sandstone and

mudstone of the Sespe Formation (Dibblee, 1966, 1982). Geomorphology Alluvial fans are categorized as streamflow fans, debris fans, and composite fans based on their geomorphology and origin (Bull, 1977; NRC, 1996). Debrisflow–dominated fans have steeper gradients (generally ࣙ6°, or 10.5 percent) built by successive debris flows and sediment-gravity deposits, with water-borne sediment concentrations generally greater than 50 percent by volume (Pierson and Costa, 1987; Iverson, 1997). Catchment morphometric parameters such as Melton’s number when combined with planimetric length can be used to classify fans that are susceptible to debris flows and debris floods (Wilford et al., 2004; Welsh and Davies, 2011). Alluvial fans dominated by fluvial processes, debris floods and debris flows may be categorized by Melton’s numbers of <0.3, 0.3– 0.6 and >0.6, respectively, with planimetric lengths of <2.7 km for debris-flow–dominated watersheds and >2.7 km for fluvial- and debris-flood–dominated watersheds (Welsh and Davies, 2011).

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 3–27

The 9 January Debris Flow, Santa Barbara County, California Table 1. Watershed data including planimetric length, relief ratio, Melton’s number, and drainage area, showing a general lengthening of watersheds from canyon mouth to crest, and a general decrease in relief ratio and Melton’s number moving eastward from Montecito through Carpinteria, CA.

Watershed

Planimetric Drainage Length Relief Melton’s Area (km) Ratio Number (km2 )

Figure 4. Longitudinal proﬁles from west to east (left to right) through Montecito piedmont aligned to the mountain front/fan apices, showing the steeper composite fans separated by the more gently sloping streamﬂow fans due to the crustal warping on the Mission Ridge Fault Zone; Cold Spring and Montecito creeks are combined, and Hot Springs Creek is not shown.

Montecito area Cold Spring Hot Springs Oak Creek San Ysidro “Buena Vista” Romero Summerland and Carpinteria East Toro Arroyo Paredon Santa Monica Franklin Creek Carpinteria Creek Gobernador Creek Rincon Creek

The geomorphic character of the study area is dominated by large south-draining watershed catchments on the south ﬂank of the Santa Ynez Mountains and the broad piedmont of the Santa Barbara Coastal Plain, which extends from the mountain front to the coastline (see Figures 4 and 5 for Montecito and Carpinteria/Summerland areas, respectively). Plani-

metric lengths of the watersheds range from about 1 to almost 3.7 km above Montecito; near Toro Canyon, the mountain front shifts to the south, and the primary catchments increase in length, ranging up to 9 km above Carpinteria at Rincon Creek (see Table 1). Toro Canyon is an antecedent drainage that cuts through

3.55 2.16 1.13 3.67 2.26 3.03

0.20 0.30 0.34 0.22 0.30 0.23

0.27 0.57 0.57 0.32 0.56 0.38

9.35 1.62 0.48 7.80 1.76 4.89

4.39 5.20 5.21 1.75 6.36 6.79 9.01

0.19 0.15 0.17 0.19 0.13 0.11 0.11

0.32 0.34 0.34 0.32 0.31 0.26 0.27

8.43 9.05 8.97 2.03 12.79 18.73 22.73

Figure 5. Longitudinal profiles through the Summerland and Carpinteria piedmonts showing a doubling in watershed length from west to east upstream of the apices. The length of the entire Toro flow path is 6.5 km, and the length of the entire Carpinteria flow path is 13.4 km. Gobernador Creek, a tributary to Carpinteria Creek, is not shown.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 3–27

Lancaster, Swanson, Lukashov, Oakley, Lee, Spangler, Hernandez, Olson, DeFrisco, Lindsay, Schwartz, McCrea, Roﬀers, and Tran

a ridge uplifted along the Arroyo Parida Fault. Arroyo Paredon is deflected eastward by the same uplifted ridge, and canyons farther east show varying degrees of less prominent reorientation along the fault trace. Longitudinal channel profiles for each drainage are illustrated in Figures 4 and 5. South of the mountain front, much of the piedmont is underlain by alluvial-fan deposits that formed where steep, topographically confined drainages transition to flatter, unconfined terrain. Debris from sediment-laden flows and debris flows accumulates along distributary channels to construct the landform over time, and fans from adjacent canyons coalesce to form a bajada. Fan surfaces range from active to relict, and channels are typically entrenched into older deposits at the fan apices. At Montecito, the piedmont plain is bisected by a discontinuous alignment of low hills uplifted along the Mission Ridge Fault (MRF) during Quaternary time (see pink shaded area on Figure 6). Alluvial fans grade from their apices at the mountain front to the MRF, where distributary fan drainage patterns transition to several confined antecedent channels through this zone of uplift. North of the uplift, upper fan channel gradients generally increase with decreasing drainage size, ranging from 5.3 percent to 6.7 percent (2.9° to 3.8°) below the large canyons of Cold Spring and San Ysidro canyons, respectively, to 7.6 percent (4.3°), 9.1 percent (5.2°), and 12 percent (6.8°) below the smaller canyons of Romero, Oak Creek, and Buena Vista Creek, respectively (Figure 6). Alluvial fans in the Montecito plain range in size from approximately 0.20 km2 to 1.75 km2 . South of the MRF, the drainage system is less constrained, forming a series of secondary fan apices and streamflow fans of relatively lower gradient. Lower fan channel gradients range from 2.7 percent to 3.1 percent (1.5° to 1.8°) south of Mission Ridge. These streamflow fans range in size from 0.2 km2 to 0.9 km2 , with the largest emerging from Montecito Creek (active surfaces make up less than 0.5 km2 of the fan), although San Ysidro Creek has a larger active surface at 0.7 km2 . East of Montecito, upper fan gradients are generally lower and increase with decreasing drainage size, ranging from 1.8 percent to 2.6 percent (1.0° to 1.5°) below the larger canyons of Carpinteria and Gobernador, respectively, and 3.4 percent to 10.3 percent (2.0° to 5.9°) below the smaller drainages. Where the drainage systems are less constrained, and late Pleistocene to Holocene units consist primarily of distal lower fan deposits and alluvial valley deposits, gradients range from 0.8 percent to 5.4 percent (0.5° to 3.1°) (Figures 5 and 7). The Arroyo Paredon fan is similar in size to the larger fans in Montecito, at approximately

1.7 km2 , while to the east, the alluvial fan of Santa Monica Creek is the largest in the study area at 2.8 km2 . To the east and west of these fans, both Franklin and Toro Canyon alluvial fans are much smaller at about 0.3 km2 . Morphometric Relationships Morphometric parameters, including Melton’s number, planimetric length, and fan gradient, were extracted for the Santa Barbara Coastal Plain watersheds to assess flow processes controlling alluvial-fan formation (see Table 1 and Figure 8). In general, Melton’s number ranges from 0.26 to 0.57, and planimetric length ranges from 1.3 to 9.0 km, with the median of the data at 0.37 and 4.1 km, respectively, which suggests that most of these watersheds are dominated by fluvial/debris-flood processes per criteria in Welsh and Davies (2011). Planimetric lengths of these watersheds increase and the Melton’s number decreases systematically from west to east, suggesting that they become progressively more streamflow dominated to the east. In contrast, Hot Springs, Oak, and Buena Vista Canyon watersheds above Montecito exhibit Melton’s numbers and planimetric length of 0.3 to 0.6 and less than 2.7 km, respectively, which indicate their fan landforms are debris flow dominated. Upper alluvial fan gradients at Montecito and at Toro and Santa Monica canyons to the east are generally ࣙ4° (7 percent), indicating that past debris-flow processes are a component of fan formation, which was confirmed by field observations. Wildfire-induced perturbations are known to systematically influence morphometric thresholds correlated with debris-flow generation. For example, a regression study by Gartner et al. (2014) suggested that PFDF occurrence and magnitude are correlated with Melton’s number ranging from 0.05 to 0.71 (with an average of 0.25), and the planimetric length ranges from 0.2 to 12.9 km (with an average of 3.4 km). The watershed morphometric data from the Santa Barbara watersheds generally fall within these ranges, but they indicate the influence of wildfire on geomorphic thresholds as they fall well below the debris-flow thresholds described in Welsh and Davies (2011). Fire and Debris-Flow History Debris flows and/or flooding have inundated portions of Santa Barbara County more than twenty times since 1900 (USACE, 1999; Gurrola and Rogers, 2020), with wildfire exacerbating the effects of flooding and damaging debris flows on the coastal plain in the winters of 1914, 1926, 1964, 1969, 1971, 2017, and 2018. Events are not always widespread, with localized

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 3–27

The 9 January Debris Flow, Santa Barbara County, California

Figure 6. Topography and hillshade map derived from 1-m-resolution light detection and ranging (LiDAR) data of the Montecito area showing inundation in purple. Radial contours (in meters) indicate the presence of alluvial fans. The interaction of surge fronts with bridges, culverts, and roads on alluvial fans is apparent by the numerous large and small avulsions creating the distributary ﬂow paths occupying roads and drainages. Mission Ridge Fault Zone is depicted as pink area across the middle portion of the piedmont; note: this is not a regulatory fault zone (fault data from USGS, 2019). Other red lines depict faults, solid where accurately located, dashed where inferred, dotted where concealed. Names of canyons are in black with white outline, and the abbreviations for the debris retention basins (debris basins) are in black. Debris basin abbreviations: CSDB = Cold Spring; MONB = Montecito; SYDB = San Ysidro; ROMB = Romero. Damaged structures are shown as colored points from DINS data set (CAL FIRE, 2018b), representing state of damage as identiﬁed by the CAL FIRE–led damage assessment team: green = slight; yellow = moderate; orange = high; red = destroyed. Inset rectangles and numbers are keyed to Figure 10.

storms causing substantial damages in one area while completely bypassing another (USACE, 1999). The 9 November 1964 post-fire storm event on the Coyote burn area caused flooding and debris flows within Montecito and San Ysidro creeks, resulting in approximately $500,000 in damages (in 1964 dollars; USACE, 1974, 1999). To the west, a debris flow destroyed 12 homes and six bridges on Mission Creek in Santa Barbara (Burns, 2018). Eyewitnesses reported 20-ft-tall (6.1 m) walls of debris moving down the affected channels at approximately 4.6 m/s (USACE, 1974). On 25 January 1969, a winter storm impacted areas recovering from the 1964 Coyote Fire, flooding a large portion of Carpinteria and causing debris flows within

the major tributaries of Montecito Creek (SBCFCD, 1969). Debris flows along Romero Creek avulsed from the stream channel just north of State Route 192 (SR 192), resulting in downstream flooding, including over U.S. 101. Approximately 100 homes in Montecito, and 250 homes and 20 commercial structures in Carpinteria were damaged as a result. A 40 minute period of heavy rain on 27 December 1971 initiated flooding and debris flows carrying heavy loads of debris and mud in Romero, Garapata, Toro Canyon, Santa Monica, Franklin, and Carpinteria creeks, originating from watersheds that had been burned in the Romero Fire earlier that year (NOAA, 1971; CADWR, 1973). U.S. 101 near Carpinteria was blocked for 8 hours after a 3 ft (0.91 m) wall of mud

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 3–27

Lancaster, Swanson, Lukashov, Oakley, Lee, Spangler, Hernandez, Olson, DeFrisco, Lindsay, Schwartz, McCrea, Roﬀers, and Tran

Figure 7. Topography and hillshade map of the Summerland and Carpinteria area showing inundation in purple. Radial contours (in meters) indicate the presence of alluvial fans that appear steeper, prominent, and active between Toro and Santa Monica canyons, and gently sloping, less prominent, and incised between Franklin and Carpinteria creeks. The Arroyo Parida fault zone is depicted as a pink area traversing the lower portion of the Santa Ynez Mountains; note: this is not a regulatory fault zone (fault data from USGS, 2019). Other red lines depict faults, solid where accurately located, dashed where inferred, dotted where concealed. Debris basin abbreviations: UWTB = upper west Toro; LWTB = lower west Toro ETCB = east Toro; OCDB = Oil Canyon; ARPB = Arroyo Paredon; SMDB = Santa Monica; FRKB = Franklin; GODBD = Gobernador. Damaged structures are shown as colored points from CAL FIRE DINS data set, representing state of damage as identiﬁed by the CAL FIRE–led damage assessment team: green = slight; yellow = moderate; orange = high; red = destroyed. According to the data set, no damaged private structures were mapped east of Arroyo Paredon. Inset rectangles and numbers are keyed to Figure 12.

and water pushed across it toward the ocean. Several roads and bridges were blocked and damaged. In the Carpinteria area, approximately 10–15 families were evacuated, and their homes were damaged, but no deaths or serious injuries occurred. Photos of Toro Canyon Park indicate the presence of boulder fields and woody debris, suggesting debris flows occurred in this event (USACE, 1974). Prehistoric debris flows have been documented based on surficial deposits in the Santa Barbara and Montecito areas. These include a sequence of debris flows thought to be as old as 90–130 ky, a series in Rattlesnake Canyon thought to be 8 to 30 ky, a debrisflow deposit at Rocky Nook Park thought to be 1 ky, and a younger event in Rattlesnake Canyon thought

to be less than 500 years before present (Keller et al., 2020a). While these surficial and stratigraphic prehistoric debris-flow data help to frame minimum recurrence intervals, stratigraphic records are often incomplete due to changes in loci of deposition and erosion in the fan environment (Coe et al., 2003; Wagner et al., 2012). THE 9 JANUARY 2018 EVENT During the storm, debris flows issued from numerous watersheds within the Santa Ynez Mountains (Figures 6 and 7). The first report of debris flows was a 911 call received at 3:47 am (PST) for an explosion resulting from debris-flow impacts to a gas

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 3–27

The 9 January Debris Flow, Santa Barbara County, California

Figure 8. Plot of watershed planimetric length vs. Melton’s number for watersheds north of Montecito and Carpinteria. Horizontal gray line is equal to a planimetric length of 2.7 km. Watersheds having a Melton’s number falling between 0.3 and 0.6, with a planimetric length of <2.7 km, fall into the debris-flood category developed for non-fire settings. All of the data fall within the maximum and minimum values corresponding to post-fire debris-flow events documented in Gartner et al. (2014).

line buried below San Ysidro Creek. A security video camera showed that the debris flow reached a back yard on upper Cold Spring Creek at 3:49 am (Kean et al., 2019). Within the Montecito and Carpinteria areas, the debris flows continued past the canyon mouths and inundated portions of adjacent urbanized alluvialfan surfaces and abandoned channels. In some locations, debris flows traveled up to 5 km to the Pacific Ocean. Of the destroyed structures, 79 had complete structural damage, including 41 structures that were swept off their foundations (Kean et al., 2019). Debris accumulated in low sections of U.S. 101, a major transportation corridor, rendering the section through Montecito impassable by vehicle for 13 days. Between 9 January and 22 January 2018, first-responder personnel conducted search and rescue operations, during which approximately 1,300 individuals were evacuated, and 700 sheltered-in-place (SBC, 2018a, 2018b). The Thomas Fire Presidential Disaster Declaration was amended on 10 January 2018 to include flooding, mudflows, and debris flows. Meteorological Observations A weak atmospheric river was present at the time of the event, demonstrating that catastrophic hydrologic impacts can occur even in the absence of substan-

tial water vapor transport (i.e., a strong atmospheric river) due to synoptic-to-mesoscale forcing. Observations suggest that an NCFR (a narrow band of highintensity rainfall in the vicinity of the cold front) and, to a lesser extent, orographic forcing produced the high-intensity rainfall in this event, with the weak atmospheric river serving as a moisture source (Oakley et al., 2018). At approximately 1:00 am PST on 9 January, the north-to-south–oriented cold front was located just offshore of Point Conception and was propagating eastward across the southern California Bight. Radar imagery revealed the characteristic intense rainfall cores and lighter rainfall gaps of an NCFR parallel to the cold front (Figure 9). The NCFR impacted Montecito and Carpinteria beginning around 3:45 am PST and then began to dissipate near the Santa Barbara– Ventura County line around 4:18 am PST (Figure 9). The subsequent weakening of rainfall intensity likely spared other portions of the burn area from additional catastrophic debris flows. Precipitation in this event was extreme at the 5 and 15 minute durations, with three locations recording >13 mm in 5 minutes, and several records were set at a few stations with 50+ years of observations (Oakley et al., 2018). Several stations demonstrated notable differences in the average return interval for these two durations. Triggering rainfall amounts at the 15 minute duration were as high as 26.16 mm in the Montecito area (Doulton Tunnel gauge) and 21.84 mm in Carpinteria (Carpinteria FS gauge), corresponding to return intervals of 50 to 100 years (Oakley et al., 2018). The Montecito station reported a 200 year precipitation event of 13.72 mm at the 5 minute duration and a 100 year event of 18.54 mm at the 15 minute duration (Oakley et al., 2018; SBCFCD, 2018). Field Observations and Mapping Methodology Field observations, including documentation of debris-flow runout, area, and depth, are often used to classify debris-flow magnitudes and calibrate debrisflow models, and thus they were a primary consideration in our assessment (Hungr et al., 2007; Cannon et al., 2010; Scheidl and Rickenmann, 2010; Kean et al., 2011; Gartner et al., 2014; and Iverson, 2014). The distribution, characteristics, and source-area conditions of the debris flows were documented during three phases. The first “event response” phase was completed as a collaboration between the USGS and the CGS in the first 12 days after the event, documenting perishable features in the Montecito inundation area at 1,584 stations (Kean et al., 2019). The second

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 3–27

Lancaster, Swanson, Lukashov, Oakley, Lee, Spangler, Hernandez, Olson, DeFrisco, Lindsay, Schwartz, McCrea, Roﬀers, and Tran

Figure 9. National Weather Service radar showing (A) the NCFR impacting the western portion of the Thomas Fire burn area at 3:45 am PST, 9 January 2018; (B) NCFR impacting the Summerland and western Carpinteria area at 4:07 am PST; and (C) NCFR breaking up at 4:18 am PST as it crosses the Santa Barbara and Ventura County line as it moved east. Annotations on left frame are Santa Barbara (SB), Montecito (M), Carpinteria (C), Santa Barbara County line at Rincon Point (RP), Ventura (V), and the National Weather Service Radar location (KVTX). Data are from: https://www.cnrfc.noaa.gov/radarArchive.php.

phase consisted of field mapping debris-flow source areas and channel conditions from Montecito to Rincon Creek at 458 stations. Phase three consisted of assessing the extent of debris-flow initiation and inundation across the remainder of the Thomas Fire area, establishing triggering rainfall amounts, and refining damage and cost data from Lukashov et al. (2019). In all phases, field observation data on debris-flow occurrence, scour depth, maximum flow depth, channel gradients, peak flow heights, and super-elevation were collected in the channel areas; field observation data on geometry and extent of rills, sheet-flow erosion, microlevees, inter-rill areas, soil depth, slope gradient, and slope geometry were collected in hillslope areas above Montecito. Montecito Area Hillslopes Hillslopes are generally underlain by three primary material types with respect to erosion characteristics, which correspond to Maymen-Rock Outcrop Complex soils (either 75–100 percent or 50–75 percent slopes). These include: (1) steep slopes (>75 percent or 36.9°)

dominated by rock outcrops that are underlain primarily by resistant sandstone beds of the Coldwater or Matilija formations; (2) slopes mantled by angular to subangular, cobble- to boulder-size rocky debris, where this debris was derived from underlying weathered sandstone beds of the Coldwater or Matilija formations, sandstone interbeds of the Juncal Formation or Cozy Dell Shale, or secondary debris-fan and colluvial deposits; and (3) moderately steep slopes (47 to 75 percent or 25° to 37°) underlain by dominantly finegrained sections of the Juncal Formation and Cozy Dell Shale, which commonly form more geomorphically uniform slopes with a thin mantle of soil. Slopes Underlain by Rock Outcrops—Surface runoff on slopes underlain by rock outcrops coalesced into intervening gullies that were typically scoured of loose debris in the upper reaches, exposing either bedrock or consolidated gravelly soils. In the lower reaches, the gullies either fed into larger channels, or they conveyed the scoured debris through rocky colluvial deposits or small fan deposits, sometimes leaving rocky levee deposits above the channel banks. Slopes Mantled by Rocky Debris—On rocky slopes, runoff was observed either to flow around the clasts,

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 3–27

The 9 January Debris Flow, Santa Barbara County, California

scouring away much on the intervening finer debris, or to form coarse, meandering rills. The majority of cobble and larger size clasts on rocky slopes did not appear to have moved much during the debris-flow event. This lack of mobilization suggests that nearly all of the large debris (boulders) observed in the inundation zone was entrained from existing coarse deposits within the channel areas. Slopes Underlain by Fine-Grained Bedrock— Slopes underlain by fine-grained bedrock typically exhibited varying degrees of rill (discrete flow paths) or sheet-flow erosion, depending on the soil thickness, slope gradient, and slope geometry (Figure 10A). Rill geometry varied depending on whether the plan-form slope geometry was convex (divergent), concave (convergent), or planar. Rill and/or sheet erosion began just a short distance below the ridgelines. The proportion (area) of slopes displaying rills or sheet flow typically increased with slope gradient, and the width of erosion paths commonly increased where the gradient increased downslope. The depth of erosion typically varied with the original depth of soil that was disaggregated by heating during the fire, which was controlled in part by original soil thickness and soil burn severity, and hence by slope gradient and pre-fire vegetation density. Soils remaining below the disaggregated layer were typically more cohesive and contained fine roots, which impeded erosion during the heavy but brief rain event. On some heavily eroded slopes, rills scoured below the disaggregated soils into low-cohesion soils with fine/small roots. Rill depth only increased slightly downslope on planar or divergent slopes. However, rill depths did increase on convergent slopes and swales, where rills merged downslope to form deeper rills and gullies, or where slope gradient decreased and/or the depth of loose soil increased downslope, commonly resulting in deeper, narrower, more meandering rills. The crosssectional area of rills did commonly increase downslope on all slope types, primarily reflecting the development of wider rills in the loose, disaggregated soils, as noted in Keller et al. (2020b). Rill and inter-rill dimensions were recorded at six, 30-m-long transects located on soils of the Juncal Formation in the upper reaches of Cold Spring and San Ysidro canyons on slope gradients ranging from 17° to 37°. The ratios of rill width to maximum depth are summarized in Figure 11. The percent of slope eroded ranged from 51 to 83 percent, although the presence of concentrated gravel mantling inter-rill areas suggests that some finer soils were also winnowed away between rills. Median width-to-depth ratios ranged from about 8 to 13, with median rill widths ranging from 50 to 100 cm and median rill depths ranging from 4 to 10 cm at each transect.

Micro-levees were locally observed along the margins of rill and sheet-flow paths. Micro-levees appeared to be most common where soils were thicker and/or where slope gradients decreased downslope; microlevees were typically less than 10 cm in height in gravelly and sandy soils, but they locally ranged up to about 20 cm high along the larger rills and gullies. Observation of soils in unburned areas indicated that little to no rilling or other erosion occurred under the vegetation canopy during the rainfall event and that gravel was commonly present in the soil profile, but gravel pavement was not well developed in comparison with inter-rill areas in the burn area. Natural hydrophobicity was observed where tested at the base of the shallow organic horizon in these unburned areas. Source Channels The maximum depth of inundation during the debris-flow event was recorded at several stations in each of the subject canyons north of Montecito. In the two largest canyons (Cold Spring and San Ysidro), inundation depths ranged up to about 10 m in the lower reaches and generally decreased upstream into the headwaters; however, greater inundation depths of up to 14.7 m were locally observed where channel constrictions or bends caused super-elevated conditions. In the smaller drainages, inundation depths reached observed maxima of 9.3 m in Romero Canyon, 6.4 m in Hot Springs Canyon, and 6.2 m in Buena Vista Canyon (Table 2). The source channels were dominated by transport and scour processes with overbank deposits preserved on localized inset alluvial surfaces as thin drapes of sandy mud and gravel up to 20 cm thick and boulders stranded against trees or other obstacles. Scour depths typically ranged from 0.5 to 2.0 m with an observed maximum of 3.0 m, and exposed bedrock along more constricted reaches (Figure 10E). Trees surviving in the flow path were marked by mud patinas and scarring, which helped to define flow depths. Evidence of significant past historic flows includes healing tree scars in Romero Canyon and pre-event overbank boulder deposits and levees in Hot Springs Canyon. Older debris-flow deposits were widely observed in channel banks, and pre-event levees were found at many locations above the channel banks. Debris Basins Constructed debris-retention basins (debris basins) present at the time of the 9 January debris flow included basins at Cold Spring, San Ysidro, and Romero canyons, and on Montecito Creek (Figure 4). All of these basins were reportedly cleared of debris prior to

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 3–27

Lancaster, Swanson, Lukashov, Oakley, Lee, Spangler, Hernandez, Olson, DeFrisco, Lindsay, Schwartz, McCrea, Roﬀers, and Tran

Figure 10. Photo locations from inset rectangles on Figure 4. (A) Aerial photo showing deep rills and gullies in soils formed on shale and siltstone beds (Tj on geologic map), where eroded U.S. Forest Service San Ysidro Trail on hillslope is roughly 1.5 m in width (San Ysidro Canyon, north is up). (B) View west toward bridge deck knocked off abutments by debris-flow surge front on Hot Springs Creek; note: boulder remnant on left side of abutment were stranded after surge front removed deck. (C) View south of car jammed into home on Hot Springs Creek. (D) View northeast of lower piedmont gently sloping streamflow fan with broad fine-grained debris deposition through citrus orchards and U.S. 101 to the south. (E) View south of channel alluvium scoured down to moderately cemented Coldwater Sandstone bedrock (Tcw on geologic map) on upper Romero Creek; note: cobble-boulder alluvium from past debris flows mantling Tcw bedrock. (F) Aerial photo of San Ysidro Creek on the east side and running north-south and parallel to Randall Road, where numerous homes were destroyed or swept away; San Ysidro Creek crosses East Valley Road (SR 192) in the lower right side of the photo, and arrow points to boulder having a maximum diameter of about 6 m. (G) View of boulder debris field indicating surge-front deposition up to eves of a home located on Montecito Creek.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 3–27

The 9 January Debris Flow, Santa Barbara County, California Table 2. Montecito inundation data based on field observations and mapping, modified from Kean et al. (2019). Runout distance was taken as the full length of the flow path on which debris flows were observed. Surge-front distances were taken from the topographic break at the mountain front to the first and last series of boulder fields deposited in the creeks and on the alluvial fans, as interpreted from observations in this study and described in Kean et al. (2019). Maximum inundation depths represent confined channels rather than overbank areas observed after the event and do not consider channel modification by scour and transport due to the debris flows and subsequent clear-water recessional flows. A k value of 3 was used in the forced vortex equation to estimate debris-flow velocity.

Channel Name Montecito Cold Spring Hot Springs Oak Creek San Ysidro Buena Vista Romero

Runout Distance (km)

Runout of Boulder Surge-Front Deposition (km)

Maximum Depth (m) and Velocity (m/s) in Canyon

Maximum Depth (m) and Velocity (m/s) on Fan

Maximum Boulder Dimension on Piedmont (m)

4.2 NA NA 1.8 4.2 4.4 5.0

0.5–2.5 NA NA 0.15–1.0 0.75–1.45 0.0–0.7 0.0–1.4

NA/NA 10.1/6.4 6.4/4.7 1.5/NA 14.7/6.4 4.6/5.4 9.3/6.6

10.0/NA NA/NA NA/NA 3.0/3.0 7.0/2.5 5.4/NA 7.0/4.0

4.8 4.8 2.2 1.2 6.0 3.0 2.1

NA = not applicable.

the storm, but they were filled to capacity and overtopped by debris during the 9 January debris-flow event. Inundation Zone The debris-flow inundation covered an area of 3.15 km2 , and flows from individual canyons reached the ocean up to 5.0 km from the mountain front

Figure 11. Box-and-whisker plots showing width-to-depth ratios of hillslope rills based on hillslope erosion transects in Cold Spring and San Ysidro canyons. Central orange line is median of data, and upper and lower limits of box are deﬁned by 25th and 75th percentiles; whiskers illustrate upper and lower quartiles, and dots represent statistical outliers. All transects were measured in the Juncal Formation, corresponding to Maymen-Rock Outcrop Complex soil series on generally planar to slightly convex slopes in plan curvature (CAL FIRE, 2018). Plot locations are CS-T1: 34.4880°N, 119.6664°W; CS-T2: 34.4854°N, 119.6554°W; CS-T3: 34.4733°N, 119.6393°W; SY-T1: 34.4718°N, 119.6015°W; SY-T2: 34.4708°N, 119.6021°W; SY-T3: 34.4709°N, 119.6070°W. Data were originally plotted by Francis Rengers, USGS.

(Table 2). Surge-front depths in the upland canyons ranged from 1.5 to 14.7 m in the confined canyons to 3 to 10 m on the piedmont. Trunk channels acted as primary conduits of the event flows, but flows were commonly at or above the capacity of previously existing channel banks and overflowed onto elevated alluvial-fan or terrace surfaces bordering the trunk channels (Figures 6 and 10). Overbank flows and avulsions occurred at channel constrictions, including culverts, bridges, and associated bridge abutments (Figure 10B), and outer banks of channel bends, where the phenomenon of super-elevation generated increased debris-flow height. Overbank flows inundated adjacent fan surfaces and opportunistically occupied roads and dormant distributary channels (Figures 6, 10D and F). Prominent avulsions occurred at Buena Vista Creek, at Lilac Drive and Alisos Drive, San Ysidro Creek at Randall Road, at U.S. 101 from both San Ysidro and Montecito creeks, and at Oak Creek north of San Ysidro Road (Figure 6). Channels were generally pathways of debris transport and variable scour with local deposition upstream of choke points, while overbank surfaces were dominantly areas of transport and deposition. The maximum inundation elevation on alluvial surfaces was consistently higher than the final surface elevation of deposits left after the event due to the passing of elevated surge fronts and dewatering of deposits. Debris-flow deposits observed on inundated surfaces were typically bimodal. The fines fraction consisted of dark-gray, muddy sand with ash and variable gravel content; gravel content diminished with increasing distance from the mountain front; laboratory testing of debris-flow matrix samples indicated a classification of silty sand (SM) per the Unified Soil Classification System (Kean et al., 2019). The coarse

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 3–27

Lancaster, Swanson, Lukashov, Oakley, Lee, Spangler, Hernandez, Olson, DeFrisco, Lindsay, Schwartz, McCrea, Roﬀers, and Tran

fraction consisted of small to large boulders, up to a maximum documented diameter of 6 m; larger boulders may have been captured in the debris basins, but the dimensions of boulders removed from the basins were not systematically documented. Cobblesize clasts were conspicuously uncommon among the surface deposits, but they may have been more plentiful buried within the debris basins. Boulders composed of strongly indurated sandstone derived primarily from the Matilija and Coldwater formations were deposited on alluvial surfaces and channel banks, accumulating as levees, boulder fields, stacked rocks, or isolated rocks in more distal areas (Figure 10F and G) and providing evidence of surge-front deposition. Boulders were subangular to subrounded and oblong to tabular in shape. Large boulder field deposits were concentrated at the confluence of Cold Spring and Hot Springs canyons, north of SR 192, near Hot Springs Road along Montecito Creek, and along San Ysidro Creek both north and south of SR 192, including Randall Road (Figure 10F). Less concentrated boulder deposits were observed elsewhere along these creeks down to San Benito Way along Montecito Creek and down to San Leandro Way along San Ysidro Creek. Boulder deposits were also observed along Buena Vista Canyon north of SR 192, and sporadically along the upper reaches of Romero Canyon and below the confluence of Romero and Picay creeks.

Summerland and Carpinteria Source Channels Slope conditions in the upper reaches of the canyons east of Montecito were not observed in detail in the field during this study owing to limitations on time and access. However, review of imagery and light detection and ranging (LiDAR) data indicate that extensive rilling and erosion occurred during the 9 January storm, similar to the canyons north of Montecito. Field observations and imagery indicate that debris flows formed in each of the canyons as well; however, the down-canyon extent of the flows decreased from west to east. This in part reflects decreasing rain intensity as the rain band dissipated to the east, the increasing length of the trunk drainages to the east, and smaller proportions of watershed burned. The longer drainages at lower gradient allowed more coarse debris to drop out of the flows and transition into hyperconcentrated flows or flood flows downstream. Inundation depths up to 6.7 m were locally observed where channel constrictions or bends caused super-elevated conditions (Table 3).

Debris Basins Constructed debris basins are present on Toro Canyon east, Toro Canyon west (upper and lower), Oil Canyon and Arroyo Paredon (north of the confluence with Oil Canyon), Santa Monica Canyon, Franklin Creek, and Gobernador Creek (Figures 7, 12B and C). These debris basins, except for Oil Canyon, which has not been maintained, were reportedly cleared of debris prior to the event. Debris flows tested or exceeded the capacity of all the debris basins except the lower west Toro, Franklin, and Santa Monica basins. Farther east at Gobernador Canyon, bouldery debris mostly settled out within the upper portion of the broad, unconfined basin footprint (Fig 12C), but flood flows carried large woody debris over the bollards, much of which accumulated against bridges downstream. Inundation Zone Substantial debris flows and sediment-laden flows occurred in all of these canyons except Rincon Creek (Figure 7), with inundation covering an area of 2.41 km2 . Surge-front depths ranged from 5.0 to 6.9 m in the confined canyons to 1.2 to 5 m on the piedmont (Table 3). Out-of-channel flow and inundation of structures due to avulsion were less extensive than in the Montecito area, and none resulted in fatalities. The only gauged runoff in the entire coastal plain was recorded in Carpinteria Creek at Casitas Pass Road, with a maximum discharge at 4:45 am PST reaching a reported stage height of 16.7 ft (5.0 m) and initially a peak discharge of about 15,000 ft3 /s (424.8 m3 /s), which is currently under review (USGS Gauge ID# 11119500) (Laber, 2018). This was roughly 40 minutes after the peak 5 minute duration rainfall was recorded at the Carpinteria rain gauge. Wood- and sedimentladen surge fronts were conveyed for 2.4 km in Carpinteria Creek and deposited at U.S. 101 and Carpinteria Avenue. Debris-flow inundation occurred due to numerous avulsions along the lower portion of the piedmont south of Toro Canyon (Figure 7). Farther east, a debris-flow avulsion occurred on an inset alluvial surface of the Arroyo Paredon fan (Figure 12A). At the location of the apex avulsion, additional boulder debris was deposited and backfilled the previously entrenched fan head; downstream, additional boulders and woody debris were deposited against the bridge crossing at Foothill Road (SR 192), and this constriction enhanced avulsions into adjacent downstream areas. Flows that reached the frontage road (Via Real) north of U.S. 101 were mostly muddy with woody debris. Farther east at Santa Monica, Franklin, and Carpinteria creeks, flows were contained near the

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 3–27

The 9 January Debris Flow, Santa Barbara County, California Table 3. Summerland and Carpinteria inundation data based on field observations and mapping. Runout distance was taken as the full length of the flow path on which debris flows were observed. Surge-front distances were taken from the topographic break at the mountain front to the first and last series of boulder fields deposited in the creeks and on the alluvial fans, as interpreted from observations. A value of zero for runout distance indicates that debris flows did not travel beyond the fan apex or topographic break at the mountain front. Significant flows of finer-grained sediment and woody debris discharged from spillways at both the Santa Monica and Franklin debris-retention basins. These flows were deposited in the Carpinteria Salt Marsh Preserve.

Channel Name

Runout Distance (km)

Apparent Boulder Surge-Front Deposition (km)

Maximum Depth (m) and Velocity (m/s) in Canyon

Maximum Depth (m) and Velocity (m/s) on Fan

Maximum Boulder Dimension on Piedmont (m)

Toro Arroyo Paredon Santa Monica Franklin Carpinteria Gobernador Rincon

4.0 2.6 0.0 0.0 3.4 NA 0.0

0.0–1.0 0.0–0.8 Debris basin Debris basin In canyon Debris basin No debris ﬂow

5.0/5.7 4.3/NA 6.7/4.2 NA/NA NA/NA 4.0/5.0 No debris ﬂow

3.5/3.3 3.0/NA 1.2/NA NA/NA 5.0* /NA 4.7/4.5 2.3/NA

3.0 3.9 NA/NA NA/NA 3.0 3.0 NA/NA

NA = not applicable. * Depth from Carpinteria Creek is from USGS gauge record.

mountain front by debris basins, and outflows below the mountain front consisted of primarily finegrained sediment and woody debris with scattered boulders. Debris-Flow Volume and Magnitude The total volume of debris generated by flows in the Santa Barbara Coastal Plain is estimated at 1,498,000 m3 . This volume includes the reported total volume of 260,342 m3 of debris removed from debris basins (SBCFCD, 2018) and 126,008 m3 removed from creek channels (USACE, 2018), but it does not include the volumes of debris removed from U.S. 101, or the volumes that reached the ocean, which were not directly documented. In addition, sediment thicknesses recorded at 373 post-event field stations were used to estimate debris volume distributed over inundated areas in Montecito and Summerland. The total mapped inundation area is 5,560,000 m2 (revised from Lukashov et al., 2019), and the average deposit depth is 0.3 m with a standard deviation of 0.53 m. Because the data do not have a normal distribution and are heavily skewed by large outliers, a median thickness of 0.2 m was adopted (Schiff and D’Agostino, 1996), which gives an estimated deposited debris volume of 1,112,000 m3 . Jakob (2005) classifies boulder debris flows in a volume-based logarithmic scale up to magnitude 6, where magnitudes 7–10 are used only for volcanic debris flows. However, the distributed nature of source areas combined with depositional overlap of debrisflow deposits restrict the ability to separate material by watershed as used in Jakob (2005), so they are considered here as an aggregate event. The cumulative es-

timated volume for this event is greater than 1,498,000 m3 , and the aggregate inundation area is over double the value used for the magnitude 6; therefore, we classify the event as a magnitude 6 debris flow. In comparison, flows following the 1964 Coyote Fire and the 1971 Romero Fire each impacted an estimated area of greater than 2 km2 for a magnitude 5 classification (Jakob, 2005; Lancaster, 2018). Volumes of flows from individual canyons were poorly constrained because of the depositional overlap, and most debris basins were overwhelmed or lacking. Two exceptions were west Toro Canyon (upper and lower) and Santa Monica Canyon, where the basins were large enough to contain the debris-flow material. Using the empirical model of Gartner et al. (2014), mean volume estimates of 40,681 and 100,437 m3 were developed using 40 mm/hr rainfall intensity at west Toro Canyon and Santa Monica Canyon, respectively (Table 4). The rainfall intensity documented near these canyons was between 79.24 and 104.64 mm/hr (Oakley at al., 2018). The observed volume was 53 percent of the predicted volume at west Toro and 119 percent of the predicted volume at Santa Monica, which are within the standard deviation reported by the USGS. Large amounts of riparian woody debris were scoured from channel networks and transported into the built environment, creating debris dams at culverts and bridges. The quantity of woody debris is unknown and cannot reliably be taken as a percentage of the total volumes presented, as woody debris was largely absent from the overtopped debris basins, and was not disseminated among the debris-flow deposits, but rather formed piles and rack deposits against structures, trees, and other obstacles.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 3–27

Lancaster, Swanson, Lukashov, Oakley, Lee, Spangler, Hernandez, Olson, DeFrisco, Lindsay, Schwartz, McCrea, Roffers, and Tran Table 4. Debris basin removal volumes compared with USGS mean debris-flow volume estimates using empirical equations from Gartner et al. (2014) and the maximum rainfall intensity input in the model. Percent of USGS predicted volume was calculated between clean-out volume and the predicted volume. “Greater than” symbol denotes where debris flows filled and overtopped the debris basin, and thus locations where the removed volume accounts for a part of the volume that inundated downstream areas.

Debris Basin* Cold Spring San Ysidro Romero W Toro Upper & Lower E Toro Arroyo Paredon Santa Monica Franklin Gobernador§

Construction Year

Removal Volume† (cm)

Capacity Exceeded or Contained

USGS Mean Volume (cm) at 40 mm/hr†

Percent of USGS Predicted Volume

1964 (USACE) 1964 (USACE) 1971 (USACE) 1971 (USACE) 1971 (USACE) 1971 (USACE) 1977 (USDA) 1971 (USACE) 1971/2008

18,947 16,627 33,041 21,555 11,584 18,371 119,271 888 8,722

Exceeded Exceeded Exceeded Contained Exceeded Exceeded Contained Contained Contained

60,207 84,883 45,387 40,681‡ 29,055 43,012 100,437‡ 14,309 240,303

>32% >20% >73% 53% >40% >43% 119% 6% 4%

Most basins were constructed with grouted boulder dam and culvert; Montecito debris basin volume (1,778 cm) and Plunge Pool volume (9,557 cm) not included. † Volume of debris removed after January 9 event, as reported by USACE. ‡ USGS mean volume compiled from multiple contributing sub-watersheds. § Gobernador was originally constructed in 1971 with 25,275 cubic yard (19,324 m3 ) capacity, and then it was modiﬁed in 2008 to allow ﬁsh passage.

Debris-Flow Velocities Debris-flow velocity estimates were calculated from observations of super-elevation at 52 channel bend locations, expanding on the data set for Montecito summarized in Kean et al. (2019). The super-elevation, or the slope between the high- and low-flow marks of a debris flow around a curving channel, can be used to back-calculate and estimate velocity with the forced-vortex equation (Chow, 1959; Hungr et al., 1984; Prochaska et al., 2008; Kean et al., 2019); procedures described in Kean et al. (2019) were followed herein, including adoption of a k correction factor of 3. Estimated debris-flow velocities ranged from 4.2 to 6.6 m/s within the source canyons, and from 2.5 to 4.0 m/s on fan areas at Montecito, Summerland, and Carpinteria (see Tables 2 and 3). In addition, review of video footage of unconfined flows along Olive Mill Road (asphalt-concrete paved) from the lower portion of Montecito Creek revealed a velocity of 3.7 m/s, which is within the observed range of super-elevation velocities estimated elsewhere in fan areas. Frequency To develop quantitative hazard information, a focus on Holocene climate, anthropogenic climate change, historic and prehistoric data is needed (Jakob et al., 2017; Williams et al., 2019). Charcoal accumulation rates from marine sedimentary cores off the coast of Santa Barbara and documented historic wildfires indicate the average return period for large wildfires

(>50,000 acres) is approximately 21 years for preEuropean settlement periods, 29 years for the European settlement period, and 23 years for the 1900s (Byrne, 1979; Mensing et al., 1999). A review and statistical analysis of wildfires confined to the Santa Barbara Coastal Plain (from Gaviota Beach to the Ventura County line) suggest that from 1913 to 2017, more than 63 wildfires greater than 0.04 km2 (10 acres) occurred in the region (CAL FIRE, 2018c; Lancaster, 2018). Using these wildfire data, a statistical analysis of wildfire return period suggests that the annual return period for a wildfire of 7.8 km2 (2,500 acres) is about 2.7 years, an area slightly larger than the San Ysidro Creek watershed and similar in size to the El Capitan watershed that produced a PFDF on 20 January 2017 (Schwartz, 2017). Whereas the statistical return period equivalent to the area within coastal-draining portions of Santa Barbara County consumed by the Thomas Fire (89 km2 or 22,000 acres) is about 10 years (Table 5). Table 5. Fire size and annual return period for the Santa Barbara Coastal Plain using fire-history data from CAL FIRE (2018c) and analysis from Lancaster (2018).

Fire Size (acres) 500 5,000 10,000 22,300 50,000

Annual Probability of Exceedance

Annual Return Period (years)

53.74 28.39 20.75 11.92 3.03

1.86 3.52 4.82 8.39 32.96

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 3–27

The 9 January Debris Flow, Santa Barbara County, California

duration for the 9 January storm rainfall (Perica et al., 2011; Oakley et al., 2018) and the fire frequency data presented in Table 5, the joint probability for debris flows being triggered in a single watershed, where joint probability equals P(A) × P(B), would be 0.0066 (150 year RI) and 0.0033 (300 year RI) for the 50 and 100 year storm event, respectively. The joint probability for debris flows being triggered in multiple watersheds at the scale of the Thomas Fire’s footprint in Santa Barbara County would be 0.002 (500 yr RI) and 0.001 (1000 yr RI) for the 50 and 100 year storm event, respectively. Considering that debris flows in multiple watersheds occurred after wildfires in 1964, 1971, and 2018, an average return period for widespread damaging or catastrophic debris flows based on physical evidence may be less than 20 years. The difference in recurrence between coupled fire and triggering precipitation and historic PFDF data highlights the importance of a multi-faceted assessment approach, including a cumulative magnitude-frequency analysis, as suggested in Jakob et al. (2017). DEBRIS-FLOW SOURCES AND RAINFALL

Figure 12. Photo locations from inset rectangles on Figure 5. (A) View north of Arroyo Paredon Creek north of Foothill Road, showing boulder fields from surge-front deposition in previously entrenched channel. Maximum boulder size observed in this general area was 3.9 m. (B) View south of Santa Monica debris basin (SMDB on map) full of fine-grained slurry with piles of racked trees and woody debris around the spillway; massive boulder field located with arrow and shown as inset photo indicates where some of the surge fronts came to rest as flow front lost both confinement and gradient (photo taken by Lael Wageneck, SBCFCD). (C) View north of Gobernador debris basin (GODB on map) with concrete breakers at the outlet, and boulder deposition from debrisflow surge fronts at the upstream junction of the creek and debris basin; largest boulders are 2.5 m in maximum dimension.

The historical frequency of both wildﬁre and triggering rainfall provides context to quantitative hazard assessments (Kean and Staley, 2019). Using the return intervals (RI) for rainfall intensity at the 15 minute

Sources of debris-flow material include residual and colluvial soils from steep slopes, and alluvial wash and valley deposits (Cannon et al., 2003; Santi et al., 2008). Hillslope sediment is entrained and delivered to channels by rilling and gullying that result from intense surface erosion during runoff. Research suggests that rills develop as micro-slope failures initiated by small slips at depth due to increases in pore pressure (e.g., Wells, 1987; Gabet, 2003). Within channel networks, debris-flow initiation is poorly understood. As with hillslope rill initiation, several researchers suggest that debris-flow initiation may result from mass failure of the channel bed or, alternatively, as a result of grain-by-grain entrainment and bulking by hydrodynamic forces from a critical discharge of water (e.g., Armanini and Gregoretti, 2005; Lamb et al., 2008; and Kean et al., 2013). Kean et al. (2013) examined the dynamics of debris flows as initiated by small sediment surges, which appear to result from low-gradient slopes temporarily storing sediment as “sediment capacitors.” This dynamic includes minute slope failures, while still allowing inferences to be drawn about rain intensities needed to trigger different surge magnitudes (Kean et al., 2013). Once a debris flow is triggered, it gains mass from sediment bulking by scour and erosion in source areas and confined piedmont channels. Numerous researchers have attempted to quantify hillslope sediment contributions, finding that as little as 3 percent to as much as 93 percent of total debris comes from hillside erosion (Santi et al., 2008; Staley et al., 2014).

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 3–27

Lancaster, Swanson, Lukashov, Oakley, Lee, Spangler, Hernandez, Olson, DeFrisco, Lindsay, Schwartz, McCrea, Roﬀers, and Tran

Figure 13. Maximum 15 minute rainfall amounts (mm) over a 1 km2 grid representing the cumulative linear kilometers of debris-flow source gullies greater than 1.5 m in width; debris flows in gullies less than 1.5 m wide or those interpreted with questionable confidence, are not represented in this figure. The rain gauges within the burn perimeter annotated on this figure are summarized on Figure 14. Note: A portion within the northern-most burn perimeter in proximity to the county line and the northern-most gauge were mapped by the U.S. Forest Service as unburned based on the soil burn severity map (CAL FIRE, 2018a), and thus post-fire debris-flow activity would not be expected.

Debris-flow source area mapping utilized (Lukashov et al., 2019) an interpretive and digitizing scale between 1:500 and 1:1,250, based on a systematic review of hillshade and aerial photos that were field validated at 184 sites and included a debrisflow confidence classification matrix following Wills et al. (2017). In total, 1,222.3 km of source gullies (ࣙ1.5 m width) were mapped throughout the burn area. Approximately 600 km were mapped in Santa Barbara County and 620 km in Ventura County. Source gullies interpreted as definite were mapped with an accuracy of 98 percent, probable gullies had an accuracy of 69 percent, and questionable gullies had an accuracy of 22 percent. Debris-flow sources were intersected with a 1 km2 grid, weighted by their confidence, and summarized for each grid cell (Figure 13). Peak 15 minute rainfall amounts from the storm were collected from 46 gauges maintained by the Santa Barbara County Public Works Department and Ventura County Watershed Protection District. These amounts, irrespective of time, were plotted and interpolated with the inverse-weighted distance method in ArcGIS (Chen et al., 2017) and are shown in black on

Figure 13. Furthermore, the rainfall depths are plotted against maximum source density data on Figure 14. Corresponding to damaging impacts in Montecito, Figure 13 shows a relatively dense east-west–trending band of debris-flow source gullies in tributary watersheds of the Santa Ynez Mountains, which appears to correlate with high rainfall amounts of 10.2 mm up to 26.16 mm. In contrast, the 5.1 and 7.6 mm contours define what appears to be the threshold of debris-flow occurrence. DEBRIS-FLOW DAMAGES Quantification of debris-flow damages and costs is used for risk assessments, planning for future disasters, and in making decisions about allocating money for pre-disaster mitigation mapping and prevention. However, representing PFDF damages is a challenge because there is no way to quantify the loss of human life, and damages to physical structures are difficult to compile given the variety of entities involved (Taylor and Brabb, 1972; Fleming and Taylor, 1980; and Godt, 1999). Still, damage and cost data are needed to optimize mitigation efforts by comparing the costs of

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 3–27

The 9 January Debris Flow, Santa Barbara County, California

Figure 14. Plot of maximum debris-flow source density against 15 minute rainfall amounts (mm) for the entire burn area using data from Figure 13. Maximum source density was taken as the highest value in length divided by area (km/km2 ) within the grid cell in which the rain gauge was located, or within any of the adjacent eight grid cells (Figure 13). The plot shows the dramatic increase in source gully development at depths greater than 5 mm. The rain gauges within the burn perimeter are labeled as follows: DT = Doulton Tunnel, JD = Jameson Dam, ET = Edison Trail, OM = Old Man Mountain, UMC = Upper Matilija Canyon, LC = La Conchita, WG = Wheeler Gorge, MC = Matilija Canyon, MD = Matilija Dam. Low source density corresponding to high rainfall amounts at JD can be explained due to the precipitation gauge being located in the valley away from steeply sloping source areas. Gauge at LC showed features suggesting that debris-flow generation was absent. The cluster of gauges showing 15 minute depths between 2.5 and 10 mm but showing zero debris-flow activity falls within the Sulphur and Santa Paula Ridge areas, south of Ojai and north of Ventura.

damage to the costs of mitigation (McCoy et al., 2016), as well as in addressing risks in economically restricted areas (Santi et al., 2011). Our compilation of damages and costs followed the general methodology of Fleming and Taylor (1980), including direct, indirect, and undetermined debrisflow damages within Santa Barbara County (damages were unavailable for Ventura County as of January 2020). Direct damages are those necessary to repair and to fully restore all structures and land sustaining physical damage immediately resulting from the debris flows. Indirect damages include secondary losses and results of the debris flows, such as measures taken to mitigate additional damages, and impacts to the local economy from loss of income or taxes. Undetermined damages are those that may be direct or indirect but are inseparable from the

Thomas Fire damages based on available information. As there was no significant time separating the events, many losses are reported as a combination of the two events. Significant structural damage to public and private infrastructure and property occurred within Montecito and in portions of the Carpinteria coastal plain (Figures 6 and 7). Impacts to public infrastructure consisted of seven destroyed and/or damaged bridges along SR 192 and U.S. 101, damages to local water, sewage, gas, and electrical utility infrastructure, and debris deposition in public rights-of-way, debris basins, and channels. Damage to private property was widespread near creek channels, with 558 building structures damaged, more than 162 of which were considered destroyed (CAL FIRE, 2018b). Approximately 2,323 first responders were deployed (SBC, 2018b), which included officers, firefighters, Coast Guard helicopters, armored vehicles, and search and rescue teams with up to 39 search and cadaver dogs, all of whom were involved with 800 rescues, with over 100 rescues done by air (SBC, 2018a, 2018b; Welsh, 2018). The total estimated direct cost of the debris flows alone, as of January 2020, is $982,950,000, with indirect costs of $49,300,000, and possible additional costs of up to $200,000,000 coming from undetermined costs (Table 6). An additional fluctuating impact is a $330,000,000 reduction in assessed property values as a result of the debris flows (RDN, Inc., 2019). About 67 percent of the direct costs comes from private property insurance claims. Damages summarized and plotted over a 1/16 km2 grid (Figure 15) are composed of data from the U.S. Army Corps of Engineers (USACE), Caltrans, Santa Barbara County, the Department of Insurance, and CAL FIRE. USACE debris removal costs directly attributed to 11 debris basins and channels are estimated at over $78,000,000, with additional unattributed county debris removal costs of $24,000,000 (SBCFCD, 2019). A storm clean-up progress map, dated June 2018, was used to delineate the boundaries for debris removal in channels (SBCFCD, 2019). Channels were mapped as polygons approximately bank to bank, for the extents of debris removal, and then intersected by the 1/16 km2 grid. Bridge repair costs are estimated at $55,000,000 from Caltrans District 5 News Releases (Caltrans, 2018a, 2018b) and are broken down according to local reports (Edhat, 2018). Note, $20 million of this is quoted as a repair contract for five bridges, so this cost was divided equally among these bridges (Caltrans, 2018a; Edhat, 2018). At the time of this report, the Cal FIRE Damage Inspection Database (DINS) data set provided the most detailed spatial distribution of damaged private structures (CAL FIRE, 2018b). This data set was combined

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 3–27

Lancaster, Swanson, Lukashov, Oakley, Lee, Spangler, Hernandez, Olson, DeFrisco, Lindsay, Schwartz, McCrea, Roﬀers, and Tran Table 6. Summary of direct, indirect, and undetermined costs related to damage caused by 8 January 2018 debris ﬂows in Santa Barbara County. Damage

Cost (USD 2018)

Direct Damage Costs U.S. 101 debris removal Property insurance claims Debris, basin, and channel removal County emergency protective measures Basin and channel permanent restoration SR 192 bridge repairs County road and bridge repair, 20 miles (32 km) City of Montecito emergency response Water district Sewer district Trail restoration Reduction in assessed property values Indirect Damage Costs Lost wages due to U.S. 101 closure New Randall Road debris basin Installation of 6 debris ring nets Undetermined Damage Costs SC Edison settlement to public entities Disaster assistance loans Emergency watershed protection (EWP) *

$11,250,000 $658,000,000 $102,000,000 $12,600,000 $120,000,000 $55,000,000 $8,500,000 $7,000,000 $5,500,000 $1,600,000 $1,500,000 $330,000,000 $25,000,000 $20,000,000 $4,300,000 $150,000,000 $50,000,000 $824,131*

Data Source Caltrans (2018a) Mills et al. (2018) SBCFCD (2019) SBCFCD(2019) Horwitz (2018) Caltrans (2018b) SBC Board of Supervisors (2018) RDN, Inc. (2018) Montecito Water District, (2019) Herrick, (2019) Herrick, (2019) RDN, Inc. (2019) RDN, Inc. (2018) Farnsworth (2020) KSBY (2019) Yamamura and Smith (2019) U.S. Small Business Administration (2018) NRCS (2019)

This value is for the entire ﬁre area.

Figure 15. Damage intensity map showing the cumulative estimated cost per 1/16 km2 grid cell in area impacted by debris-ﬂow deposition. This map includes all direct damages to private property and public infrastructure, including the cost of debris removal. Costs associated with private property are preliminary and based on an evaluation of CAL FIRE DINS data (CAL FIRE, 2018b) as a proxy for ﬁnal reconstruction costs. The damage data around Carpinteria Salt Marsh Preserve are due to woody debris that issued from Santa Monica and Franklin creeks. Indirect costs, such as economic losses, are not included in this map.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 3–27

The 9 January Debris Flow, Santa Barbara County, California

with the $658,000,000 in insurance claims by the Department of Insurance to provide a normalized average cost per damaged structure (Mills et al., 2018), which was calculated as approximately $2,140,482 per destroyed structure, $1,228,795 per highly damaged structure, $436,024 per moderately damaged structure, and $158,554 per slightly damaged structure. The majority of damage occurred downstream of the burn perimeter in Montecito, Summerland, and Carpentaria, but several structures sustained damage due to debris deposition within the burn perimeter, including buildings on the flanks of Cold Spring Canyon Creek, and the Gobernador Creek Debris Basin. The greatest areas of debris deposition and damage were along Montecito Creek and San Ysidro Creek, with moderate debris deposition and damage along Cold Spring Creek, Hot Springs Creek, Oak Creek, Romero Creek, Toro Canyon Creek, and Arroyo Paredon, the Santa Monica Creek Debris Basin, and the Gobernador Creek Debris Basin. Little deposition occurred in the Franklin Debris Basin and Franklin wash. The geographic intersection of slope with the DINS data (CAL FIRE, 2018b) indicates that 106 of 162 destroyed structures occurred on alluvial fans inclined between 2° and 6° in the general range of surge-front deposition. An additional 26 structures were destroyed on surfaces inclined less than 2°, which shows that surge fronts can be conveyed long distances in confined alluvial channels. SUMMARY AND CONCLUSIONS The large-magnitude PFDF of 9 January 2018 claimed 23 lives, damaged or destroyed 558 structures, and caused severe damage to infrastructure in Montecito and Carpinteria, CA. The combination of large, steep watersheds historically susceptible to debris flows positioned above developed alluvial fans of the Santa Barbara Coastal Plain, widespread moderate soil burn severity, and uncommonly intense rainfall immediately following the Thomas Fire was a recipe for disaster. Debris-flow initiation, severity, and distribution were controlled primarily by the track of extreme precipitation associated with the NCFR, causing inundation over a 3.15 km2 area on the Montecito piedmont and 2.41 km2 area in Summerland and Carpinteria. Peak-flow depths in these areas were up to 10 m and 5 m, respectively. The total estimated volume of soil, ash, rocks, and woody debris exceeded 1,498,000 m3 in the Santa Barbara Coastal Plain, which classifies as a magnitude 6 event, if taken as an aggregate (Jakob, 2005), and debris volumes from individual canyons were locally constrained to 21,555 m3 at west Toro Canyon and 119,271 m3 at Santa Monica Canyon.

Watershed morphology, geology and soils, and alluvial-fan morphology played important roles in controlling the magnitude, timing, distribution, and runout distance of the debris flows. Watersheds with steep rugged slopes (Melton’s number ࣙ0.27) and short planimetric lengths exhibited rapid (5 to 10 minute) response times at Montecito, whereas more gently sloping elongate watersheds (Melton’s number ࣘ0.31) with longer lengths (ࣙ6.0 km) at Carpinteria exhibited 30 to 40 minute response times. The dominance of debris-flow behavior, maximum transported boulder size, and runout distance also diminished eastward, mimicking the transition from more debris flow– to streamflow-dominated fans over the same area. Debris-flow material consisted of a matrix of muddy, ashy sand with gravel derived primarily from rilling of burned hillslopes, and sandstone boulders ranging up to 6 m in diameter derived from alluvial channel deposits. Overbank deposits consisted primarily of boulders and ubiquitous thin patinas of gravelly mud with a durable paste-like consistency. Destruction was most prominent in Montecito due to avulsions and overbank flows along Montecito, San Ysidro, Oak, Buena Vista, and Romero canyons, which locally followed roads for up to 1 km. Flow severity and damages decreased eastward toward Carpinteria, reflecting the transition of flow behavior below the mountain front from debris flows to sedimentladen flooding, coupled with lower population density and fewer drainage crossings. Where present, existing debris basins reduced surge-front depth maxima and associated impacts on downstream urbanized alluvial fans, but most were not engineered for extreme events. Preliminary estimates as of January 2020 indicate that this event resulted in direct and indirect costs exceeding $1 billion dollars, but existing data suggest that the final cost may exceed $1.5 billion dollars when all insurance claims are included. Although the January 9 debris-flow event is the most catastrophic on record, multiple watersheds in the same general area were also impacted by debris flows following wildfires in 1964 and 1971, suggesting the average return period for widespread catastrophic debris flows may be less than 20 years. This event was the result of extreme rainfall on watersheds that were still burning into the winter season, severely stressing the emergency preparedness community and generating a catastrophe on par with the largest magnitude PFDF events documented in California. While in general, the National Weather Service provided adequate forecast lead time, additional recognition of atmospheric conditions that produce extreme high-intensity rainfall is of great importance

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 3–27

Lancaster, Swanson, Lukashov, Oakley, Lee, Spangler, Hernandez, Olson, DeFrisco, Lindsay, Schwartz, McCrea, Roﬀers, and Tran

in a changing climate, as real-time rainfall data cannot provide adequate warning time for debris flows (Kean et al., 2011; Oakley et al., 2017). ACKNOWLEDGMENTS We would like to acknowledge the three anonymous peer reviewers for their constructive comments, and those who contributed information and data to this investigation: Jason Kean, Dennis Staley, Jeffrey Coe, and Francis Rengers (collectively co-leading the phase 1 inundation assessment team from the U.S. Geological Survey); Drew Coe and Pete Cafferata (CAL FIRE); Shawn Johnson, Lael Wageneck, and Jon Frye (Santa Barbara County Public Works); Mark Bandurraga, Bruce Rindahl, and Ron Marotto (Ventura County Watershed Protection District); and Larry Gurrola (engineering geologist). REFERENCES CITED Alluvial Fan Task Force (AFTF), 2010, The Integrated Approach: For Sustainable Development on Alluvial Fans: Water Resources Institute, California State University, San Bernardino, CA, for the California Department of Water Resources, 183 p. Armanini, A. and Gregoretti, C., 2005, Incipient sediment motion at high slopes in uniform flow condition: Water Resource Research, Vol. 41, No. 12, 8 p., https://doi.org/10.1029/ 2005WR004001. Bull, W. B., 1977, The alluvial fan environment: Progress in Physical Geography, Vol. 1, pp. 222–270. Burns, M., 2018, When ‘The Big One’ Isn’t an Earthquake: Edhat Santa Barbara, January 16, 2018: Electronic document, available at https://www.edhat.com/news/when-the-big-oneisn-t-an-earthquake Byrne, R., 1979, Fossil Charcoal from Varved Sediments in the Santa Barbara Channel—An Index of Wildfire Frequencies in the Los Padres National Forest (735–1520 AD): Pacific Southwest Forest and Range Experiment Station Report PSW-47, Berkeley, CA, 70 p. California Department of Forestry and Fire Protection (CAL FIRE), 2018a, Thomas Fire Watershed Emergency Response Team—Final Report: California Department of Forestry and Fire Protection, Sacramento, CA, Report CA-VNC-103156, 172 p. California Department of Forestry and Fire Protection (CAL FIRE), 2018b, Damage Inspection Database (DINS): California Department of Forestry and Fire Protection, Office of the State Fire Marshall, Sacramento, CA, ArcGIS files dated 26 April 2018. California Department of Forestry and Fire Protection (CAL FIRE), 2018c, Historic Fire Perimeter Database: Electronic data, available at https://frap.fire.ca.gov/frapprojects/fire-perimeters/ California Department of Water Resources (CADWR), 1973, California High Water 1971–1972: California Department of Water Resources, Sacramento, CA, pp. 69–72. Caltrans, 2018a, News Release, Monday, October 8, 2018: Electronic document, available at http://www.dot.ca.gov/ dist05/paffairs/santabarbara/100818.pdf

Caltrans, 2018b, Caltrans Presentation of U.S. 101 recovery to SBCAG Board, February 15, 2018: County of Santa Barbara Access Television, CSBTV20. Cannon, S. H.; Boldt, E. M.; Kean, J. W.; Laber, J.; and Staley, D.M., 2010, Relations between Rainfall and Postfire DebrisFlow and Flood Magnitudes for Emergency-Response Planning, San Gabriel Mountains, Southern California: U.S. Geological Survey Open-File Report 2010-1039, 31 p. Cannon, S. H.; Gartner, J. E.; Parrett, C.; and Parise, M., 2003, Wildfire-related debris-flow generation through episodic progressive sediment-bulking processes, western USA. In Rickenmann, D. and Chen, C. (Editors), Debris-Flow Hazards Mitigation: Mechanics, Prediction, and Assessment: Millpress, Rotterdam, Netherlands, pp. 71–82. Chawner, W. D., 1934, The Montrose–La Crescenta Flood of January 1, 1934, and its Sedimentary Aspects: M.S. Thesis, California Institute of Technology, Pasadena, CA, 78 p. Chen, T.; Ren, L.; Yuan, F.; Yang, X.; Jiang, S.; Tang, T.; Liu, Y.; Zhao, C.; and Zhang, L., 2017, Comparison of spatial interpolation schemes for rainfall data and application in hydrological modeling: Water, Vol. 9, 18 p., https://doi.org/10.3390/w9050342. Chow, V. T., 1959, Open-Channel Hydraulics: McGraw-Hill Book Company, New York, NY, 680 p. Coe, J. A.; Godt, J. W.; Parise, M.; and Moscariello, A., 2003, Estimating debris flow probability using fan stratigraphy, historic records, and drainage basin morphology, Interstate 70 highway corridor, central Colorado, U.S.A. In Rickenmann, D. and Chen, C. (Editors), Debris-Flow Hazards Mitigation: Mechanics, Prediction, and Assessment: Millpress, Rotterdam, Netherlands, pp. 1085–1096. Dibblee, T. W., Jr., 1966, Geology of the Central Santa Ynez Mountains, Santa Barbara County, California: California Division of Mines and Geology Bulletin 186. Dibblee, T. W., Jr., 1982, Geology of the Santa Ynez–Topatopa Mountains, southern California. In Fife, D. L. and Minch, J. A. (Editors), Geology and Mineral Wealth of the California Transverse Ranges: South Coast Geological Society, Annual Symposium and Guidebook Number 10, Mason Hill Volume, pp. 40–56. Edhat, 2018, Caltrans Continues to Rebuild/Repair Bridges on State Route 192: Edhat Santa Barbara: Electronic document, available at https://www.edhat.com/news/caltranscontinues-to-rebuildrepair-bridges-on-state-route-192 Farnsworth, B., 2020, Work on Randall Road Debris Basin Expected Summer of 2021: KEYT News: Electronic document, available at https://keyt.com/news/2020/01/07/randallroad-will-transform-into-a-debris-basin/ Fleming, R. W. and Taylor, F. A., 1980, Estimating the Costs of Landslide Damage in the United States: U.S. Geological Survey Circular 832, 21 p. Gabet, E. J., 2003, Post fire thin debris flows: Sediment transport and numerical modeling: Earth Surface Processes Landforms, Vol. 28, pp. 1341–1348. Gartner, J. E.; Cannon, S. H.; and Santi, P. M., 2014, Empirical models for predicting volumes of sediment deposited by debris flows and sediment-laden floods in the Transverse Ranges of southern California: Engineering Geology, Vol. 176, pp. 45–56, https://doi.org/10.1016/j.enggeo.2014.04.008. Godt, J. W., 1999, Maps Showing Locations of Damaging Landslides Caused by El Niño Rainstorms, Winter Season 1997– 98, San Francisco Bay Region, California: Pamphlet to accompany U.S. Geological Survey Miscellaneous Field Studies Maps MF-2325-A-J.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 3–27

The 9 January Debris Flow, Santa Barbara County, California Gurrola, L. and Rogers, J. D., 2020, Geologic hazards due to landslide dams in the Cold Springs and Hot Springs watersheds, County of Santa Barbara, California: Geological Society of America Abstracts with Programs, Vol. 52, No. 4, paper 26-7. Herrick, K. M., 2019, Village Beat: The Year in Review 2018: Montecito Journal, Vol. 24, No. 52: Electronic document, available at https://issuu.com/santabarbarasentinel/ docs/mj_52_24_full Horwitz, H., 2018, Santa Barbara County Public Works Responds to Devastating Mudflow: APWA Reporter, May 2018: Electronic document, available at https://www.apwa.net/ Library/Reporter/201805_ReporterOnline.pdf Hungr, O.; McDougall, S.; Wise, M.; and Cullen, M., 2007, Magnitude–frequency relationships of debris flows and debris avalanches in relation to slope relief: Geomorphology, Vol. 96, No. 3–4, pp. 355–365. Hungr, O.; Morgan, G. C.; and Kellerhals, R., 1984, Quantitative analysis of debris torrent hazards for design of remedial measures: Canadian Geotechnical Journal, Vol. 21, pp. 663– 677, https://doi.org/10.1139/t84-073. Iverson, R. M., 1997, The physics of debris flows: Reviews of Geophysics, Vol. 35, pp. 245–296. Iverson, R. M., 2014, Debris flows: Behavior and hazard assessment: Geology Today, Vol. 30, No. 1, pp. 15–20, https://doi.org/10.1111/gto.12037. Jakob, M., 2005, A size classification for debris flows: Engineering Geology, Vol. 79, pp. 151–161. Jakob, M.; Weatherly, H.; Bale, S.; Perkins, A.; and MacDonald, B., 2017, A multi-faceted debris-flood hazard assessment for Cougar Creek, Alberta, Canada: Hydrology, Vol. 4, No. 7, 33 p. Kean, J. W.; McCoy, S. W.; Tucker, G. E.; Staley, D. M.; and Coe, J. A., 2013, Runoff-generated debris flows: Observations and modeling of surge initiation, magnitude, and frequency: Journal Geophysical Research: Earth Surface, Vol. 118, pp. 2190–2207, https://doi.org/10.1002/jgrf.20148. Kean, J. W. and Staley, D. M., 2019, Estimating the annual probability of post-fire debris flow in southern California: American Geophysical Union, Abstracts with Programs, abstract NH33E-0968. Kean, J. W.; Staley, D. M.; and Cannon, S. H., 2011, In situ measurements of post fire debris flows in southern California: Comparisons of the timing and magnitude of 24 debris-flow events with rainfall and soil moisture conditions: Journal Geophysical Research, Vol. 116, No. 4, 21 p., https://doi.org/10.1029/2011JF002005. Kean, J. W.; Staley, D. M.; Lancaster, J. T.; Rengers, F. K.; Swanson, B. J.; Coe, J. A.; Hernandez, J. L.; Sigman, A. J.; Allstadt, K. E.; and Lindsay, D. N., 2019, Inundation, flow dynamics, and damage in the 9 January 2018 Montecito debris-flow event, California, USA: Opportunities and challenges for post wildfire risk assessment: Geosphere, Vol. 15, No. 4, pp. 1140–1163, https://doi.org/10.1130/GES02048.1. Keller, E.; Adamaitis, C.; Alessio, P.; Anderson, S.; Goto, E.; Gray, S.; Gurrola, L.; and Morell, K., 2020a, Applications in geomorphology: Geomorphology, Vol. 366, pp. 106729, https://doi.org/10.1016/j.geomorph.2019.04.001. Keller, E.; Adamaitis, C.; Alessio, P.; Goto, E.; and Gray, S., 2020b, Montecito debris flows of 9 January 2018: Physical processes and social implications. In Heermance, R.V. and Schwartz, J.J. (Editors), From the Islands to the Mountains: A 2020 View of Geologic Excursions in Southern California: Geological Society of America Field Guide 59, p. 95–114.

KSBY, 2019, Montecito Debris Flow Nets Could Be Installed Next Week: KSBY.com: Electronic document, available at https://ksby.com/news/local-news/2019/04/30/montecitodebris-flow-nets-could-be-installed-next-week Laber, J., 2018, An Overview of the January 9, 2018, Flash Flood and Debris Flow Event in Montecito, CA: Electronic document, available at https://www.alertsystems.org/presentations/ Conf2018/Session1-Recent_Events/Laber_REC.pdf Lamb, M. P.; Dietrich, W. E.; and Venditti, J. G., 2008, Is the critical Shields stress for incipient sediment motion dependent on channel-bed slope?: Journal Geophysical Research, Vol. 113, pp. F02008, https://doi.org/10.1029/2007JF000831. Lancaster, J. T., 2018, The Santa Barbara County 1/9 Debris Flow of 2018 Extreme Runoff Response to Extreme Precipitation: Presentation to the Southwest Extreme Precipitation Symposium, La Jolla, CA, March 28, 2018: Electronic document, available at https://swepsym.org/proceedings-2018-droughtsfloods.php#Lancaster. Lancaster, J. T.; Spittler, T. E.; and Short, W. R., 2015, Alluvial Fan Flooding Hazards: An Engineering Geology Approach to Preliminary Assessment: California Geological Survey Special Report 227, 46 p. Lukashov, S. G.; Lancaster, J. T.; Oakley, N. S.; and Swanson, B. J., 2019, Post fire debris flows of 9 January 2018, Thomas Fire, southern California: Initiation areas, precipitation and impacts. In Kean, J. W.; Coe, J. A.; Santi, P. M.; and Guillen, B. K. (Editors) Debris-Flow Hazards Mitigation: Mechanics, Monitoring, Modeling, and Assessment: Proceedings of the Seventh International Conference on Debris-Flow Hazards Mitigation, June 10–13, 2019: Association of Environmental and Engineering Geologists, Golden, CO, pp. 774–781. McCoy, K.; Krasko, V.; Santi, P.; Kaffine, D.; and Rebennack, S., 2016, Minimizing economic impacts from post fire debris flows in the western United States: Natural Hazards, Vol. 83, pp. 149–176, https://doi.org/10.1007/s11069-016-2306-0. Mensing, S. A.; Michaelson, J.; and Byrne, R., 1999, A 560year record of Santa Ana fires reconstructed from charcoal deposited in the Santa Barbara Basin, California: Quaternary Research, Vol. 51, pp. 295–305. Mills, E.; Lamm, T.; Sukhia, S.; Elkind, E.; and Ezroj, A., 2018, Trial by Fire: Managing Climate Risks Facing Insurers in the Golden State: California Department of Insurance, Sacramento, CA: Electronic document, available at https://www.law.berkeley.edu/wp-content/ uploads/2018/09/Trial-by-Fire-September-2018.pdf Minor, S. A.; Kellogg, K. S.; Stanley, R. G.; Gurrola, L. D.; Keller, E. A.; and Brandt, T. R., 2009, Geologic Map of the Santa Barbara Coastal Plain Area, Santa Barbara County, California: U.S. Geological Survey Scientific Investigations Map 3001, 1 sheet, scale 1:25,000, pamphlet 38 p. Montecito Water District (MWD), 2019, Montecito Water District Update: Montecito Association Board Meeting, February 12, 2019: Electronic document, available at https:// www.montecitowater.com/news/presentation-to-montecitoassociation/ Moody, J. A. and Martin, D. A., 2001, Initial hydrologic and geomorphic response following a wildfire in the Colorado Front Range: Earth Surfaces Processes Landforms, Vol. 26, pp. 1049–1070. National Oceanic and Atmospheric Administration (NOAA), 1971, Climatological Data, Vol. 77, No. 12: NOAA, Boulder, CO. National Research Council (NRC), 1996, Alluvial Fan Flooding: Committee on Alluvial Fan Flooding, Water Science and Technology Board, Commission on Geosciences,

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 3–27

Lancaster, Swanson, Lukashov, Oakley, Lee, Spangler, Hernandez, Olson, DeFrisco, Lindsay, Schwartz, McCrea, Roffers, and Tran Environment, and Resources, National Research Council, National Academy Press, Washington, D.C., 172 p. Natural Resources Conservation Service (NRCS), 2019, personal communication. Oakley, N. S.; Cannon, F.; Munroe, R.; Lancaster, J. T.; Gomberg, D.; and Ralph, F. M., 2018, Brief Communication: Meteorological and climatological conditions associated with the 9 January 2018 post fire debris flows in Montecito and Carpinteria, California, USA: Natural Hazards, Vol. 18, pp. 3037–3043, https://doi.org/10.5194/nhess-18-30372018. Oakley, N. S.; Lancaster, J. T.; Kaplan, M. L.; and Ralph, F. M., 2017, Synoptic conditions associated with cool season post fire debris flows in the Transverse Ranges of southern California: Natural Hazards, Vol. 88, pp. 327–354, https://doi.org/10.1007/s11069-017-2867-6. Perica, S.; Dietz, S.; Heim, S.; Hiner, L.; Maitaria, K.; Martin, D.; Pavlovis, S.; Roy, I.; Trypaluk, C.; Unruh, D.; Yan, F.; Yekta, M.; Zhao, T.; Bonnin, G.; Brewer, D.; Chen, L.; Parzybok, T.; and Yarchoan, J., 2011, PrecipitationFrequency Atlas of the United States, California: National Oceanic and Atmospheric Administration Atlas 14, Vol. 6, Version 2.0. Pierson, T. D. and Costa, J. E., 1987, A rheologic classification of subaerial sediment-water flows. In Costa, J. E. and Wieczorek, G. F. (Editors), Debris Flows/Avalanches: Process, Recognition, and Mitigation: Geological Society of America, Reviews Engineering Geology, Vol. 7, pp. 1–12. Prochaska, A. B.; Santi, P. M.; Higgins, J. D.; and Cannon, S. H., 2008, A study of methods to estimate debris flow velocity: Landslides, Vol. 5, pp. 431–444, https://doi.org/10.1007/s10346-008-0137-0. Robert D. Niehaus, Inc. (RDN, Inc.), 2018, The Economic Impacts of the Montecito Mudslides: A Preliminary Assessment: Electronic document, available at http://www. rdniehaus.com/rdn/wp-content/uploads/2018/03/RDN_ Montecito_Mudslides_Impacts-1.pdf Robert D. Niehaus, Inc. (RDN, Inc.), 2019, The Montecito Debris Flow One Year Later: Housing Market Impact and Recover: Electronic document, available at http://www. rdniehaus.com/rdn/wp-content/uploads/2019/03/RDN_ Montecito-Debris-Flow_Housing-Market-Impact_03-19.pdf Santa Barbara County (SBC) Board of Supervisors, 2018, Agenda Letter from Public Works Director, Scott D. McGolpin, November 6, 2018: Electronic document, available at https://santabarbara.legistar.com/ (accessed July 2019). Santa Barbara County (SBC), 2018a, Santa Barbara County 1/9 Debris Flow Disaster Response and Recovery: Central Coast Climate Change Collaborative, April 24, 2018: Electronic document, available at http://www.centralcoastclimate.org/wpcontent/uploads/2018/05/5.-Matt-Pontes.pdf Santa Barbara County (SBC), 2018b, Thomas Fire and 1/9 Debris Flow After-Action Report and Improvement Plan: Electronic document, available at https://www.countyofsb. org/asset.c/4550 Santa Barbara County Flood Control District (SBCFCD), 1969, 1969 Floods: Electronic document, available at https://www.countyofsb.org/pwd/waterdownloads.sbc Santa Barbara County Flood Control District (SBCFCD), 2018, personal communication, 130 East Victoria Street Suite 200, Santa Barbara, CA 93101. Santa Barbara County Flood Control District (SBCFCD), 2019, personal communication, 130 East Victoria Street Suite 200, Santa Barbara, CA 93101.

Santa Barbara County Flood Control District (SBCFCD), 2020, Santa Barbara—Annual Rainfall (Stn #234) 1900–2019: Electronic document, available at https://www.countyofsb. org/pwd/water/downloads/hydro/graphs/234graph.pdf Santi, P. M.; deWolfe, V. G.; Higgins, J. D.; Cannon, S. H.; and Gartner, J. E., 2008, Sources of debris flow material in burned areas: Geomorphology, Vol. 96, pp. 310–321, https://doi.org/10.1016/j.geomorph.2007.02.022. Santi, P. M.; Hewitt, K.; VanDine, D. F.; and Barillas Cruz, E., 2011, Debris-flow impact, vulnerability and response: Natural Hazards, Vol. 56, pp. 371–402. Scheidl, C. and Rickenmann, D., 2010, Empirical prediction of debris-flow mobility and deposition on fans: Earth Surface Processes Landforms, Vol. 35, pp. 157–173. Schiff, D. and D’Agostino, R. B., 1996, Practical Engineering Statistics: John Wiley & Sons, New York, NY, 309 p. Schwartz, J., 2017, Assessment of the post fire debris flow and flooding events of January 20, 2017 in the Sherpa Fire burned area. In Mikulovsky, R. (Editor), Diggin Deep, No. 313: Minerals & Geology, U.S. Forest Service, Washington, pp. 9–11. Shuirman, G. and Slosson, J. E., 1992, The fire-flood sequence; a deadly combination. In Forensic Engineering, Environmental Case Histories for Civil Engineers and Geologists: Academic Press, San Diego, CA, pp. 186–206. Staley, D. M., 2018, personal communication. Staley, D. M.; Kean, J. W.; Cannon, S. H.; Schmidt, K. M.; and Laber, J. L., 2013, Objective definition of rainfall intensity– duration thresholds for the initiation of post fire debris flows in southern California: Landslides, Vol. 10, pp. 547–562, https://doi.org/10.1007/s10346-012-0341-9. Staley, D. M.; Wasklewicz, T. A.; and Kean, J. W., 2014, Characterizing the primary material sources and dominant erosional processes for post-fire debris-flow initiation in a headwater basin using multi-temporal terrestrial laser scanning data: Geomorphology, Vol. 214, pp. 324–338. Taylor, F. A. and Brabb, E. E., 1972, Maps Showing Distribution and Cost by Counties of Structurally Damaging Landslides in the San Francisco Bay Region, California, Winter of 1968– 69: U.S. Geological Survey Miscellaneous Field Studies Map MF-327. U.S. Army Corps of Engineers Los Angeles District (USACE), 1974, Flood Plain Information, Montecito Streams, Vicinity of Montecito, Santa Barbara, County, California: USACE, Department of the Army, Washington, D.C., 37 p. U.S. Army Corps of Engineers Los Angeles District (USACE), 1999, Santa Barbara County Streams, Lower Mission Creek Flood Control Feasibility Study, Draft Main Report & EIS/EIR: USACE, Department of the Army, Washington, D.C., 39 p. U.S. Army Corps of Engineers Los Angeles District (USACE), 2018, Debris Basins, Santa Barbara County, CA January 16, 2018: Electronic document, available at https://www.spl. usace.army.mil/Missions/Emergency-Management/SantaBarbara-County-Debris-Removal/ U.S. Geological Survey (USGS), 2005, Southern California— Wildfires and Debris Flows: U.S. Geological Survey Fact Sheet 2005-3106. U.S. Geological Survey, 2019, Quaternary Fault and Fold Database for the United States: Electronic data, available at https://earthquake.usgs.gov/hazards/qfaults/ U.S. Small Business Administration (SBA), 2018, SBA Tops $50 Million in Disaster Assistance Loans: Disaster Press Release CA 15438-05: Electronic document, available at https://www. sba.gov/offices/disaster/dfocw/resources/1624696

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 3–27

The 9 January Debris Flow, Santa Barbara County, California Wagner, D. L.; Lancaster, J. T.; and DeRose, M., 2012, The Oak Creek Debris and Hyperconcentrated Flows of July 12, 2008, Inyo County, California: A Geologic Investigation: California Geological Survey Special Report 225, 63 p. Wells, W. G., 1987, The effects of fire on the generation of debris flows in southern California: Reviews in Engineering Geology, Vol. 7, pp. 105–114., https://doi.org/10.1130/REG7-p105. Welsh, A. and Davies, T., 2011, Identification of alluvial fans susceptible to debris-flow hazards: Landslides, Vol. 8, pp. 183– 194, https://doi.org/10.1007/s10346-010-0238-4. Welsh, N., 2018, Search dogs comb hills of Montecito for survivors and remains: Santa Barbara Independent, 14 January 2018: Electronic document, available at https://www.independent. com/2018/01/14/search-dogs-comb-hills-montecitosurvivors-and-remains/ Western Regional Climate Center (WRCC), 2017, Cooperative Climatological Data Summaries: Electronic data, available at https://wrcc.dri.edu/cgi-bin/cliMAIN.pl?ca0619 Wilford, D. J.; Sakal, M. E.; Innes, J. L.; Sidle, R. C.; and Bergerud, W. A, 2004, Recognition of debris flow,

debris-flood and flood hazard through watershed morphometrics: Landslides, Vol. 1, pp. 61–66. Williams, A. P.; Abatzoglou, J. T.; Gershunov, A.; GuzmanMorales, J.; Bishop, D. A.; Balch, J. K.; and Lettenmaier, D. P., 2019, Observed impacts of anthropogenic climate change on wildfire in California: Earth’s Future, Vol. 7, pp. 892–910, https://doi.org/10.1029/ 2019EF001210. Wills, C. J.; Roth, N. E.; McCrink, T. P.; and Short, W. R., 2017, The California landslide inventory database. In DeGraff, J. and Shakoor, A. (Editors), Proceedings of the Third North American Symposium on Landslides: Association of Environmental and Engineering Geologists, Roanoke, VA, pp. 666– 674. Yamamura, J. and Smith, D., 2019, Santa Barbara City and County part of $360 million Edison settlement: Santa Barbara Independent, 13 November 2019: Electronic document, available at https://www.independent.com/2019/11/13/ santa-barbara-city-and-county-part-of-360-million-edisonsettlement/

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 3–27

Alluvial Fan Alteration Due to Debris-Flow Deposition, Incision, and Channel Migration at Forest Falls, California KERRY CATO* Department of Geological Sciences, California State University, San Bernardino, 5500 University Parkway, San Bernardino, CA 92407

BRETT GOFORTH Department of Geography, California State University, San Bernardino, 5500 University Parkway, San Bernardino, CA 92407

Key Terms: Debris Flows, Hyperconcentrated Flows, Debris-Flow System ABSTRACT Historical patterns of debris flows have been reconstructed at the town of Forest Falls in the San Bernardino Mountains using a variety of field methods (mapping flow events after occurrence, dendrochronology evidence, soil chronosequences). Large flow events occur when summer thunderstorms produce brief high-intensity rainfall to mobilize debris; however, the geomorphic system exhibits properties of non-linear response rather than being a single-event precipitationdriven process. Previous studies contrasted the relative water content of flows generated by varying-intensity summer thunderstorms to model factors controlling flow velocity and pathway of deposition. We hypothesize that sediment discharge in this geomorphic system exhibits multiple sources of complexity and present evidence of (1) thresholds of sediment delivery from sources at the higher reaches of bedrock canyons, (2) storage effects in sediment transport down the bedrock canyons, and (3) feedbacks in deposition, remobilization, and transport of sediment across the alluvial fan in dynamic channel filling, cutting, and avulsion processes. An example of the first component occurred in March 2017, when snowmelt generated a rapid translational landslide and debris avalanche of about 80,000 m3 ; this sediment was deposited in the bedrock canyon but moved no farther down gradient. The second component was observed when accumulation of meta-stable sediments in the bedrock canyon remained in place until fluvial erosion and subsequent debris flow provided dynamic instability to remobilize the mass downstream. The third component occurred on the alluvial fan below the bedrock

*Corresponding author email: kerry.cato@csusb.edu

canyon, where low-water-content debris flows deposited sediments that filled the active channel, raising the channel grade level to levee elevation, allowing for subsequent spread of non-channelized flows onto the fan surface and scouring new channel pathways down fan. A conceptual model of spatial and temporal complexities in this debris-flow system is proposed to guide future study for improved risk prediction. INTRODUCTION Debris-flow risk prediction entails considerable spatial and temporal uncertainties. For example, hydrologic responses for debris-flow generation can be highly dependent on variable slope conditions of antecedent moisture and pore pressure relative to rainfall intensities (Johnson and Sitar, 1990). Such geomorphic systems that exhibit non-linear dynamics are said to be complex, and multiple sources of complexity have been revealed by long-term studies of geomorphic processes (Phillips, 2003; Murray and Fonstad, 2007; and Temme et al., 2015). This article re-examines the functioning of a previously studied debris-flow system in the Transverse Range of southern California by identifying multiple sources of complexity. The damaging effects of Snow Creek canyon debris flows on the Forest Falls community are well documented in relation to high-intensity summer rainstorms on Yucaipa Ridge (Morton and Hauser 2001; Morton et al., 2008). While the role of “summer monsoon” and “cloud burst” rains identified by these previous studies is significant for debris-flow generation, recent events of winter season mass wasting raised new questions about antecedent sources of complexity in this geomorphic system. We report new observations to identify sources of complexity in (1) thresholds of sediment delivery from sources at the higher reaches of bedrock canyons, (2) storage effects in sediment transport down the bedrock canyons, and

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 29–41

Cato and Goforth

(3) feedbacks in deposition, remobilization, and transport of sediment across the alluvial fan in dynamic channel filling, cutting, and avulsion processes. A conceptual model of spatial and temporal complexities in this debris-flow system is proposed to guide future study for improved risk prediction. PREVIOUS STUDIES This article uses creek names shown on the U.S. Geological Survey (USGS, 1994) 7.5 Forest Falls quadrangle (after USGS, 2018a). This use is consistent with current U.S. Forest Service and local emergency response usage. The names “Snow Creek Canyon” (SC) and “Rattlesnake Canyon” (RC) have been used interchangeably in some publications. Previous research concluded that, cumulatively, debris-flow events in SC and RC (next channel east of SC) occur on average every 3.5 years with some years having two episodes. This average was based on documented events from 1951 to 2005 that included 15 large debris flows as well as tree-ring dates for abrasion scars and tree age dates of surfaces over the past 300 years (Morton and Hauser, 2001; Morton et al., 2008). These studies attributed the sediment source of rocky debris-filled canyons with numerous debris chutes that contained fine-grained disintegrated rock. The California Geological Survey (CGS) and San Bernardino County have mapped the slopes surrounding Forest Falls, including SC, as a landslide hazard zone (Tan, 1990; San Bernardino County, 2010). SITE DESCRIPTION The Transverse Range is a west-to-east–oriented mountainous physiographic region of California associated with the transform boundary of the Pacific and North American tectonic plates (Harden, 1998). The San Andreas Fault zone makes the “Big Bend” step-over in southern California, resulting in regional transpressional tectonic stress that has uplifted the San Bernardino Mountains (Yule and Sieh, 2003). The study site at SC (lat. 34.0669°N, long. 116.9103°W) is located on the north slope of Yucaipa Ridge in the San Bernardino Mountains on lands of the San Bernardino National Forest (place-names after USGS, 2018a). The bedrock in this area of Yucaipa Ridge is mapped as Mesozoic quartz monzonite and Precambrian gneiss (Dibblee, 1964; Gutierrez, 2010). The study site has experienced extensive tectonic activity associated with the San Andreas Fault zone, but other faults at the site are also considered to be Holocene active. The Mission Creek Fault (North Branch of the San Andreas Fault) crosses SC (USGS, 1994; Figure 1b). To the north, downslope along the

base of a bajada, is the Mill Creek Fault. The main South Branch of the San Andreas Fault, located approximately 3.5 km south of SC, trends along the south side of the Yucaipa Ridge. According to the Alquist-Priolo Earthquake Fault zone maps (CGS, 1974), all three of the above-mentioned faults are classified as potentially active during the Holocene period. The tectonism produced not only high-relief uplifted mountains (and hence the source of the debris) but also surface faulting and earthquake activity that have also produced highly fractured and locally deeply weathered rock prone to mass wasting. The CGS and County of San Bernardino classified the northern and southern flanks of Mill Creek Valley as a landslide hazard zone (Tan, 1990; San Bernardino County, 2010). On the north side of Yucaipa Ridge at 2,657 m, SC originates and flows into Mill Creek at 1,663-m elevation, over a horizontal distance of 2.6 km. The bedrock portion of SC is a steeply inclined “V-notched” channel incised into bedrock, with this channel section extending down to elevation 1,797 m from the top of Yucaipa Ridge. Below this elevation, SC flows at a gentler gradient across its alluvial fan, which is one of several drainages that form a bajada along the base of Yucaipa Ridge. The bajada trends downhill in a northward direction, through the community of Forest Falls, to the grade level of Mill Creek Valley (Figure 1a and b). The only ingress to the town is along Valley of the Falls Drive, a two-lane paved, county-maintained road built on the bajada and intersected by numerous active flow channels that drain off the northern side of Yucaipa Ridge. The active channel of SC is presently located on the west-central side of the fan. The age of fan surface deposits generally increases away from this active channel in the central portion of the fan, as evidenced by chronosequence studies of soil development and tree-ring age dates of surface deposits (Turk et al., 2008). The far western edge of the fan is bounded by an inactive debris-flow channel that is cut off by the active channel at the fan apex and has well-developed soils on the fan surface, indicating lack of recent debris-flow activity. Forest Falls has a Mediterranean-type climate with warm, dry summers and cold, wet winters. Yucaipa Ridge results in significant orographic enhancement of precipitation. Winter season precipitation occurs during the passage of mid-latitude cyclones, and summer monsoonal rainfall occasionally occurs in thunderstorms with brief intense showers. Flood-intensity rain events are infrequent but typically occur during summer and fall in outbreaks of monsoonal thunderstorm activity, especially when the region is impacted by remnants of a dissipating tropical cyclone, as well as in winter, when a Pacific Atmospheric River circulation pattern develops.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 29–41

Complexities in a Debris-Flow System

Figure 1. (a) Map showing Forest Falls location between two strands of the San Andreas Fault in the San Bernardino Mountains of southern California (modiﬁed from Morton et al., 2008). (b) Aerial photo showing Snow Canyon (SC), the community of Forest Falls, and nearby features. (Source: Modiﬁed from USGS, 1994).

A 12-year record of precipitation (2006–2018) provided by the San Bernardino County Flood Control District (SBCFCD) by a remotely automated tipping bucket rain gauge atop Yucaipa Ridge at 2,753m elevation registered a mean annual precipitation of 85.8 ± 24.9 cm (standard deviation) for the October– September hydrologic year. Approximately 84 percent of precipitation occurs from October through June. On average, the wettest month of the year is December (21.7 cm), and the driest month is June (0.4 cm). METHODOLOGY Site data were obtained from multiple sources, and ﬁeld surveys were repeated from 2017 to 2020. Meteo-

rological data were obtained from a continuously operating station of the SBCFCD with a tipping bucket rain gauge located on Yucaipa Ridge (SBCFCD, 2017). This station is located on Yucaipa Ridge approximately 0.3 km from the headwall of SC. A geographic information system (GIS) was used to model dimensions and topographic relief of the catchment source area for debris-ﬂow generation (Figure 4a and b). The catchment area of SC Canyon was analyzed from a 10-m-cell-resolution digital elevation model (DEM) downloaded from the National 3D Elevation Program (USGS, 2018b) and processed with ArcMap 10.3 software. The Surface tool in ArcMap Spatial Analyst derived slope angles of each grid cell. The Watershed tool was used to delineate the planimetric area of

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 29–41

Cato and Goforth

the catchment upstream from a fan apex pour point. The Stream order tool was used to determine the rank of the trunk stream channel at the fan apex using a 1,000-grid-cell flow accumulation rule. The Surface Volume tool in ArcMap 3D Analyst was used to compute a 2D map area encompassed by the catchment and a 3D surface area of the canyon slopes. The greatest density of field survey observations occurred on the alluvial fan surface. We monitored which channels were active (i.e., showed geomorphic change) by noting bar deposition, channel bank erosion, vertical surface alluviation, levee formation, channel abandonment, and areas of channel avulsion and deposition onto older fan surfaces. We attempted to visit the site after each major winter storm occurred, after each summer thunderstorm, and if a mass-wasting event was reported (such as the 2017 landslide). Topographic and geomorphic information described above was documented in two forms: open traverse topographic surveys and using LiDAR scans of the active alluvial fan channel areas. To establish the ephemeral location of the alluvial fan channel, open traverse surveys of topography on the fan over the active channel and extent of debris flow were recorded. This included, noting the formation of debris-flow levees, alluviation and increases in thalweg elevations and providing Global Positioning System (GPS) control for both small unmanned aerial vehicle (sUAV [drone]) and LiDAR surveys. Submeterprecision GPS roving receivers (Trimble GEO7x with Zephyr antenna, ArcPad 10 software) were used to perform open traverse surveys of topography on the fan over the active channel, and the extent of debris flow was recorded. The GPS positional data (x, y, z) were recorded only if a real-time differential correction was obtained, and the 3D position dilution of precision was <3.0. Elevation (z) was calibrated using the GEOID03 model offset of −30.414 m reported by the National Geodetic Survey (NGS) at the latitude and longitude coordinates of Forest Falls CA relative to height above the WGS84 ellipsoid (NGS, 2017–2018). A total of 3,022 GPS survey points were recorded across the fan on September 23, 2017, and the survey was repeated on September 28, 2018 (2,362 GPS survey points). Only GPS survey points with estimated positional error ࣘ1 m were used for geospatial analysis of the fan surface elevation change. This allowed for use of 2,002 points (66.2 percent) from the 2017 survey and 2,113 points (89.4 percent) from the 2018 survey. These Global Navigation Satellite System (GNSS) data for both years were exported to ArcMap and converted into 10-m-cell grids using the Point to Raster tool, which averaged all elevations within each cell. The Spatial Analyst Raster Calculator tool was used to compute the difference in elevation ( Z) for 2018

minus 2017. Cells with values of −1 m or more indicate areas where net erosion exceeded the range of GPS measurement imprecision (Figure 4b). Likewise, positive values greater than 1 m indicate where net deposition measurably occurred. In January 2020, repeat cross sections across the alluvial fan were surveyed using the GPS methods previously described. Each surveyed point of elevation in a cross section was averaged for 25 GNSS positions. Low-altitude digital photographic remote sensing (30–100 m above ground level) was conducted by sUAV. The sUAV surveys used DJI Mavic Pro for repeat aerial photographs to monitor stream dynamics in comparison to the GPS ground surveys. Digital aerial photos obtained with the sUAV were processed using the Metashape point cloud software, which performs photogrammetric processing of digital images and generates 3D spatial data. This method worked reasonably well for this site except in areas of dense forest canopy on the alluvial fan, where the ground surface is diﬃcult to observe. The sUAV was extremely useful in the upper drainage and bedrock channel area, where the extreme topographic relief makes ground-based access diﬃcult. Volume estimates of the February 2017 mass movement in the upper slopes of SC utilized drone-based aerial photos processed with Metashape. The middle- and lower-alluvial-fan areas were scanned using terrestrial and handheld LiDAR to create the imagery shown in Figure 8. Small-scale scans were obtained from OpenTopography.org (National Center for Airborne Laser Mapping, 2014) to cover the entire channel length at approximately 1-m resolution for baseline coverage. A tripod-based terrestrial LiDAR scanner (Teledyne Polaris LR laser scanner) and handheld laser scanner (Geoslam Horizon) were used for large-scale scans to record greater detail of alluvial fan channel features. RESULTS Our GIS model indicates that SC has local relief of 862 m from the fan apex (1,795 m) to the crest of the highest headwall at Yucaipa Ridge (2,657 m) over a total length of 1.6 km horizontal distance (27.8° gradient). Slope angles of DEM grid cells range from 2.7° to 69.5°, with mean of 39.8°. Although the catchment encompasses 67.2 ha of planimetric (horizontal) area upstream from the fan apex, the 3D surface model provides an estimate that there is 93.6 ha of total land surface area available for erosion by including coverage of the steeply sloped surfaces within this canyon. SC has a nearly linear drainage pattern, and the trunk channel ranks as a second-order stream at the fan apex below the base of the canyon. From the fan apex, the length

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 29–41

Complexities in a Debris-Flow System

Figure 2. (a) Mass-wasting area in Snow Creek after ﬁrst episode (on February 17, 2017) (area of movement outlined in yellow). (b) Second and larger movement event in Snow Creek (about 2 months later). Trees on slide mass are approximately 25 m high. (Photo source: Tom McIntosh, personal communication, 2017).

of the active channel is approximately 530 m, with an average alluvial fan gradient of 9.9°. This channel has a sharp curve in the stream course from 317 to 342 m below the fan apex, and the outer cut-bank section of the curve is the location of recurrent levee breaches during recent ﬂows. On February 17, 2017, the ﬁrst of two mass-wasting events occurred in the upper reaches of SC (Figure 2a). Less than 2 months later, a second mass movement was fortuitously recorded by video camera at the same upper basin, bedrock sidewall of SC (Figure 2b). These events occurred while snowpack was melting on moderately steep slopes with 45°–50° inclinations. Subsequent inspection of the landslide scar showed that the material appeared to break free as a translational landslide movement of intact bedrock with a colluvial veneer. Some detrital material was deposited in an avalanche chute at the base of the slope, which at the surface consisted of 1–3-m-diameter angular boulders (Figure 3a and b). The video showed a rapid runout of sediment by debris avalanche that extended

850 m farther down the chute into a narrow bedrock canyon. We estimate from this video that the larger clastic material in the debris avalanche reached a velocity on the order of 100–140 km/hr and heights of up to 50 m above the canyon bottom. Video documentation of this process alerted us to the presence of significant sediment storage within the bedrock channel. The following three seasons (summer 2017, fall 2017, and winter 2018) were abnormally dry, and repeat observations of the bedrock slope detected small amounts of sediment ravel. Most of the debris deposited in the canyon from these two mass-movement events have remained stationary and have not been remobilized. About a year and a half later, on August 16, 2018, an intense thunderstorm over Yucaipa Ridge resulted in 5 cm of rain over a 2-hour period; this was the most significant hydrologic loading to occur after the masswasting event in spring 2017. The runoff from this storm produced a hyper-concentrated mudflow that forced closure of Valley of the Falls Road approximately 0.2 km down the channel by a layer of approximately 0.1–0.25 m of sediment deposited on the road. Post-flow inspection of the alluvial fan upstream of the road found that sediment filled the channel and formed new levee structures associated with channel avulsion of SC, with flow shifted west onto the fan at the location of a curve in the channel. While most of the sediment was transported into the Mill Creek trunk drainage below the fan, a substantial volume of sediment was deposited as fluvial bars along the SC channel and adjoining fan surfaces (after Wells and Harvey, 1987). These observations indicate that vertical aggradation of the fan occurred in areas upstream of Valley of the Falls Drive. The sediment source for the flow included the erosion scar area of the 2017 masswasting event and finer-texture (medium-sized sand to fine gravel) material that accumulated 600–700 m downstream after being flushed from coarser debris interstices. DISCUSSION Three landscape development components are observed to operate in the SC debris-flow system, and each appears to have its own controls, but inputs and throughputs may link the components or only influence that component. The components include (1) thresholds of sediment delivery from sources at the higher reaches of bedrock canyons, (2) storage effects in sediment transport down the bedrock canyons, and (3) feedbacks in deposition, remobilization, and transport of sediment across the alluvial fan in dynamic channel filling, cutting, and avulsion processes.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 29–41

Cato and Goforth

Figure 3. (a) Source area of mass wasting taken in May 2017 about 2 months after last movement occurred. Rock mass structure is evident, as is a moderate to high degree of weathering. Slope inclination is approximately 50°. (b) Trees and large boulders were mobilized by the winter–spring 2017 mass-wasting event. The landslide source area is several hundred meters up channel from this location. Angular boulder diameters range from 1.0 to 3.0 m in diameter.

Mass-Wasting Sediment Production from the Bedrock Canyon The winter–spring 2017 mass-wasting event in SC occurred on the west bedrock canyon wall at an elevation of 2,347 m. A variety of conditions likely contributed to the mass-wasting event, including triggering meteorological conditions, such as warm temperatures that produced rapid snowmelt with abundant antecedent moisture from seasonally high precipitation rates, and contributing geological conditions, such as rock weathering, steep terrain–inducing movement, tectonic activity, and associated rock mass weakening. These Mesozoic quartz monzonite and Precambrian gneiss rocks, as mapped by both Dibblee (1964) and Gutierrez et al. (2010), were observed to be highly fractured and weathered. The area of movement is a mapped landslide complex with slope inclination in the failure area at 50°–55°. A 1–2-m-thick colluvial layer overlies a highly fractured bedrock mass (Figure 3a and b). The mass movement occurred on a plane that

is oriented subparallel to the valley wall. At the base of the slope, the block size of the angular boulders ranges from 1.5 to 3.5 m. By contrast, at the toe of the debris runout, located 600–700 m downslope, the block diameter of the angular boulders exposed at the surface was 0.6 to 1.5 m. While material size may have been sorted with distance from the slide area, the rapid and violent nature of this event appears to have produced a signiﬁcant breakdown of particle sizes. This was most likely enhanced by the rock mass’s in situ fracturing and weathering prior to mass wasting. Farther downstream on the alluvial fan surface, clast diameters typically range from 10 cm to 0.5 m along with a considerable sand and silt–sized fraction. Tectonic strain, frequent seismic shaking, and topographic ampliﬁcation of earthquakes can trigger rock slope failures (after Keefer, 1984; see review by Murphy, 2015). There are three Holocene active faults located within 3.5 km of the site of the 2017 mass movement, with the closest being the Mission Creek Fault (North Branch of the San Andreas fault) at about

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 29–41

Complexities in a Debris-Flow System

Figure 4. LiDAR imagery of Snow Creek and alluvial fan. (a) Entire system downstream of mass-movement area with features annotated. Area A is the active channel area where sUAV point cloud imagery was acquired that shows the channel changes produced by the August 16, 2018, hyper-concentrated ﬂow and subsequent ﬂows through early 2020. (b) Enlarged area from (a) that shows close-up of alluvial fan channels.

0.50 km away and crossing the bedrock canyon (USGS, 2017). From February 2017 to December 2017, there were 24 earthquakes greater than M2.5 within a 20-km radius of SC (USGS, 2017). According to the USGS (2017) Earthquake Hazards program, a M3.4 earthquake occurred on February 10, 2017, less than 1 week prior to the mass-movement event, with an epicenter approximately 8 km southwest of the site. Data obtained from the Yucaipa Ridge rainfall gauge show that Forest Falls received 81 cm of precipitation from October 23, 2016 to February 18, 2017 (SBCFCD, 2017). In the weeks prior to the masswasting event, there were no recorded precipitation events; however, the warm temperatures during this time melted much of the snowpack, and this additional hydrologic loading likely facilitated movement, as has been reported elsewhere in the Transverse Range (e.g., Morton and Campbell, 1974). Note the diﬀerence in snow coverage due to snowmelt over a 2-month period that is shown in photographs in Figure 2a and b. Based on our observations and conclusions by O’Keefe (2017), the 2017 mass-wasting event appears to combine aspects of ﬂow and slide movements from the position of slope failure, including a debris avalanche (Cruden and Varnes, 1996). Prior documented large mass-wasting episodes in Forest Falls

were classified as debris flows and debris avalanches triggered by summer thunderstorms (Morton and Hauser, 2001; Morton et al., 2008). Although debrisflow processes have likely operated in this location through the Holocene and late Pleistocene, detailed stratigraphic analysis is lacking. Sediment Storage and Transport Down the Bedrock Canyons A substantial volume of sediment can accumulate in bedrock canyons from recurrent mass wasting and short-travel-distance, small-volume flows before the materials become remobilized by larger debris-flow events within the bedrock channel that travel as far as the downstream fan surface. Generally, more attention is given to the upslope mass movements that produce sediment and the events of longer-distance flow that transport sediment downslope onto the alluvial fan. However, much of the sediment produced in the winter–spring 2017 mass-wasting event became stored in the channel of the bedrock canyon, yet to be remobilized by future debris flow. Based on processed dronebased aerial photography, we estimate that approximately 120,000 m3 of sediment mobilized in the 2017 event were initially stored at the base of the landslide

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 29–41

Cato and Goforth

Figure 5. Photos of the upper drainage basin of Snow Creek; all views are looking south toward the crest of Yucaipa Ridge. (a) Field photo taken on January 8, 2020, with snow present in the east and west avalanche chutes. Top of sediment body in the channel that is exposed above the snow is denoted by curved line with arrows in east avalanche chute. Source area of 2017 debris avalanche noted in the west avalanche chute. (b) Vertical aerial view of same area as in (a) but with no snow; double-headed arrow is same length in both photos (a and b). (c) Close-up of sedimentary bar in east avalanche chute with approximate dimensions of 500 m long and 8–28 m wide.

for up to 850 m downstream within the bedrock channel. The August 2018 hyper-concentrated ﬂow reached Valley of the Falls Drive. Field assessments of sediment depths and maps of the ﬂow indicate that about 15,000 m3 of earth material were deposited on and downstream of Valley of the Falls Drive, combined with an estimated 10,000 m3 on the alluvial fan surface. Acknowledging uncertainties in such estimates, it would suggest more than three-quarters (95,000 m3 )

of sediment from the 2017 mass-wasting event remain stored in the canyon’s bedrock channel. Sediment Deposition and Transport of Sediment on and across the Debris-Flow Fan Several changes were observed to have occurred from 2017 to 2020 in the area of the main debrisﬂow channel, including aggradation, levee formation,

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 29–41

Complexities in a Debris-Flow System

Downstream of this curve, the channel became deactivated in 2017 such that sediment is no longer transported through this abandoned distributary channel onto Valley of the Falls Drive. This change activated erosional processes of incision and headcutting into fan deposits, forming new channels for ﬂow that connect to the outlet of the bedrock channel located approximately 800 m upstream of Valley of the Falls Drive. Figures 6–8 show photographs and LiDAR imagery of this area.

Discussion of System

Figure 6. January 2020 views of processes observed in the Area A–active alluvial channel (see photo index map and Figure 4). (a) Transition zone within 200 m of the channel curve where the channel has incised and active bank cutting occurs (area of trees with double-headed arrow) and farther downstream where the channel opens up with less incision and levees have formed on both sides of the channel. (b) sUAV photo of wide channel area just upstream of entry to abandoned channel. For scale the person in the circle is approximately 2 m tall. Active SC trunk channel now migrated to the west. This abandonment and new channel formation occurred within a 2-year period (2018–2020).

channel abandonment, incision and headcutting, and formation of new channels as the overall active channel location shifts westward. The changes were most pronounced at the curve in the stream course that is located from 317 to 342 m below the fan apex. Here, a raised levee of deposited debris (Figure 7) occurred on both sides of the channel at transects 1, 2, 5, and 6 (Figure 9). The channel filled with deposits up to the top of levee at transects 3 and 4. West of the main channel at transect 3, a headcut gully eroded fan material as flows overtopped and breached the channel’s west levee, resulting in flow diversion onto the fan surface. Erosion of multiple gullies into the fan surface occurred west of the main channel at transects 4–6.

Previous work (Morton and Hauser, 2001; Morton et al., 2008) observed that high-intensity summer monsoon rainfall events on Yucaipa Ridge produced 15 large debris flows that impacted Forest Falls from 1951 to 2005. Yet such storm events appear to occur more often than debris-flow incidents have been documented. The results of our study provide new insights on complexities that occur in this geomorphic system antecedent to debris flows. Figure 10 summarizes conceptual relationships between events of mass wasting, debris flow, and hyper-concentrated flow to the complexities observed for sediment production, transport, and storage in the SC debris-flow system. These complexities can entail non-linear responses over multiple years, indicating that the SC debris-flow system does not only function in single-event precipitation-driven processes of large debris-flow generation. For example, two mass-wasting events in the winter of 2017 associated with snowmelt produced a large sediment deposit within the canyon bedrock channel that remained meta-stable after monsoonal rainfall in August 2018 produced hyper-concentrated flow. A large percentage of the remaining material was coarse-grained boulders to cobbles, demonstrating that sediment loads within the debris-flow system at Forest Falls are contained in multiple components of erosion, transport, and deposition. Quantitative “thresholds” of rainfall intensity are required for sediment remobilization, recognizing that “storage effects” occur in the bedrock canyon and active drainage channel that accumulate significant loads of sediment from antecedent massmovement processes as well as by deposits of smaller water-limited debris flows that travel short distances. Debris flows from the canyon deposited sediments that aggraded within the active drainage channel and breached the active channel levee, allowing for subsequent spread of non-channelized flows onto the fan. Thus, a “feedback” process of channel avulsion progressed in subsequent flow events that eroded new channel pathways down fan.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 29–41

Cato and Goforth

Figure 7. Comparison of views over time (2013–2019) in the Area A–active Snow Creek alluvial fan channel (see photo index map for location). (a) May 10, 2013: Snow Creek fan, active channel, view south to Yucaipa Ridge, photo taken down gradient of curve at edge of east levee in abandoned channel. Black arrow points to the same clast on west levee (∼1 m) shown in (b). (b) July 19, 2019: Repeat photo is slightly left offset because of levee erosion toward east. Channel bed has also infilled with debris deposits that raised the surface elevation. Deforestation is evident (dead and knocked-down trees) from flow deposits that overtopped levee directly ahead and to the west (down gradient of curved channel). (c) Forest on the Snow Creek fan before channel avulsion process began on May 10, 2013. (d) Repeat photograph taken on January 8, 2020. White arrows point to the same trees (Pinus coulteri). Downslope flow was from right to left, and ∼1 m of debris was deposited upslope (right) of the two trees, where flow was blocked by a log. The near tree was located at the edge of a new channel cut into the fan, and the tree trunk exhibited abrasion scars from impact of recurrent flows. Note the removal of a smaller tree in the foreground where the new channel eroded and ingrowth of a sapling (Calocedrus decurrens) in the background beyond the area of deposition near the levee of the inactive channel.

Sediment produced by mass wasting of bedrock channel walls and stored within these steep bedrock channel sections does not appear to be an anomaly. The occurrence of discrete sediment deposits in bedrock channel systems has been observed in most nearby alluvial fans at Forest Falls (Ramirez, 2019). Even within other areas of the SC drainage basin, this occurs; for example, the other avalanche chute to the main SC channel is shown in Figure 5 to also contain a discrete sedimentary body. This sediment deposit is approximately 500 m long and ranges from 8 to 30 m wide and thus, depending on depth, could be thousands of cubic meters in volume. This would be consistent with the D1-facies (debris ﬂow) of Wells and Harvey (1987).

At broader timescales, recurrent debris flows, hyperconcentrated flows, and fan sediments eventually become redistributed over the fan surface. This occurs not only by superposition of flow deposits onto surficial chronosequences but also by cross-cutting relations where older debris deposits formerly buried at depth become exhumed and remobilized by erosional processes that activate when channel pathways move over the fan. CONCLUSIONS For almost two decades, the damaging effects of Snow Creek canyon debris flows on the Forest Falls community has been recognized, with the primary

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 29–41

Complexities in a Debris-Flow System

Figure 8. January 2020 LiDAR image of Area A shown in Figure 6. Imagery obtained using a Geoslam Horizon scanning unit. Camera locations for photos in Figure 6 are indicated by triangles. Darker areas from the architectural rendering indicate ground surface where debris deposition has occurred since 2013; also, individual tree canopies and fallen tree trunks appear dark. Lighter areas on the rendering are open channel areas with no tree canopy.

Figure 9. Cross-section proﬁles of the main channel of Snow Creek and fan surface to the west eroded by the channel avulsion process below a curve in the stream course recorded by GPS survey on January 8, 2020. The solid black triangles are locations of the channel center on each transect. The stream course curves west to east from transects 3 to 4, and the fan gradient trends downslope from south to north. The vertical:horizontal scale is 2:1.

causative mechanism attributed to high-intensity summer storms on Yucaipa Ridge. However, the nonlinear dynamics of 2017–2018 flow events suggests that this debris-flow system also functions with multiple sources of complexity. We describe the complexities of three system components to guide ongoing empirical field study of the debris-flow system: (1) thresholds of sediment delivery from sources at the higher reaches of bedrock canyons, (2) storage effects in sediment transport down the bedrock canyons, and (3) feedbacks in deposition, remobilization, and transport of sediment across the alluvial fan in dynamic channel filling, cutting, and avulsion processes. Complexity in geomorphic systems involves multiple possibilities for delayed and non-linear responses that require long-term study to parameterize in models. High-resolution geospatial data from techniques of GPS, LiDAR, InSAR, sUAVs, and GIS analyses are useful to map geomorphic features resulting from complexities in dynamic geomorphic systems and identify significant processes that are not evident when lacking such contexts of observation. This approach

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 29–41

Cato and Goforth

Figure 10. Flowchart for debris-flow system components and actions. (A) Mass movements as translational landslide or debris avalanche are triggered by seismic shaking or rapid snowmelt on weathered unstable slopes of the canyon rock mass; relatively dry sediment has limited transport. (B) High-intensity rainfall triggers fluvial erosion and transport of coarse rock and fine sediments in debris flows and hyper-concentrated flows. (C) Alluvial bar deposits in Snow Creek channel are remobilized and redeposited over shorter distances within the fan landform, including outcomes of channel filling, channel cutting, levee building, and channel avulsion. (D) Output of remobilized sediment in the Snow Creek channel from bedrock canyons above the fan, downstream directly into the Mill Creek trunk channel, typically as hyper-concentrated flow events. (E) Output of remobilized sediment from alluvial fan storage in the Snow Creek channel and adjoining fan surfaces subjected to flows. Actions A–C can result in inputs of sediment to the Snow Creek alluvial fan system. Actions B–E cause outputs of sediment from the Snow Creek alluvial fan system discharged into the Mill Creek trunk channel. From a safety standpoint, the focus is on what process will produce sediment onto the alluvial fan and into Mill Creek, where there is greater risk to loss of life or of property damage.

can also provide repeat measurements of event-speciﬁc changes at multiple spatial and temporal scales that would not otherwise be possible to survey given the large area and hazards. RECOMMENDATIONS FOR FUTURE RESEARCH The following recommendations for future work are identiﬁed by this study:

r The amount of sediment stored in the bedrock channel needs to be quantiﬁed and monitored for change. LiDAR and InSAR may allow more precise calculations so that the hypotheses developed can be tested. r The amount and location of sediment that is stored on the alluvial fan. The forested nature of this geomorphic surface has limited the use of historic

stereophotography or new high-resolution sUAV photography for this purpose. LiDAR and InSAR may provide the ability for this purpose. r A sediment budget to determine how much sediment is being produced in which area, where it is being temporarily being stored, and how much sediment passes through the system. r The locations where individual mass-wasting and debris-flow events are triggered need to be determined with higher spatial and temporal resolution. A time series of such data could potentially be useful to provide a statistical frequency distribution for occurrences of sediment mobilization, contrast different scales of flow events, and determine the flow distances contained within this system. An array of field instrumentation may be useful, such as infrasound acoustic sensors for mass-wasting events and geophones for ground vibration detection caused by debris flows (after Hurlimann et al., 2003; Abanco et al., 2014; and Havens et al., 2014), and could be calibrated with events documented by the aerial surveys and GPS surveys. REFERENCES Abanco, C.; Hurlimann, H.; and Moya, J., 2014, Analysis of the ground vibration generated by debris flows and other torrential at the Rebaixader monitoring site, central Pyrenees, Spain: Natural Hazards Earth System Sciences, Vol. 14, pp. 929–943. doi:10.5194/nhess-14-929-2014. California Geologic Survey (CGS), 1974, Special Study Zone (Alquist-Priolo Earthquake Fault Zone Map), SW ¼ San Gorgonio Mountain Quadrangle (15-minute series): CGS (formerly known at the California Division of Mines and Geology). Cruden, D. M. and Varnes, D. J., 1996, Landslide Types and Processes: Landslides Investigation and Mitigation: Transportation Research Board Special Report 247, 36 p. Dibblee, T. W. (Compiler), 1964, The San Gorgonio Mountain Quadrangle, San Bernardino and Riverside Counties, California: U.S. Geological Survey, Miscellaneous Geologic Investigations Map I-43. Gutierrez, C.; Bryant, W.; Saucedo, G.; and Wills, C. (Compilers), 2010, California Geological Survey 150th Anniversary Geologic Map of California: California Geologic Survey. Harden, D. 1998. California Geology. Prentice Hall, Englewood Cliffs, NJ. Havens, S.; Marshall, H.; Johnson, J.; and Nicholson, B., 2014, Calculating the velocity of a fast-moving snow avalanche using an infrasound array: Geophysical Research Letters, Vol. 41, pp. 6191–6198. doi:10.1002/2014GL061254. Hurlimann, M.; Rickenmann, D.; and Graf, C., 2003, Field and monitoring data for debris-flow events in the Swiss Alps: Canadian Geotechnical Journal, Vol. 40, pp. 161–175, doi:10.1139/T02-087. Johnson, K. and Sitar, N., 1990. Hydrologic conditions leading to debris-flow initiation: Canadian Geotechnical Journal, Vol. 27, pp. 789–801. doi:10.1139/t90-092. Keefer, D., 1984, Landslides caused by earthquakes: Geological Society America Bulletin, Vol. 95, pp. 406–421.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 29–41

Complexities in a Debris-Flow System Morton, D. M.; Alvarez, R. M.; Ruppert, K. R.; and Goforth, B., 2008, Contrasting rainfall generated debris ﬂows from adjacent watersheds at Forest Falls, southern California, USA: Geomorphology, Vol. 96, pp. 322–338. doi:10.1016/j.geomorph.2007.03.021. Morton, D. M. and Campbell, R. H., 1974, Spring mudﬂows at Wrightwood, Southern California: Quarterly Journal Engineering Geology, Vol. 7, pp. 377–384. Morton, D. M. and Hauser, R. M., 2001, A Debris Avalanche at Forest Falls, San Bernardino County, California, July 11, 1999: U.S. Geological Survey Open File Report 01-146, https://pubs.usgs.gov/of/2001/0146/pdf/of01-146.pdf. Murphy, B., 2015, Coseismic landslides. In Davies, T. (Editor), Landslide Hazards, Risks, and Disasters: Elsevier, Amsterdam, Netherlands, pp. 91–125. Murray, B. and Fonstad, M., 2007, Preface: Complexity (and simplicity) in landscapes: Geomorphology, Vol. 91, pp. 173– 177. doi:10.1016/j.geomorph.2007.07.011. National Center for Airborne Laser Mapping, 2014, NCALM Seed- Airborne Laser Swath Mapping (ALSM) Survey of the San Andreas Fault (SAF) System of Central and Southern California: Electronic document, available at http://opentopo. sdsc.edu/raster?opentopoID=OTSDEM.052016.26911.1 National Geodetic Survey, 2017–2018. Online Computation Page for GEOID03 GEOID Height: Electronic document, available at https://www.ngs.noaa.gov/cgi-bin/GEOID_STUFF/ geoid03_prompt1.prl O’Keefe, K., 2017, Characterization of the Winter–Spring 2017 Mass Movement Episode in Snow Canyon, Forest Falls, San Bernardino County, California: Unpublished Undergraduate Research Project, Department of Geological Sciences, California State University, 29 p. Phillips, J. D., 2003, Sources of non-linearity and complexity in geomorphic systems: Progress Physical Geography, Vol. 27, pp. 1–23. doi:10.1191/0309133303pp340ra. Ramirez, O., 2019, Forest Falls Debris Flows and Potential Hazard to Residents: Unpublished Undergraduate Research Project, Department of Geological Sciences, California State University, 21 p.

San Bernardino County, 2010, San Bernardino County General Plan, Geologic Hazard Overlays: Electronic document, available at http://cms.sbcounty.gov/lus/Planning/ ZoningOverlayMaps/GeologicHazardMaps.aspx San Bernardino County Flood Control District (SBCFCD), 2017, Meteorological Sensor ID: 2900: Electronic document, available at http://www.sbcounty.gov/dpw.pwg/alert/ reports.html Tan, S. S. (Compiler), 1990, Landslide Hazards in the Yucaipa and Forest Falls Quadrangles, San Bernardino County, California: Division of Mines and Geology Landslide Hazard Identiﬁcation Map No. 18. Temme, A.; Keiler, M.; Karssenberg, D.; and Lang, A., 2015, Complexity and non-linearity in Earth surface processes: Concepts, methods, and applications: Earth Surface Processes Landforms, Vol. 40, pp. 1270–1274. doi:10.1002/esp.3712. Turk, J. K.; Goforth, B. R.; Graham, R. C.; and Kendrick, K. J., 2008, Soil morphology of a debris ﬂow chronosequence in a coniferous forest, southern California, USA: Geoderma, Vol. 146, pp. 157–165. doi:101016/j.geoderma.2008.05.012. U.S. Geological Survey, 1994, Forest Falls Quadrangle: 7.5 Minute Topographic Series, 1970 (photo revised, 1988; minor revisions 1994). U.S. Geological Survey, 2017, Quaternary Fault and Fold Database for the United States: Electronic document, available at https://earthquake.usgs.gov/hazards/qfaults U.S. Geological Survey, 2018a, Geographical Names Information System: Electronic document, available at https://geonames.usgs.gov/domestic/index.html. U.S. Geological Survey, 2018b. National Map Server: Electronic document, available at https://www.usgs.gov/core-sciencesystems/ngp/tnm-delivery/gis-data-downloadOnline Wells, S. and Harvey, A., 1987, Sedimentologic and geomorphic variations in storm-generated alluvial fans, Howgill Fells, Northwest England: Geological Society America Bulletin, Vol. 98, pp. 182–198. Yule, D. and Sieh, K., 2003, Complexities of the San Andreas fault near San Gorgonio Pass: Implications for large earthquakes: Journal Geophysical Research, Vol. 108, No. B11, p. 2548.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 29–41

Time Since Burning and Rainfall Characteristics Impact Post-Fire Debris-Flow Initiation and Magnitude LUKE A. MCGUIRE* University of Arizona, Department of Geosciences, 1040 East 4th Street, Tucson, AZ 85721

FRANCIS K. RENGERS U.S. Geological Survey, 1711 Illinois Street, Golden, CO 80401

NINA OAKLEY Western Regional Climate Center, Desert Research Institute, 2215 Raggio Parkway, Reno, Nevada 89512, USA

JASON W. KEAN DENNIS M. STALEY U.S. Geological Survey, 1711 Illinois Street, Golden, CO 80401

HUI TANG Section 4.7, Earth Surface Process Modeling, German Research Center for Geosciences (GFZ), Telegrafenberg, Building A 27, 14473 Potsdam, Germany

MARIAN DE ORLA-BARILE Center for Western Weather and Water Extremes, Scripps Institute of Oceanography, 9500 Gilman Drive, La Jolla, CA 92037

ANN M. YOUBERG University of Arizona, Arizona Geological Survey, 1955 East 6th Street, Tucson, AZ 85721

Key Terms: Wildfire, Threshold, Infiltration, Erosion, Geomorphology ABSTRACT The extreme heat from wildfire alters soil properties and incinerates vegetation, leading to changes in infiltration capacity, ground cover, soil erodibility, and rainfall interception. These changes promote elevated rates of runoff and sediment transport that increase the likelihood of runoff-generated debris flows. Debris flows are most common in the year immediately following wildfire, but temporal changes in the likelihood and magnitude of debris flows following wildfire are not well constrained. In this study, we combine measurements of soil-hydraulic properties with vegetation survey data and numerical modeling to understand how debris-flow threats are

*Corresponding author email: lmcguire@email.arizona.edu

likely to change in steep, burned watersheds during the first 3 years of recovery. We focus on documenting recovery following the 2016 Fish Fire in the San Gabriel Mountains, California, and demonstrate how a numerical model can be used to predict temporal changes in debris-flow properties and initiation thresholds. Numerical modeling suggests that the 15-minute intensityduration (ID) threshold for debris flows in post-fire year 1 can vary from 15 to 30 mm/hr, depending on how rainfall is temporally distributed within a storm. Simulations further demonstrate that expected debris-flow volumes would be reduced by more than a factor of three following 1 year of recovery and that the 15-minute rainfall ID threshold would increase from 15 to 30 mm/hr to greater than 60 mm/hr by post-fire year 3. These results provide constraints on debris-flow thresholds within the San Gabriel Mountains and highlight the importance of considering local rainfall characteristics when using numerical models to assess debris-flow and flood potential.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 43–56

McGuire, Rengers, Oakley, Kean, Staley, Tang, Orla-Barile, and Youberg

INTRODUCTION Wildfire is a well-documented catalyst for change in hydrologic and geomorphic systems (e.g., Shakesby and Doerr, 2006). Post-wildfire reductions in infiltration capacity (Ebel and Moody, 2017) and canopy interception (Stoof et al., 2012) promote increased runoff. Increased runoff, combined with the effects of lower critical thresholds for sediment entrainment (Moody et al., 2005) and a high percentage of bare soil, leads to a substantial increase in debris-flow likelihood after a wildfire. Post-wildfire debris flows are often generated when runoff concentrates in steep channels and mobilizes large volumes of sediment in contrast to debris flows that initiate from shallow landslides (e.g., Meyer and Wells, 1997; Cannon et al., 2008; Gabet and Bookter, 2008; and Kean et al., 2011). With few exceptions (e.g., Cannon et al., 2008), previous work has focused on the threats posed by runoff-generated debris flows in the first year following disturbance by wildfire; therefore, the extent to which debris-flow hazards persist into subsequent years is not well understood. Rainfall intensity-duration (ID) thresholds are commonly used to assess post-wildfire debris-flow potential, with debris flows often initiating once a critical rainfall intensity is exceeded (Cannon et al., 2008; Staley et al., 2013). Staley et al. (2017) recently developed an empirical model to predict debris-flow likelihood as a function of terrain attributes, soil burn severity, and rainfall intensity (averaged over 15, 30, or 60 minutes). However, it is not clear how rainfall ID thresholds change with time following wildfire because data regarding debris-flow occurrence are most common in the first post-wildfire year and because there is no clear connection between the magnitude of empirically derived rainfall ID thresholds and the hydrologic and geomorphic variables that are changing as the landscape recovers. Since overland flow is a necessary condition for runoff-generated debris flows, it is critical to understand how wildfire-driven changes to soil infiltration capacity vary with time since burning and how the magnitude of those changes translates into changes in debris-flow potential. Post-wildfire reductions in infiltration capacity are often attributed to surface soil sealing (Larsen et al., 2009), hyper-dry conditions (Moody and Ebel, 2012), or increased soil water repellency (DeBano, 2000; Shakesby and Doerr, 2006), which may persist for up to 5 years but typically decays over timescales of 1–2 years (e.g. Larsen et al., 2009). The percentage of bare soil, which is initially high following wildfire and decreases as vegetation recovers, is also likely to be a key factor in determining debris-flow potential since bare soil on hillslopes is particularly vulnerable to erosion. Hillslope erosion can account for a substantial amount of the sediment within post-

wildfire debris flows in certain cases (e.g. Smith et al., 2012; Staley et al., 2014; and Rengers et al., 2016b) and contribute to sediment bulking in the channel that increases flow depth and discharge. In addition to a lack of data regarding how debrisflow likelihood varies with time following a fire, there is also a general need to assess the extent to which debris-flow properties can be predicted by simple metrics that summarize rainfall characteristics, such as the peak 15-minute average rainfall intensity (I15 ). It is common practice to use rainfall intensity, averaged over a specified duration, to assess the potential for debris flows (Staley et al., 2013). Similarly, empirical models used to predict the volume of post-fire debris flows rely on simple measures of rainfall characteristics, including I15 (Gartner et al., 2014). Such methods have proved invaluable for rapid assessments of post-fire debris-flow hazards. There is also reason to believe that I15 is a particularly useful metric for predicting the initiation and magnitude of post-fire debris flows because runoff generates these flows and the magnitude of post-wildfire runoff at the watershed scale correlates well with rainfall over timescales of 10–15 minutes (Kean et al., 2011; Raymond et al., 2020). However, given the sensitivity of runoff and sediment transport to rainfall intensity, it is reasonable to assume that two rainstorms could have the same peak 15-minute rainfall intensity but still produce different debris-flow responses. Consider the extreme case where the rainfall intensity is 10 mm/hr for 5 minutes, rapidly increases to 100 mm/hr for 5 minutes, and then returns to an intensity of 10 mm/hr for 5 minutes. The runoff and debris-flow response resulting from this type of storm, which would have a peak I15 of 40 mm/hr, may be quite different from that of a storm where the rainfall intensity is constant at 40 mm/hr. Rapid increases in rainfall intensity over short durations are not uncommon in nature, including in southern California, where they typically accompany rainstorms classified as narrow cold-frontal rain bands (NCFRs). NCFRs are narrow bands (often <5 km wide) of high-intensity rainfall occurring parallel to and in the proximity of a cold front (Figure 1). NCFRs have a history of triggering post-fire debris flows in the Transverse Ranges of southern California (Oakley et. al., 2017, 2018). In January 2018, an NCFR impacted the burn scar of the 2018 Thomas Fire, triggering debris flows that produced widespread damage and resulted in 23 fatalities (Oakley et al., 2018; Kean et al., 2019). Here, we provide a simple example using two distinct hyetographs, one for an NCFR and one with a Gaussian distribution, to evaluate how the shape of the rainfall hyetograph (e.g., I5 /I15 ) influences debris-flow response for the same I15 . To the extent that particular rainfall hyetograph characteristics can be linked to

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 43–56

Post-Fire Debris-Flow Initiation

Figure 1. (a) Base reﬂectivity (dBZ) from the NEXRAD weather radar in Santa Ana (KSOX) at 19:32 UTC on January 2, 2006, showing a narrow cold-frontal rain band. Warmer colors indicate higher-intensity precipitation, while cooler colors indicate lighter precipitation. (b) Time series of the 1-minute and 15-minute average rainfall intensity based on measured rainfall at the Clear Creek School gauge. Time is in minutes from 18:16 UTC on January 2, 2006.

storm type (e.g., NCFR, isolated convective cells), this is a first step toward understanding the role of storm type on post-fire debris-flow initiation and magnitude. Understanding the storm types most likely to produce impactful debris flows would improve the situational awareness of weather forecasters who may be tasked with the responsibility of issuing warnings for communities downstream of burned areas. In this study, we combine site measurements of soilhydraulic properties and canopy/ground cover with a physically based numerical model to explore how changing site characteristics influence the initiation and magnitude of runoff-generated debris flows. The numerical model, developed by McGuire et al. (2017), represents the coupled processes of runoff, sediment transport, and debris-flow initiation. In addition to changing site characteristics, we force this model with two different rainstorms—one that is an idealized storm with a rainfall time series that has the shape of a Gaussian distribution and one that is more representative of rainfall conditions associated with an NCFR—to assess the sensitivity of results to differences in rainfall characteristics. We hypothesize that (1) the I15 threshold for debris-flow initiation will be sensitive to the temporal distribution of rainfall (i.e., NCFRs versus a Gaussian distribution of rainfall), (2) the I15 threshold will increase with time following wildfire, and (3) the typical volume of debris flows will decrease with time following wildfire. Moreover, we aim to quantify the timescale for substantial recovery of soil-hydraulic properties following wildfire and pro-

vide physical explanations for any trends observed between time since burning, debris-flow initiation thresholds, and debris-flow volume. STUDY AREA The study area, which we refer to as Las Lomas, is located near the headwaters of a 0.1-km2 watershed that drains into the Las Lomas debris basin (Figure 2). Data from the Las Lomas study site are used to quantify changes in soil-hydraulic properties and canopy/ground cover with time following the 2016 Fish Fire. The Fish Fire, which started on June 21, 2016, burned 4,253 acres of the Angeles National Forest in the San Gabriel Mountains (SGM) near Los Angeles, California (Figure 2). The wildfire burned in rugged terrain with steep hillslopes dominated by chaparral vegetation. Soils in the SGM are generally thin (0.5–1 m), rock outcrops are common, and a highly weathered layer of saprolite is occasionally exposed on the hillslopes (Staley et al., 2014). Based on particle size analysis of hillslope sediment at the site, the soil texture is classified as sandy loam (Tang et al., 2019). Repeat measurements of soil-hydraulic properties were conducted on a roughly 40° hillslope in an area that, based on field observations and the severity indicators described by Parson et al. (2010), experienced moderate to high soil burn severity during the Fish Fire. No vegetation canopy remained, all litter and duff at the surface had been consumed by the wildfire, and soils were generally water repellent at/near

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 43–56

McGuire, Rengers, Oakley, Kean, Staley, Tang, Orla-Barile, and Youberg

Figure 2. (a) Overview of study area. (b–d) Photos looking across the hillslope along a transect at the Las Lomas site where inﬁltration measurements were conducted.

the surface. A series of rainstorms between December 2016 and February 2017 incised a network of rills through the study area and produced a number of debris flows and floods at the outlet of the drainage basin (Tang et al., 2019). METHODS Field Measurements Following the Fish Fire Field-saturated hydraulic conductivity (Ks ) and sorptivity (S) were determined through in situ measurements conducted with a mini disk tension infiltrometer over a 33-month period following the wildfire. The tension infiltrometer has a disk with a radius of 2.25 cm. The suction head was set to 1 cm for all measurements. Measurements were made during site visits to the Las Lomas study area in September 2016, November 2016, January 2017, February 2017, July 2017, March 2018, and March 2019. Measurements were performed every 1 m along a 20-m transect that extended in the cross-slope direction, with the exception of those made in September 2016. When time permitted, additional measurements were made on the hillslope in the vicinity of the established transect. In September 2016, the transect had not yet been established, and measurements were made in nearby areas burned at moderate or high severity. A total of 22, 31, 26, 37, 20, 21, and 20 infiltration measurements were made during the site visits in September 2016, November 2016, January 2017, February 2017, July 2017, March 2018, and March

2019, respectively. During each measurement, the total volume of water infiltrated is tracked as a function of time and must later be post-processed to infer field-saturated hydraulic conductivity (Ks ) and sorptivity (S) (e.g., Zhang, 1997). Estimates of Ks and S were derived following the methodology of Zhang (1997). Letting I denote the total volume infiltrated at time √ t during the measurement, I = C1 t + C2 t, where C1 = A1 S, C2 = A2 Ks , and A1 = 3.89 and A2 = 1.04 are empirical coefficients whose values depend on soil texture. Since I and t are known at various times throughout each measurement, we determined C1 and C2 using the three different curve-fitting techniques proposed by Vandervaere et al. (2000). For each mini disk measurement, this results in three estimates of S and Ks that are then averaged to obtain single values for S and Ks . The wetting front suction head (hf ), a parameter in the Green-Ampt infiltration model, can then be estimated as hf = S2 /2Ks θs (Ebel and Moody, 2017), where θs = 0.4 denotes the volumetric water content at saturation. In some instances, however, it was clear that the infiltration data did √ not follow the trend suggested by the equation I = C1 t + C2 t. This could be due to error during the measurement process, such as poor contact between the mini disk and the soil surface, or could be the result of a layered system (e.g., a thin waterrepellent layer on top of a more wettable soil) that is not well described by the assumed infiltration model. In these cases, we could not obtain estimates for S and Ks . In March 2018 and March 2019, we conducted vegetation surveys on a hillslope adjacent to the

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 43–56

Post-Fire Debris-Flow Initiation

infiltration transect using the point-intercept method (e.g., Crocker and Tiver, 1948). A measuring tape was extended, and measurements were taken on 20-cm intervals along the transect (105 measurements in 2018 and 76 in 2019). We sighted directly down toward the surface with a laser pointer and recorded the first obstacle that intercepted the light. The laser hit either the vegetation canopy, bare soil, litter, or a rock. Litter was classified as any loose plant material on the soil surface. Any sediment with a diameter greater than 5 mm was classified as rock cover. If the laser hit any portion of the canopy, the maximum height of that vegetation was recorded. Numerical Model The numerical model represents fluid flow using the shallow-water equations, which contain additional source terms to account for changes in flow resistance as a function of sediment concentration (McGuire et al., 2016, 2017). The model is described in detail by McGuire et al. (2016, 2017) and is only briefly summarized here. Infiltration is modeled with the GreenAmpt equation, using estimates of Ks and hf obtained from field measurements. The infiltration capacity of the soil (Ic ) is given by Zf + hf + h , Ic = Ks Zf where h denotes flow depth, Zf = V/(θs − θi ) is the depth of the wetting front, V is the total depth of water infiltrated, and θi is the initial volumetric soil water content. Hydraulic roughness was represented using a depth-dependent Manning friction coefficient (Mügler et al., 2011). More specifically, the Manning coefficient varies in space and time according to −0.33 h h ≤ hc , n = n0 h c h > hc , n0 where hc = 0.003 m is a critical flow depth. The coefficient n0 is calibrated. The Hairsine-Rose model (Hairsine and Rose, 1992a, 1992b) is used to account for sediment entrainment and deposition, as described in detail by McGuire et al. (2016). In the Hairsine-Rose model, particles can be detached and entrained into the flow via raindrop impact or flow-driven detachment. The rate at which sediment is detached by raindrops is a function of flow depth, rainfall intensity, and raindrop diameter, while the rate of flow-driven sediment detachment is a function of stream power. Since the canopy and ground cover (e.g., litter) can shield the underlying soil from raindrop impact, changes in ground and canopy cover will also influence the rate of

raindrop-driven sediment detachment (e.g., McGuire et al., 2016). Sediment being transported by the flow can also influence flow rheology. In regions of flow where the sediment concentration is below 20 percent, flow resistance is accounted for solely through the above Manningtype equation. In regions of flow where the sediment concentration is above 40 percent, we account for additional resistance associated with debris flow using a Coulomb friction approach (e.g., Iverson and Denlinger, 2001) where the effective basal normal stress is modified by pore-fluid pressure within the flow. In all simulations, the ratio of pore-fluid pressure to total basal normal stress (λ) is set to a constant value of λ = 0.65. A debris flow is identified within the model as flow with a sediment concentration above 40 percent. At intermediate sediment concentrations between 20 and 40 percent, the magnitude of the debris-flow resistance term is scaled by a multiplicative factor that ranges from 0 at a sediment concentration of 20 percent to 1 at a sediment concentration of 40 percent. Varying flow resistance terms as a function of sediment concentration enables the model to better represent the transition from water-dominated flow to debris flow. The onset and magnitude of runoff generated by the model depends on the infiltration capacity of the soil as predicted by the Green-Ampt equation and the rainfall hyetograph. Runoff, in turn, facilitates sediment transport that leads to changes in topography and, potentially, to debris flow initiation. The model is capable of generating debris flows through two different mechanisms. First, it is possible that hydrologic conditions and sediment availability promote substantial entrainment and limited deposition of sediment. In this case, sediment concentration (c) may exceed that typically associated with debris flows in the model (e.g., c > 40 percent) as a result of the progressive addition of sediment to the flow column. This style of debris flow initiation is similar to the progressive bulking mechanism that has been described in past studies (e.g., Cannon et al., 2001; Gabet and Bookter, 2008). Debris flows may also initiate within the model as a result of the en masse failure of bed sediment along a failure plane (e.g., Takahashi, 1978; Kean et al., 2013). This scenario often develops in the model when there are areas of preferential deposition, such as channel reaches with relatively low slopes, that lead to the formation of a sediment dam. If the deposited sediment becomes unstable as a mass, it fails and can lead to the formation of a debris flow surge if the added sediment volume is sufficient to locally increase the sediment concentration above 40 percent. The stability of a deposited mass of sediment is determined locally using a factor of safety based on the ratio of resisting

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 43–56

McGuire, Rengers, Oakley, Kean, Staley, Tang, Orla-Barile, and Youberg

Figure 3. (a) Rainfall time series typical of narrow cold-frontal rain bands with diﬀerent peak 15-minute rainfall intensities. (b) Idealized rainstorms with rainfall intensity shaped like a Gaussian function. For display purposes, only three curves are shown; however, for modeling, we used curves with peak 15-minute rainfall intensity (I15 ) varying from 10 to 60 mm/hr in intervals of 5 mm/hr. (c) Shaded relief map of the Arroyo Seco watershed.

and driving forces acting on a column of sediment. If the driving forces are greater than the resisting forces at a particular model grid cell, then the column of sediment is determined to be unstable, and the sediment in that grid cell is instantaneously added to the above flow. A complete description of the failure criteria used in the model is given by McGuire et al. (2017). Debris-Flow Simulations In model simulations, we applied the vegetation and hydrologic measurements to simulate runoff, sediment transport, and debris-flow initiation within a 0.012-km2 catchment (Figures 2 and 3c). Although this particular catchment did not burn in the Fish Fire, it is located within the SGM and was chosen for several reasons. First, the distribution of slope values within the catchment, referred to as Arroyo Seco, is similar to other debris-flow–producing basins in the SGM (Kean et al., 2011). Second, a terrestrial laser scanner survey of the entire Arroyo Seco site (Staley et al., 2014) provided a high-resolution digital elevation model (DEM) that could not be matched by any watershed at the Fish Fire site. Finally, the small size of the basin makes it

possible to perform the large number of simulations needed for hypothesis testing. We coarsened the DEM from its original resolution of 2 cm (Staley et al., 2014) to a grid spacing of 37.5 cm to further increase computational efficiency. Since the goal of this study is to understand debris-flow initiation thresholds and magnitude, as well as their sensitivity to rainfall characteristics and recovery, the location of the DEM relative to the perimeter of the Fish Fire is not critical; we focused on finding a DEM that is representative of the areas where post-fire debris flows tend to initiate in the SGM. Runoff is driven by a set of idealized rainstorms (Figure 3b), with peak 15-minute rainfall intensities (I15 ) varying from 10 to 60 mm/hr in increments of 5 mm/hr, as well as a family of rainstorms that are designed to have characteristics typical of an NCFR (Figures 2 and 3a). The storm events designed to represent an NCFR were generated through a two-step process. First, we identified a representative NCFR event impacting the SGM and extracted the rainfall data for that event. The selected event occurred on January 2, 2006, and had a well-defined narrow band of high intensity rainfall along the cold front.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 43–56

Post-Fire Debris-Flow Initiation Table 1. Model parameters used for simulations of runoff, sediment transport, and debris-flow initiation. Notation follows McGuire et al. (2016). When appropriate, values for years 1, 2, and 3 are presented and separated by commas. Median values are reported for Ks and hf . Parameter Name (Symbol) Roughness coefficient (n0 ) Raindrop detachability (a0 ) Raindrop redetachability (ad0 ) Fraction of effective stream power (F) Fraction canopy cover Fraction bare soil Field-saturated hydraulic conductivity (Ks ) Wetting front suction head (hf )

Units 1/3

s/m kg/m/s kg/m/s — — — mm/hr m

It also exhibited the “gap and core” structure—cores of high-intensity precipitation separated by gaps of low-intensity precipitation—that is characteristic of an NCFR (e.g., Jorgensen et al., 2003) (Figure 2). Rainfall data from 18:16 and 19:46 UTC were extracted from the gauge at Clear Creek School (34°16 40, 118°10 15), which is maintained by the Los Angeles County Department of Public Works. The peak I15 of this storm is approximately 53 mm/hr (Figure 2). We then scaled the rainfall intensity of this NCFR by the amount needed in order to generate a series of NCFR-type rainstorms with peak values of I15 that match those of the family of idealized storms. The objective of this analysis is to take a first step toward exploring how different types of rainstorms may influence debris-flow response, even if their peak I15 is identical. The ratio of peak I5 to peak I15 is a simple way to quantify a key difference between the family of idealized NCFRs and the family of designed rainstorms with a Gaussian distribution hyetograph, hereafter referred to simply as NCFRs and the designed rainstorms, respectively. The ratio I5 /I15 is approximately two to one for the NCFRs and one to four for the designed rainstorms. All simulations were performed using the same parameters and model setup as reported in McGuire et al. (2017) unless otherwise noted (Table 1). Simulations begin with no runoff and an initial volumetric soil moisture content of 0.05. McGuire et al. (2016) calibrated the Hairsine-Rose sediment transport parameters at the Arroyo Seco site by comparing simulated erosion patterns to those generated from repeat terrestrial laser scanning surveys (Staley et al., 2014). The calibrated model was able to reproduce measured patterns of erosion and deposition as well as the timing of runoff and debris-flow activity at the outlet of the watershed during a monitored rainstorm in the first year following the Station Fire (McGuire et al., 2016, 2017). Here, the roughness coefficient is set to a value of n0 = 0.05 s/m1/3 , which is in the range of calibrated roughness values for recently burned, low-order drainage basins in the SGM (Rengers et al., 2016a).

Value (Year 1, Year 2, Year 3)

Source

0.05 9,000 410,000 0.0065 0, 0.29, 0.77 1, 0.63, 0.2 19, 13, 28 0.006, 0.022, 0.026

Calibrated Calibrated Calibrated Calibrated Measured Measured Measured Measured

The fraction of bare soil exposed to raindrop impact is assumed to be 1.0 in the first year following the fire based on field observations of negligible vegetation and litter cover (Figure 2b). Infiltration rates were computed for year 1 using the Ks and hf values obtained in September 2016 and November 2016, while Ks and hf values obtained in July 2017 and March 2018 were used for year 2, and values measured in March 2019 were used for year 3 simulations. Each pixel within the computational domain was randomly assigned a value from the measured distribution of Ks and hf . Due to the number of pixels in the computational domain, we found that differences among simulations performed with different realizations of Ks and hf were not substantial. In locations where the slope exceeded 45°, we assumed that bedrock or saprolite was exposed at the surface and therefore set Ks = 0. A total of 7 out of 60 measurements attempted in September and November 2016 were terminated due to long measurement times (i.e., no measurable amount of water infiltrated within 5 minutes), potentially because of extreme water repellency or the presence of saprolite at very shallow depths. Including these data points in the analysis as locations with Ks = 0 and hf = 0 did not significantly influence computed debris-flow thresholds or volumes, so they were neglected. A total of 22 simulations were performed using the measured infiltration and vegetation characteristics from the first post-wildfire year. Eleven of these simulations were performed to assess how debris-flow properties change for the designed storms (Gaussian distribution of rainfall) with I15 varying from 10 to 60 mm/hr. Eleven additional simulations were used to assess how debris-flow properties change when using precipitation from an actual NCFR, with intensities scaled to give a different peak I15 varying from 10 to 60 mm/hr. A second and third set of numerical experiments, each containing 22 simulations, were performed using the measured infiltration and vegetation characteristics from the second and third post-wildfire years. Finally, we performed two final sets of simulations

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 43–56

McGuire, Rengers, Oakley, Kean, Staley, Tang, Orla-Barile, and Youberg Table 2. Summary of median and interquartile range (in parentheses) of soil-hydraulic properties as a function of time since burning. Months Parameter Name (Symbol)

Units

3 (n = 22)

5 (n = 31)

7 (n = 26)

8 (n = 37)

13 (n = 20)

21 (n = 21)

33 (n = 20)

17 (18) 23 (23) 20 (30) 15 (18) 11 (21) 21 (32) 28 (37) Field-saturated hydraulic mm/hr conductivity (Ks ) Sorptivity (S) mm/hr1/2 3 (12) 11 (12) 11 (12) 7 (13) 18 (11) 20 (13) 25 (19) 0.002 (0.011) 0.008 (0.031) 0.016 (0.023) 0.004 (0.020) 0.032 (0.080) 0.021 (0.073) 0.026 (0.049) Wetting front suction head m (hf )

using measured soil-hydraulic properties from year 2 and year 3 but vegetation characteristics (i.e., 100 percent bare ground) consistent with year 1. The goal of this last series of simulations was to quantify the relative impact of vegetation recovery and soil recovery on post-wildfire debris flows. Flow depth, discharge, and sediment concentration at the basin outlet were recorded in all cases to assess differences among simulations. Debris flows were identified at the outlet of the basin based on exceedance of a sediment concentration threshold of 40 percent. Flows with a sediment concentration (c) less than 40 percent were classified as floods. The amount of sediment exiting the basin (kg), sediment concentration (c) at the basin outlet, and peak debris-flow (i.e., flows with c > 40 percent) discharge at the outlet were stored for each simulation. Debris-flow volumes were estimated for each storm by determining the cumulative sediment discharge at the basin outlet during time periods when the sediment concentration exceeded 40 percent and then converting the resulting sediment mass to a volume by assuming a bulk density of 1,500 kg/m3 . Summarizing debris flow size in terms of a volume allows for more direct comparisons with field-based estimates of debris-flow volume and outputs from empirical models used to predict debris-flow volume (e.g., Gartner et al., 2014). RESULTS Changes in Soil-Hydraulic Properties and Ground Cover Repeat field measurements of soil-hydraulic properties reveal changes in sorptivity (S) and wetting front suction head (hf ) with time since burning. The median wetting front suction head (hf ) increased with time from roughly 0.002 m in September 2016 (3 months post-fire) to 0.026 m by March 2019 (33 months postfire). A Kruskal-Wallis test indicates statistically significant differences among the seven different groups of hf measurements (p < 0.01) made over the course of the 33-month monitoring period as well as differences among the seven groups of sorptivity measure-

ments (p < 0.01). Wilcoxon rank sum tests with a significance level of 0.05 were then used to assess differences between two groups of measurements to determine when significant changes occurred. According to a Wilcoxon rank sum test, the distributions of hf after 3 and 5 months differ significantly (p < 0.01), as do the distributions of S after 3 and 5 months (p < 0.01). The distributions of hf after 8 and 13 months also differ significantly (p < 0.01), as do the distributions of S after 8 and 13 months (p < 0.01). In addition to a change in the median of hf between 8 and 13 months, there is also a substantial increase in the interquartile range from approximately 0.02 to 0.08 m (Table 2) that is consistent with a more general increase in the spread in the hf distributions after 13 months of recovery (Figure 4). In contrast, fieldsaturated hydraulic conductivity (Ks ) does not vary as systematically with time following the wildfire (Figure 4 and Table 2). A Kruskal-Wallis test indicates that measured distributions of Ks over the 33-month monitoring period (p = 0.06) are not significantly different. Despite a lack of statistical significance at the 0.05 level, both the mean and the median values of Ks increased substantially over the monitoring period. The median Ks increased from 17 mm/hr after 3 months to 28 mm/hr after 33 months, whereas the mean increased from 24 to 43 mm/hr over that same time period. The fraction of bare soil decreased from 1.0 immediately following the wildfire to 0.63 in March 2018 after 21 months of recovery and 0.2 in March 2019 after 33 months of recovery. The reduction in bare ground was due primarily to an increase in canopy cover fraction from approximately 0 to 0.29 by March 2018 and 0.77 by March 2019. The fractions of litter cover were 0.07 and 0.03 in March 2018 and 2019, respectively, while rock cover was 0.01 and 0, respectively. Although the recovering vegetation may be effective at reducing direct raindrop impact on the soil surface, it likely had a minimal ability to intercept and store water, particularly in March 2018, since the average vegetation height was less than 10 cm. By March 2019, the mean canopy height had increased to approximately 80 cm.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 43–56

Post-Fire Debris-Flow Initiation

Figure 4. (a) Field-saturated hydraulic conductivity (Ks ), (b) sorptivity (S), and (c) suction head (hf ) derived from ﬁeld measurements at diﬀerent times following the June 2016 Fish Fire.

Simulations of Erosion and Debris Flows Debris flows initiated in response to lower-intensity rainstorms in year 1 relative to years 2 and 3 (Figure 5). The first sign of debris-flow activity at the lower outlet (i.e., debris-flow volume greater than 5 m3 ) during year 1 occurs in response to the NCFR with a peak I15 of 15 mm/hr. In contrast, no debris flows initiate in response to the designed rainstorm (Gaussian distribution of rainfall) until the peak I15 is greater than or equal to 30 mm/hr. In year 2, simulations suggest than an NCFR with a peak I15 of at least 35 mm/hr is needed to produce a debris flow, whereas the peak I15 of the designed storm must exceed 40 mm/hr. In the theoretical case where vegetation recovery is neglected in year 2 and only changes in infiltration capacity are taken into account, the I15 required to produce a debris flow is 15 mm/hr for an NCFR and 35 mm/hr for the designed storm. In year 3, both the NCFR and the designed rainstorms having a peak I15 of 60 mm/hr failed to produce a debris flow with a volume greater than 5 m3 (Figure 5). If vegetation recovery is neglected, then the I15 threshold for debris flows in year 3 would be 20 mm/hr for an NCFR and 50 mm/hr for the designed storm. The volume of debris flows generated by NCFRs increases with peak I15 (Figure 5). The volume of debris flows initiated by the designed rainstorms tends to increase initially with I15 but then decreases slightly for higher values of I15 (Figure 5). The total volume of sediment eroded, however, continues to increase with peak I15 regardless of specifics of the hyetograph (Figure 5). Debris-flow volumes and total sediment eroded for a given rainstorm are highest in year 1, as expected. For a given I15 , the mean volume of debris flows generated by NCFRs in year 2 is, on average, less than one-third of that produced in year 1. The volume of debris flows generated by the designed storm in year

2 is, on average, roughly one-fourth of that generated in year 1. Changes in vegetation cover appear to play a key role in determining debris-flow volume. If vegetation recovery is negligible from year 1 to year 2, then mean debris-flow volumes associated with NCFRs and designed storms would differ by factors of only 1.2 and 1.1, respectively. It is also noteworthy that NCFRs tend to produce debris flows with greater peak discharges (Figure 5) and peak flow depths, as seen in the modeled hydrographs (Figure 6). NCFRs also produce debris flows that are larger than those produced by the designed storm, particularly when 15-minute rainfall intensities are above 50 mm/hr. DISCUSSION Field measurements constrain recovery timescales following disturbance by wildfire in the SGM (Figure 4 and Table 2) and provide insight into how soil and vegetation recovery translate into recovery from a debris-flow hazards perspective (Figures 5 and 6). Simulations of erosion and debris-flow initiation suggest that there should be a substantial increase in rainfall ID thresholds over the first 3 years of recovery following wildfire as well as decreases in expected debris-flow volume (Figure 5). Simulations do not take into account the reductions in sediment supply or increases in hydraulic roughness that are likely to occur due to sequential flow events between the first year and the third year after a wildfire (e.g., Tang et al., 2019). Results reported here can therefore be viewed as conservative, with even greater increases in ID thresholds and reductions in volume being likely in many natural systems. Our field measurements indicate a significant change in the distributions of sorptivity (S) and suction head (hf ) between 3 and 5 months of recovery as well as between 8 and 13 months of recovery.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 43–56

McGuire, Rengers, Oakley, Kean, Staley, Tang, Orla-Barile, and Youberg

Figure 5. Model simulations of debris ﬂow properties in (a–c) year 1, (d–f) year 2, and (g–i) year 3 in response to the two rainstorms with diﬀerent peak values of I15 . The combined fraction of vegetation and ground cover is denoted by Cv .

However, it is not clear whether the change in S and hf between 3 and 5 months was due to recovery or due to the change in measurement location that took place between 3 and 5 months. Low values of S and hf in the year immediately following wildfire are consistent with a recent compilation of soil-hydraulic properties from burned soils (Ebel and Moody, 2017). Previous studies monitoring temporal changes in soilhydraulic properties after fire have often focused more on quantifying field-saturated hydraulic conductivity (Ks ). Cerdá (1998), for example, used rainfall simulation experiments to measure a doubling of Ks from 25 mm/hr after 7 months of recovery following a fire in a Mediterranean scrubland to 52 mm/hr following 64

months of recovery. Robichaud et al. (2016) also report a substantial increase in Ks from 31 mm/hr within 1 month of the Valley Complex fire to 38 mm/hr after 11 months and 84 mm/hr after 60 months. Using a Kruskal-Wallis test and a significance level of 0.05, we find no statistically significant changes in the distribution of Ks with time following the Fish Fire. Still, the median and mean values of Ks were substantially greater 33 months after the fire relative to 3 months after the fire. Combined with increases in S over that same time period, potentially large implications for runoff generation may be underestimated by examining differences in S and Ks individually. Improved understanding of the timescales over which soil-hydraulic

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 43–56

Post-Fire Debris-Flow Initiation

Figure 6. Model simulations of ﬂow stage at the Arroyo Seco basin outlet in response to (a–c) the designed storm with peak I15 = 45 mm/hr and (d–f) a narrow cold-frontal rain band with peak I15 = 45 mm/hr under diﬀerent conditions.

properties recover following fire are crucial for quantifying when recently burned landscapes will regain hydrologic function (e.g., Ebel and Mirus, 2014), including when debris-flow potential will return to that of long-unburned areas. Measurements of soil-hydraulic properties and vegetation/ground cover can be used in conjunction with the numerical model to infer year 1 debris-flow thresholds of I15 = 15 mm/hr and I15 = 30 mm/hr for an NCFR and the designed storm, respectively (Figure 5). The regional I15 threshold of 19 mm/hr for debris flow initiation in the SGM (Staley et al., 2013) falls in between these two model-derived thresholds. We attribute the lower I15 threshold for NCFRs to greater rainfall intensities over shorter durations, as quantified by the relatively high I5 /I15 ratio relative to the designed storm, which was more effective at generating runoff. It is also possible that thresholds associated with NCFRs are lower due to the presence of roughly 50 minutes of low-intensity rainfall that precedes the peak (Figure 2). However, this is unlikely to be a dominant factor in post-fire year 1, when the differences in debris-flow thresholds between the two rainstorms are most accentuated. First, the modeled I15 threshold associated with NCFRs in year 1 is 15 mm/hr. The average rainfall intensity during the first 45 minutes of that storm is approximately 7 mm/hr, which is not sufficient to generate runoff or trans-

port sediment, and the total depth of rainfall during that time period would be less than 6 mm. Even if the first 45 minutes of rainfall were eliminated from all NCFRs, the I15 threshold in year 1 would increase from only 15 to 20 mm/hr. The observation that NCFRs tend to produce debris flows at a lower I15 relative to the designed storm suggests that particular attention should be given to debris-flow potential when atmospheric conditions are conducive to producing an NCFR. Further work is needed to quantify the ability of NCFRs to produce debris flows relative to other types of storm systems known to produce intense rainfall in southern California, such as isolated convective cells. Increases in the I15 threshold associated with NCFRs from 15 mm/hr in year 1 to 35 mm/hr in year 2 can be attributed primarily to decreases in the percentage of bare ground. We documented an increase in the wetting front suction head (hf ) from year 1 to year 2 (Table 1 and Figure 4), but the I15 threshold would still remain constant at 15 mm/hr in years 1 and 2 if vegetation recovery were completed neglected (Figure 5). The ID threshold associated with the designed storm, on the other hand, increased from 30 mm/hr in year 1 to 40 mm/hr in year 2 (Figure 5). Since the threshold would be 35 mm/hr in the case where vegetation recovery is neglected in year 2, approximately half of the increase from 30 to 40 mm/hr can be

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 43–56

McGuire, Rengers, Oakley, Kean, Staley, Tang, Orla-Barile, and Youberg

attributed to vegetation recovery. The remainder of the increase can be related to changes in soil-hydraulic properties, namely, the increase in hf that took place between 8 and 13 months after the wildfire. All else being equal, lower values of hf will lead to an increase in runoff. Decreases in percent bare ground will lead to less hillslope erosion, which subsequently decreases the amount of sediment transported into the channel network (where debris flows are likely to form) and reduces the sediment-bulking processes that can increase flow depths and discharges. Simulations suggest that the ability of NCFRs to produce debris flows is particularly sensitive to changes in vegetation/ground cover. Therefore, the initiation of debris flows via NCFRs at our study site appears to be more sensitive to changes in the efficiency of hillslope erosion relative to the designed storms. Simulations also offer insight into how debris-flow magnitude can be expected to change with rainfall intensity and time since burning for the two different storms. Simulations indicate that debris-flow volume generally increases monotonically with I15 for NCFRs (Figure 5). This is consistent with Gartner et al. (2014), who found that debris-flow volumes increase with I15 based on a large data set of estimated volumes from post-wildfire debris flows throughout the Transverse Ranges of southern California. In contrast, debris-flow volume mobilized by the designed storms does decrease slightly when I15 is greater than 45 mm/hr in years 1 and 2. The difference between our model results and field observations (e.g., Gartner et al, 2014) could be due partly to the definition of debris flow employed here, which requires that the sediment concentration exceed 40 percent. In some of the model scenarios, an increase in I15 leads to more runoff and an overall reduction in sediment concentration to values less than 40 percent. Since debris-flow volumes estimated in the field are based on the amount of sediment deposited in debris basins or estimates of erosion occurring during debris-flow–producing rainstorms (e.g., Gartner et al., 2014), they may include sediment transported through a combination of water-dominated flood, debris flood, and debris-flow mechanisms. Regardless, debris flows with the greatest volume, highest peak discharge, and largest flow depths are generally produced by NCFRs with high rainfall intensities (Figures 5 and 6). Simulations also indicate a rapid decrease in expected debris-flow volume between postfire years 1 and 2. Santi and Morandi (2013) analyzed post-wildfire debris-flow volumes from California and also found a substantial decrease in volume between debris flows generated within 1 year of a fire and those generated between 1 and 3 years after fire, with the median sediment yield from debris flows decreasing from 10,156 to 4,006 m3 /km2 .

While we focus on a particular geographic region, the SGM in southern California, the modeling framework presented here can be used in combination with estimates of post-wildfire infiltration rates from other regions (e.g., Moody et al., 2009; Nyman et al., 2011; and Robichaud et al., 2016) to quantify the impact of changing soil-hydraulic properties on debris-flow magnitude and initiation thresholds. Similarly, satellitederived metrics of vegetation recovery, such as the enhanced vegetation index (e.g., Kinoshita and Hogue, 2011), could be used to drive temporal changes in percent ground cover within the model framework. Developing relationships between measurable hydrologic variables, ground cover characteristics, and debris-flow properties is a necessary first step toward assessing how debris-flow threats are likely to evolve with time following wildfire in different geographic regions. The prevalence of NCFRs as a primary driver of highintensity precipitation is also somewhat specific to this region, but the results here suggest that the temporal distribution of rainfall within a storm can impact rainfall thresholds for runoff-generated debris flows. Here, we explore differences between only one storm type (NCFRs) and its associated hyetograph and a Gaussian distribution of rainfall. Additional work is needed to determine the difference in debris-flow characteristics across other storm types (and their associated hyetographs) that regularly impact the Transverse Ranges. One key benefit of assessing how debris-flow response varies as a function of storm type (e.g., NCFR versus a designed storm), is that it is has the potential to increase the situational awareness of weather forecasters concerned with debris-flow hazards. Precipitation forecasts become more uncertain as lead time increases (e.g., Wick et al., 2013), and it is difficult to accurately forecast the location, timing, and intensity of short-duration, high-intensity rainfall (Doswell et al., 1996) at lead times that are useful to emergency management (at least 24– 36 hours). However, forecasters can look for the presence or absence of a certain set of atmospheric conditions that they may associate with the potential for a particular type of high-intensity rainfall event (e.g., an NCFR). If the forecaster is aware of the debris-flow response associated with the typical hyetograph of that characteristic storm type, this would provide them with additional confidence about the potential for an impactful debris flow in the area of concern. CONCLUSIONS Disturbance following wildfire leads to an increased potential for runoff-generated debris flows. The hazards posed by runoff-generated debris flows decrease

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 43–56

Post-Fire Debris-Flow Initiation

with time following wildfire as soil and vegetation recover. In this study, we monitored changes in soilhydraulic properties and percent bare ground at a site in southern California and used a numerical model to determine how temporal changes in these two variables affect debris-flow volumes and initiation thresholds. The fraction of bare ground at our study area decreased from approximately 1.0 immediately following the fire to 0.63 after 21 months and to 0.2 after 33 months. Both sorptivity and wetting front suction head increased significantly after 13 months of recovery and generally increased with time following the wildfire. Simulations suggest that the threshold I15 rainfall intensity that triggers debris flows at our study site in the SGM is sensitive to how rainfall is distributed in time. The I15 threshold for an NCFR could be as low as 15 mm/hr in post-fire year 1 and 35 mm/hr in year 2. A designed rainstorm with a time series of rainfall distributed like a Gaussian function was responsible for producing debris flows only when I15 was greater than 30 mm/hr in year 1 and 40 mm/hr in year 2. Regardless of the temporal distribution of rainfall within a storm, if debris flows do initiate in the second post-wildfire year, simulations indicate that they will be roughly one-third to one-fourth the volume of those generated in year 1. Results demonstrate how debris-flow thresholds and magnitude can be sensitive to the time series of rainfall and identify the need to quantify how/why different types of storms may be more or less likely to produce impactful debris flows. Although we focus here on post-wildfire debris flows, the methodology used to assess changes in runoffgenerated debris-flow susceptibility could be applied in other settings, including rocky alpine regions, where runoff-generated debris flows may occur. ACKNOWLEDGMENTS This work was supported in part by the National Oceanic and Atmospheric Administration (NOAA) Collaborative Science, Technology, and Applied Research (CSTAR) Program under grant NA19NWS4680004 and by the U.S. Geological Survey (USGS) Landslide Hazards Program. Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. government. Code for the numerical model used in this study (SWEHR) is available through the Community Surface Dynamics Modeling System (CSDMS) model repository. REFERENCES Cannon, S. H.; Gartner, J. E.; Wilson, R. C.; Bowers, J. C.; and Laber, J. L., 2008, Storm rainfall conditions for floods and

debris flows from recently burned areas in southwestern Colorado and southern California: Geomorphology, Vol. 96, No. 3–4, pp. 250–269. Cannon, S. H.; Kirkham, R. M.; and Parise, M., 2001, Wildfirerelated debris flow initiation processes, Storm King Mountain, Colorado: Geomorphology, Vol. 39, pp. 171–188. Cerdá, A., 1998, Changes in overland flow and infiltration after a rangeland fire in a Mediterranean scrubland: Hydrological Processes, Vol. 12, pp. 1031–1042. Crocker, R. L. and Tiver, N. S., 1948, Survey methods in grassland ecology: Grass and Forage Science, Vol. 3, No. 1, pp. 1–26. DeBano, L. F., 2000, The role of fire and soil heating on water repellency in wildland environments: A review: Journal of Hydrology, Vol. 231, pp. 195–206. Doswell, C.A. III; Brooks, H. E.; and Maddox, R. A., 1996, Flash flood forecasting: An ingredients-based methodology: Weather and Forecasting, Vol. 11, No. 4, pp. 560–581. Ebel, B. A. and Mirus, B. B., 2014, Disturbance hydrology: Challenges and opportunities: Hydrological Processes, Vol. 28, No. 19, pp. 5140–5148. Ebel, B. A. and Moody, J. A., 2017, Synthesis of soil-hydraulic properties and infiltration timescales in wildfire-affected soils: Hydrological Processes, Vol. 31, No. 2, pp. 324–340. Gabet, E. J. and Bookter, A., 2008, A morphometric analysis of gullies scoured by post-fire progressively bulked debris flows in southwest Montana, USA: Geomorphology, Vol. 96, No. 3–4, pp. 298–309. Gartner, J. E.; Cannon, S. H.; and Santi, P. M., 2014, Empirical models for predicting volumes of sediment deposited by debris flows and sediment-laden floods in the transverse ranges of southern California: Engineering Geology, Vol. 176, pp. 45–56. Hairsine, P. B. and Rose, C. W., 1992a, Modeling water erosion due to overland flow using physical principles: 1. Sheet flow: Water Resources Research, Vol. 2, No. 1, pp. 237–243. Hairsine, P. B. and Rose, C. W., 1992b, Modeling water erosion due to overland flow using physical principles: 2. Rill flow: Water Resources Research, Vol. 28, No. 1, pp. 245–250. Iverson, R. M. and Denlinger, R. P., 2001, Flow of variably fluidized granular masses across three-dimensional terrain: 1. Coulomb mixture theory: Journal of Geophysical Research: Solid Earth, Vol. 106, No. B1, pp. 537–552. Jorgensen, D. P.; Pu, Z.; Persson, P. O.; and Tao, W., 2003, Variations associated with cores and gaps of a Pacific narrow cold frontal rainband: Monthly Weather Review, Vol. 131, pp. 2705–2729. Kean, J. W.; McCoy, S. W.; Tucker, G. E.; Staley, D. M.; and Coe, J. A., 2013, Runoff-generated debris flows: Observations and modeling of surge initiation, magnitude, and frequency: Journal of Geophysical Research: Earth Surface, Vol. 118, No. 4, pp. 2190–2207. Kean, J. W.; Staley, D. M.; and Cannon, S. H., 2011, In situ measurements of post-fire debris flows in southern California: Comparisons of the timing and magnitude of 24 debris-flow events with rainfall and soil moisture conditions: Journal of Geophysical Research, Vol. 116, F04019. doi:10.1029/2011JF002005. Kean, J. W.; Staley, D. M.; Lancaster, J. T.; Rengers, F. K.; Swanson, B. J.; Coe, J. A.; Hernandez, J. L.; Sigman, A. J.; Allstadt, K. E.; and Lindsay, D. N., 2019, Inundation, flow dynamics, and damage in the 9 January 2018 Montecito debris-flow event, California, USA: Opportunities and challenges for post-wildfire risk assessment: Geosphere, Vol. 15, No. 4, pp. 1140–1163.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 43–56

McGuire, Rengers, Oakley, Kean, Staley, Tang, Orla-Barile, and Youberg Kinoshita, A. M. and Hogue, T. S., 2011, Spatial and temporal controls on post-fire hydrologic recovery in Southern California watersheds: Catena, Vol. 87, No. 2, pp. 240–252. Larsen, I. J.; MacDonald, L. H.; Brown, E.; Rough, D.; Welsh, M. J.; Pietraszek, J. H.; Libohova, Z.; de Dios BenavidesSolorio, J.; and Schaffrath, K., 2009, Causes of postfire runoff and erosion: Water repellency, cover, or soil sealing?: Soil Science Society of America Journal, Vol. 73, No. 4, pp. 1393–1407. McGuire, L. A.; Kean, J. W.; Staley, D. M.; Rengers, F. K.; and Wasklewicz, T. A., 2016, Constraining the relative importance of raindrop- and flow-driven sediment transport mechanisms in post-wildfire environments and implications for recovery time scales: Journal of Geophysical Research: Earth Surface. doi:10.1002/2016JF003867. McGuire, L. A.; Rengers, F. K.; Kean, J. W.; and Staley, D. M., 2017, Debris flow initiation by runoff in a recently burned basin: Is grain-by-grain sediment bulking or en masse failure to blame?: Geophysical Research Letters, Vol. 44, No. 14, pp. 7310–7319. Meyer, G. A. and Wells, S. G., 1997, Fire-related sedimentation events on alluvial fans, Yellowstone National Park, USA: Journal of Sedimentary Research, Vol. 67, No. 5, pp. 776–791. Moody, J. A. and Ebel, B. A., 2012, Hyper-dry conditions provide new insights into the cause of extreme floods after wildfire: Catena, Vol. 93, pp. 58–63. Moody, J. A.; Kinner, D. A.; and Úbeda, X., 2009, Linking hydraulic properties of fire-affected soils to infiltration and water repellency: Journal of Hydrology, Vol. 379, No. 3–4, pp. 291–303. Moody, J. A.; Smith, J. D.; and Ragan, B. W., 2005, Critical shear stress for erosion of cohesive soils subjected to temperatures typical of wildfires: Journal of Geophysical Research: Earth Surface, Vol. 110, No. F01004. doi:10.1029/2004JF000141. Mügler, C.; Planchon, O.; Patin, J.; Weill, S.; Silvera, N.; Richard, P.; and Mouche, E., 2011, Comparison of roughness models to simulate overland flow and tracer transport experiments under simulated rainfall at plot scale: Journal of Hydrology, Vol. 402, No. 1–2, pp. 25–40. Nyman, P.; Sheridan, G. J.; Smith, H. G.; and Lane, P. N., 2011, Evidence of debris flow occurrence after wildfire in upland catchments of south-east Australia: Geomorphology, Vol. 125, No. 3, pp. 383–401. Oakley, N. S.; Cannon, F.; Munroe, R.; Lancaster, J. T.; Gomberg, D.; and Ralph, F. M., 2018, Brief communication: Meteorological and climatological conditions associated with the 9 January 2018 post-fire debris flows in Montecito and Carpinteria California, USA: Natural Hazards and Earth System Sciences, Vol. 18, pp. 3037–3043. Oakley, N. S.; Lancaster, J. T.; Kaplan, M. L.; and Ralph, F. M., 2017, Synoptic conditions associated with cool season post-fire debris flows in the Transverse Ranges of southern California: Natural Hazards, Vol. 88, pp. 327–354. Parson, A.; Robichaud, P. R.; Lewis, S. A.; Napper, C.; and Clark, J. T., 2010, Field Guide for Mapping Post-Fire Soil Burn Severity: U.S. Department of Agriculture Forest Service General Technical Report RMRS-GTR-243, 49 p. Raymond, C. A.; McGuire, L. A.; Youberg, A. M.; Staley, D. M.; and Kean, J. W., 2020, Thresholds for post-wildfire debris flows: Insights from the Pinal Fire, Arizona, USA: Earth Surface Processes and Landforms, Vol. 45, No. 6, pp. 1349– 1360.

Rengers, F. K.; McGuire, L. A.; Kean, J. W.; Staley, D. M.; and Hobley, D., 2016a, Model simulations of flood and debris flow timing in steep catchments after wildfire: Water Resources Research, Vol. 52, pp. 6041–6061. Rengers, F. K.; Tucker, G. E.; Moody, J. A.; and Ebel, B. A., 2016b, Illuminating wildfire erosion and deposition patterns with repeat terrestrial lidar: Journal of Geophysical Research: Earth Surface, Vol. 121, No. 3, pp. 588–608. Robichaud, P. R.; Wagenbrenner, J. W.; Pierson, F. B.; Spaeth, K. E.; Ashmun, L. E.; and Moffet, C. A., 2016, Infiltration and interrill erosion rates after a wildfire in western Montana, USA: Catena, Vol. 142, pp. 77–88. Santi, P. M. and Morandi, L., 2013, Comparison of debris-flow volumes from burned and unburned areas: Landslides, Vol. 10, No. 6, pp. 757–769. Shakesby, R. A. and Doerr, S. H., 2006, Wildfire as a hydrological and geomorphological agent: Earth-Science Reviews, Vol. 74, No. 3–4, pp. 269–307. Smith, H. G.; Sheridan, G. J.; Nyman, P.; Child, D. P.; Lane, P. N.; Hotchkis, M. A.; and Jacobsen, G. E., 2012, Quantifying sources of fine sediment supplied to post-fire debris flows using fallout radionuclide tracers: Geomorphology, Vol. 139, pp. 403–415. Staley D. M.; Kean, J. W.; Cannon, S. H.; Laber, J. L.; and Schmidt, K. M., 2013, Objective definition of rainfall intensity-duration thresholds for the initiation of postfire debris flows in southern California: Landslides, Vol. 10, pp. 547–562. Staley, D. M.; Negri, J. A.; Kean, J. W.; Laber, J. L.; Tillery, A. C.; and Youberg, A. M., 2017, Prediction of spatially explicit rainfall intensity–duration thresholds for post-fire debris-flow generation in the western United States: Geomorphology, Vol. 278, pp. 149–162. Staley, D. M.; Wasklewicz, T. A.; and Kean, J. W., 2014, Characterizing the primary material sources and dominant erosional processes for post-fire debris-flow initiation in a headwater basin using multi-temporal terrestrial laser scanning data: Geomorphology, Vol. 214, pp. 324–338. Stoof, C. R.; Vervoort, R. W.; Iwema, J.; Elsen, E.; Ferreira, A. J. D.; and Ritsema, C. J., 2012, Hydrological response of a small catchment burned by experimental fire: Hydrology and Earth System Sciences, Vol. 16, No. 2, pp. 267–285. Takahashi, T., 1978, Mechanical characteristics of debris flow: Journal of the Hydraulics Division of the American Society of Civil Engineers, Vol. 104, No. 8, pp. 1153–1169. Tang, H.; McGuire, L. A.; Rengers, F. K.; Kean, J. W.; Staley, D. M.; and Smith, J. B., 2019, Evolution of debris-flow initiation mechanisms and sediment sources during a sequence of post wildfire rainstorms: Journal of Geophysical Research: Earth Surface, Vol. 124, pp. 1572–1595. Vandervaere J. P.; Vauclin, M.; and Elrick, D. E., 2000, Transient flow from tension infiltrometers II. Four methods to determine sorptivity and conductivity: Soil Science Society of America Journal, Vol. 64, pp. 1272–1284. Wick, G. A.; Neiman, P. J.; Ralph, F. M.; and Hamill, T. M., 2013, Evaluation of forecasts of the water vapor signature of atmospheric rivers in operational numerical weather prediction models: Weather and Forecasting, Vol. 28, No. 6, pp. 1337–1352. Zhang, R., 1997, Determination of soil sorptivity and hydraulic conductivity from the disk infiltrometer: Soil Science Society of America Journal, Vol. 61, No. 4, pp. 1024–1030.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 43–56

Forecasting and Seismic Detection of Proglacial Debris Flows at Mount Rainier National Park, Washington, USA SCOTT R. BEASON* Mount Rainier National Park, U.S. National Park Service, 55210 238th Avenue E, Ashford, WA 98304

NICHOLAS T. LEGG Wolf Water Resources, 1001 SE Water Avenue, Suite 180, Portland, OR 97214

TAYLOR R. KENYON ROBERT P. JOST Mount Rainier National Park, U.S. National Park Service, 55210 238th Avenue E, Ashford, WA 98304

Key Terms: Debris Flows, Glacial Outburst Floods, Hazard Mitigation and Monitoring, Hazard Forecasting, Real-Time Seismic Amplitude Measurement, Mount Rainier

ployees and visitors working and recreating in the areas downstream. Our goal is to accurately forecast the debris-ﬂow hazards up to 7 days ahead of time and then use RSAM to detect debris ﬂows within minutes of their genesis.

ABSTRACT The glaciated Mount Rainier volcano in southwestern Washington State (United States) has a rich history of outburst floods and debris flows that have adversely impacted infrastructure at Mount Rainier National Park in the 20th and 21st centuries. Retreating glaciers leave behind vast amounts of unconsolidated till that is easily mobilized during high-precipitation-intensity storms in the fall months, and during outburst floods during warm summer months. Over 60 debris flows and outburst floods have been documented between 1926 and 2019 at Mount Rainier. Debris-flow activity has led to the closure of campgrounds and visitor destinations, which has limited visitor access to large swaths of the park. This paper documents efforts to characterize and seismically monitor debris flows, map hazards, and develop forecasting approaches for wet and dry weather debris flows. Using the day-of and historic antecedent weather conditions on past debris-flow days, we developed a debris-flow hazard model to help predict those days with a higher relative hazard for debris-flow activity park-wide based on prevailing and forecasted weather conditions. Debris flows are detected in near-real-time using the U.S. Geological Survey Real-time Seismic Amplitude Measurement (RSAM) tool. If an event is detected, we can then provide evacuation alerts to em-

*Corresponding author email: scott_beason@nps.gov

INTRODUCTION Mount Rainier (MORA) is a 4,392 m (14,410 ft) stratovolcano located in southwest Washington State, United States, approximately 70 km (43 mi) southeast of Tacoma and 90 km (56 mi) south-southeast of Seattle (Figure 1). The volcano occupies most of the 956 km2 (369 mi2 ) Mount Rainier National Park and is visible from much of western Washington State. MORA has been episodically active in the last 500,000 years, including at least 10 to 12 eruptions in the last 2,600 years (Sisson and Vallance, 2009). Eruptions have initiated large lahars that have inundated areas of the Puget Lowland as far as 100 km (62 mi) from the volcano (Crandell, 1971). Because of its far-reaching lahar hazards, MORA has a “very high” threat and ranks as the third most hazardous volcano in the nation (Ewert et al., 2018). In addition to these farreaching hazards, local-scale debris flows induced by hydrologic and surficial geomorphic processes represent a significant management concern on a more frequent and local basis within MORA park boundaries. Debris flows initiated during intra-eruptive periods at MORA are generally much smaller in magnitude and impact than the large lahars that have occurred during eruptive periods (Pierson and Scott, 1985; Vallance and Scott, 1997; and Vallance, 2005). This type of debris flow is initiated when surges of water

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 57–72

Beason, Legg, Kenyon, and Jost

Figure 1. Location map of the southwest side of Mount Rainier National Park in Washington State, United States. RER, LO2, and FMW refer to the Emerald Ridge, Longmire, and Mt. Fremont seismographs, respectively. Locations mentioned in text are shown on the detail map.

recruit available loose sediment and transform into rheologically denser slurries of sediment capable of moving large grain sizes (Scott et al., 1995). These surges originate from within a glacier, referred to as glacial outburst floods, or during periods of intense and prolonged precipitation. Debris flows of this type attenuate rapidly, and the deposits are often reworked by subsequent event runoff, leaving them nearly identical to overbank flood deposits. Sometimes, these debris flows often go unnoticed in remote reaches of the park. Understanding the initiation characteristics and thus cataloging all events at MORA are the prime motivating factors in the development of the real-time detection efforts, which was one of the main objectives of this study. The glaciers in MORA are one of the strongest controlling influences on the park landscape (Lescinsky and Sisson, 1998). MORA has 29 named glaciers, which cover a total of 78.76 ± 1.11 km2 (30.41 ± 0.43 mi2 ) and encompass a total volume of 3.22 ± 0.31 km3 (0.77 ± 0.07 mi3 ) as of 2015 (Beason, 2017; George and Beason, 2017). Studies show that the glacial ice on MORA has decreased in area by 39.1 percent from 1896 to 2015 (0.44 km2 yr−1 avg.) and in volume by 45 percent from 1896 to 2015 (0.02 km3 yr−1 ) (Driedger and Kennard, 1986; George and Beason, 2017). Glacial recession contributes to increases in

glacial melt runoff and, through mechanisms not yet understood, subglacial water storage, both of which have been observed to cause glacial outburst floods and many of the debris flows recorded in the park (Walder and Driedger, 1994a, 1994b, 1995). As such, quantification of the changes in these glaciers and the impacts of newly uncovered glacial sediment stockpiles must be considered if we are to understand the hazards discussed here. This paper documents concerted efforts by the MORA park geologists and a broader research community to monitor debris flows, assess and map hazards, and forecast debris flows to minimize risks to people and infrastructure. BRIEF HISTORY OF DEBRIS FLOWS AT MOUNT RAINIER Notable and well-documented debris flows in the park provide a reference point for the processes, initiation mechanisms, and impact of these events, helping to set the stage for the more technical study to follow. The first recorded debris flow in the park occurred in the Nisqually River watershed (Figure 1), on the south side of the park, on October 16, 1926. This event was initiated by the first heavy rain at the end of the summer season (Richardson, 1968). Prior to the event, it

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 57–72

Detection of Proglacial Debris Flows at Mount Rainier

was noted that there was 33 cm (13 in.) of snow at Paradise on October 13, all of which had melted by October 16. After this melt, a warm rain event brought in 9.9 cm (3.9 in.) of rain on the day of the 16th. Between 1932 and 1976, at least six additional outburst floods or debris flows occurred in the Nisqually River, originating from the Nisqually Glacier. Most of these events were induced by precipitation, which varied from 6 to 25 cm (2.4 to 9.9 in.). Four of the events occurred in October, and two occurred in June and July. On October 14, 1932, visiting engineers from the Bureau of Public Lands witnessed a precipitationinduced debris flow, described as “a wall of water 25 ft high and 125 ft wide” and “similar to a huge mixture of concrete except darker in color” (Richardson, 1968). The force of this event moved the entire old Nisqually Glacier Bridge over 0.8 km (0.5 mi) downstream from its original location. Some of the debrisflow events were well documented, including the October 25, 1955, and July 3, 1976, events (Richardson, 1968; Samora and Malver, 1996). An event in 1955 had six pulses in 45 minutes, an estimated velocity of 6.1 m s−1 , and a discharge of 2,000 m3 s−1 , and it was estimated to be 70 percent sediment by volume (Richardson, 1968). This event also led to the construction of the current Nisqually Glacier Bridge, a tall, channel-spanning structure that exists to this day. There are an additional five events cataloged in the Nisqually watershed during the park’s history, which behaved similarly to the events listed but were much smaller and had negligible impacts on park infrastructure. These five data points contain three glacial outbursts and two “other hydrologic events.” Two of the outburst floods are wet events that were preceded by notably intense rainfall in a short period beforehand, with the other being a dry weather event that took place in July. Of these, only the dry weather event was noted to have multiple surges. The “other hydrologic events” were noted for increases in stream stage, but these were not significant enough to cause any lasting damage to infrastructure or mobilize mass-wasting events. The most recent event recorded in the Nisqually River was a precipitation-initiated outburst flood on October 27, 2012 (Beason, 2012), which caused a 1 m (3 ft) increase in river stage at Longmire, approximately 7.9 km (4.9 mi) downstream of the glacier. 1947 Kautz Mudflow The largest recorded debris-flow event in the history of MORA is the 1947 Kautz Mudflow, which had an estimated volume of 3.8 × 107 m3 . In the 24 hours prior to the event, 15 cm (5.9 in.) of heavy rain and high freezing levels were seen in the

Kautz Creek watershed on the south-southwest side of the peak (Figure 1) (Driedger and Fountain, 1989). These conditions resulted in the collapse of the lower 1.6 km (1 mi) of the Kautz Glacier and a rapid release of water stored within the glacier (Scott et al., 1995). The surge of water entrained glacial outwash material, transforming into a clay-poor debris flow (Scott et al., 1995). Deposition of the Kautz Mudflow occurred over several days and included multiple pulses of water. Debris flows were noted in other drainages during this event, including in the Nisqually River. South Tahoma Glacier Activity Tahoma Creek and the South Tahoma Glacier, on the southwest side of the peak (Figure 1), have been a notable locus of debris-flow activity in the last half century, which largely began during the summer of 1967. The summer of 1967 was noted as exceptionally warm and dry. On August 29, a short-lived outburst flood destroyed a footbridge 1.9 km (1.2 mi) below the South Tahoma Glacier. The stream rose about 0.5 m (1.5 ft) at the Tahoma Creek Campground, approximately 5.6 km (3.5 mi) downstream of the glacier. Two days later, an outburst flood swept down Tahoma Creek (Richardson, 1968). Fortunately, the campground was already closed due to fire weather danger. Between 1967 and 2019, at least 33 distinct debrisflow events have originated from the South Tahoma Glacier and flowed down Tahoma Creek (Figure 2). Walder and Driedger (1994a) noted that the record for debris flows in Tahoma Creek does have some gaps, specifically between 1967 and 1985. This is due to poor record-keeping during this time. Crandell (1971, p. 60) noted that, “Floods not associated with rainfall also moved down the [Tahoma Creek] valley from time to time during the summer of 1968.” Walder and Driedger (1994a) noted that debris flows from the years of 1971 to 1985 are described “only sketchily” in park records. Debris flows that occurred between 1986 and 1992 are well documented, largely owing to increased awareness among National Park Service staff (Walder and Driedger, 1994a). The cumulative impact of over 30 debris flows in less than half a century and a major flood event in 2006 (Bullock et al., 2007) has had remarkable impacts to human infrastructure in the Tahoma Creek valley. The 24 km (15 mi) Westside Road was closed to vehicular traffic at mile post 3 in 1988. The sudden increase in debris-flow deposition forced the westward lateral migration and avulsion of Tahoma Creek, completely decimating an old-growth forest in the process (Figure 3). Portions of the Westside Road in Figure 3 have had to be repaired numerous times due to the combined ef-

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 57–72

Beason, Legg, Kenyon, and Jost

Figure 2. Seasonal timing of debris flows in Tahoma Creek from 1967 to 2019. Dry weather debris flows refer to those initiated by glacial outburst floods generally in the summer season, whereas wet weather debris flows generally occur in the fall and winter and are associated with intense precipitation.

Figure 3. Aerial photos from 1960, 2006, and 2015 showing the westward lateral migration of Tahoma Creek along the Westside Road due to debris-ﬂow activity (related to fan deposition).

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 57–72

Detection of Proglacial Debris Flows at Mount Rainier

Figure 4. Comparison of waveforms from (a) Emerald Ridge seismograph (RER), (b) real-time seismic amplitude measurement of the Emerald Ridge seismograph (RER RSAM), and (c) Tahoma Creek soundscape monitor during the August 13, 2015, debris-ﬂow sequence. RER and RER RSAM were computed at the same geographic location, whereas the soundscapes monitor was approximately 3.7 km (2.3 mi) downstream, which accounts for the lag in arrival times for that instrument. The green line in plot (c) is the 42 day background average of 42.04 dBA (dBA is based on the intensity of sound and how the human ear responds to that sound). Individual debris-ﬂow pulses are shown in the charts by the vertical gray columns and are numerically labeled as “DF X” based on their order in the sequence.

fects of debris flows and seasonal floods. The reduction in vehicular traffic and thus foot traffic on the Westside Road has led to a rapid and dramatic decrease in the recreational use of the trails and campgrounds on this side of the park since the late 1980s. Debris Flows in 2015—Direct Observations and Monitoring Results After a lull in debris-flow activity in the Tahoma Creek basin between 2006 and 2015, four separate debris-flow sequences occurred between 09:49 AM and 12:44 PM PDT (16:49–19:44 UTC) on August 13, 2015, during the park’s dry summer season. This event is the best-documented debris flow in the park’s history due to a network of flow, sound, and seismic instruments, which recorded the event, as well as direct witness observation of debris-flow pulses. Each individual sequence was identified in seismic records from the Emerald Ridge (RER) seismograph, located near Tahoma Creek (Figure 4; see location in Figure 1). Seismic monitors, a soundscape monitor, and stream gauges downstream all recorded data relevant to each

debris-flow surge, while numerous park visitors, volunteers, and employees all witnessed and photographed the event. Several visitors, including a geology professor at Pacific Lutheran University, recorded photos and videos of individual flows. A park volunteer in the upper Tahoma Creek basin accurately recorded and documented hyperconcentrated flow surges (not recorded on the seismograph) after the four debris flows, recording a total of 12 individual hyperconcentrated flows. The first debris flow issued by the South Tahoma Glacier was witnessed by visitor Croil Anderson. Anderson described the event as being “louder than a jet” at a distance of 2.5 km (1.5 mi) from the glacier (Anderson, 2015, personal communication). Anderson also stated that the first debris flow was an “incredibly large surge of black water, ice and rock” from the terminus. Claire Todd, a geology professor from Pacific Lutheran University, was on the Tahoma Creek suspension bridge as debris flows 2a and 2b moved down the watershed (Figure 4). When arriving at the bridge, she noted “a very high water/mud mark on [the] wall of [the] channel,” quickly followed by a “loud roar

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 57–72

Beason, Legg, Kenyon, and Jost

and terrific ground shaking” (Todd, 2015, personal communication). Continuing to observe the scene, she noted “a ∼1.5 m boulder is exposed in the channel” as the flow passes, and within another minute, “roar and shaking resumes, a second flow passes, just as thick as the first—completely obscuring the large boulder again.” Professor Todd witnessed the wave pass “exposing all of the large boulder again.” Last, she recorded “a thin flow of hyperconcentrated water is passing … and a view upstream shows another low wave of hyperconcentrated flow approaching,” noting that “these minor flows are not producing the roar or shaking that the first two offered.” The Emerald Ridge seismograph is located approximately 1 km (0.6 mi) from Tahoma Creek and accurately recorded the passage of each debris-flow surge. Using the seismic data as an input, after the event, we back-calculated the U.S. Geological Survey (USGS) Real-time Seismic Amplitude Measurement (RSAM) signature (Figure 4) (Endo and Murry, 1991). RSAM summarizes seismic activity for characterizing a volcano’s changing seismicity in real time. We used it to down sample the seismic signal to an average amplitude over a set time, in this case, 30 seconds. The combination of the seismic data and RSAM calculations (Figure 4b) shows the passage of each debris-flow surge clearly. One of the most interesting findings from the August 13 debris-flow sequence is the analysis of the soundscapes data in Figure 4c. The soundscapes monitor is a research effort by the National Park Service to understand the natural and unique soundscape of the park (National Park Service, 2018). Equipment emplaced along Tahoma Creek in 2015 fortuitously recorded the background noise in the months prior to and the day of the debris flow. The monitor recorded an anomalous decrease in river noise from the background level approximately 90 minutes before the arrival of the first debris-flow surge, which suggests creek blockage, filling of a temporary reservoir, and catastrophic failure as a debris flow. Each successive surge was recorded, and the river was relatively louder after the last debrisflow surge. This coincides with visual observations that the river was flowing much more vigorously after the event than before. Park staff became aware of the debris-flow event at 12:02 PM when park volunteer Yonit Yogev called the MORA dispatch center on the radio and reported an outburst flood at Tahoma Creek trailhead. Yogev described the event as “telltale sounds of a rumbling train, a huge amount/sounds of trees, and a huge amount of water coming over the road out of the creek bed” (Yogev, 2015, personal communication; National Park Service, 2015). A park visitor, Zachary Jones, videoed the passing debris-flow surge (DF 3 in

Figure 5. Simple matrix deﬁning conceptual framework for debrisﬂow hazard mapping approach.

Figure 4) with his cellphone, which provided visual evidence of the flow. Based on all observations and data observed from this event, we postulate that this event began as a physical blockage in the normal discharge of the glacier, perhaps as either a collapse of ice within the glacier or a small landslide just downstream of the glacier. This is evidenced by the anomalous and steady decrease in river noise from the soundscapes monitor just before 09:00 AM, showing that the total input to the river had dropped below the normal background level. DEBRIS-FLOW HAZARD MAPPING Mount Rainier’s flanks and draining watersheds have varying geomorphic and watershed characteristics that, in turn, lead to varying patterns in potential for debris generation (Legg et al., 2014). Therefore, systematic mapping of relative hazards around MORA’s flanks provides a critical tool in addressing the safety of park visitors as well as infrastructure management. The hazard mapping approach used here was based on a simple framework combining measures for debris-flow initiation potential and source sediment availability, with the basic idea that higher initiation potential leads to higher debris-flow likelihood, and higher sediment availability (abundance and proximity to drainage network) leads to greater debris-flow volume. This framework can be organized into a simple matrix showing varying combinations of debrisflow likelihood and volumetric potential as shown in Figure 5. To assess debris-generating potential (likelihood), we made use of Legg et al.’s (2014) geomorphic characterization of debris-flow initiation points mapped after a major storm in November 2006. More specifically, the mapping approach presented in this current

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 57–72

Detection of Proglacial Debris Flows at Mount Rainier

Figure 6. Plot of slope versus drainage area showing the slope–drainage area domain used to map gullies with high debris-flow initiation potential (based on data by Legg et al. [2014], who measured slope [S] and drainage area [A] within source gullies identified for the seven debris flows that initiated in the November 2006 storm). The lower domain bound visually bounds their sampled slopes and drainage area measurements and parallels their regression fit of gully head measurements (thin black line fit to crosses). The regression equation is S = 1.769 × A − 0.107, and the lower threshold used for hazard mapping is S = 1.40 × A − 0.107. A second domain boundary defines a minimum drainage area below which no debris-flow initiation points were mapped.

study made use of the gully slopes and drainage areas that Legg et al. (2014) measured at several initiation points of debris flows generated during a major storm in 2006. Based on their distribution of measured slopes and drainage areas of initiation points (Figure 6), we visually assigned slope and drainage area thresholds defining the domain of high debris-flow initiation potential (these thresholds are also shown in Figure 6). Using the slope–drainage area domain defined by our assigned thresholds, we then mapped high initiation potential segments of a drainage network across the full mountain, using a digitally generated drainage network generated from 1-m-spatial-resolution light detection and ranging (LiDAR) digital elevation models from 2009 (Figure 7). Next, drainage network segments identified within the critical slope–drainage area domain were filtered and classified based on a qualitative assessment of sediment availability (addressing the second component of the Figure 5 conceptual matrix). Gullies initially identified using slope–drainage area mapping were filtered out from additional analysis if they occurred in forested settings (debris flows generated in forested settings are typically small relative to those originating from volcanic flanks), bedrock-dominated landscapes (high volcanic flanks above roughly 2,750 m [9,000 ft] above sea level are dominated by rockfall transport and show a general lack of surficial debris due to steep slopes [Czuba et al., 2012]), and glacier surfaces lacking supra-glacial debris (also with negligible sediment availability on the surface). The re-

maining gullies were then assigned a qualitative rating for relative sediment availability of “low,” “medium,” and “high,” based on available surficial geologic maps (Crandell, 1969) and aerial photographs. In general, areas recently deglaciated and near glacier margins were assigned a high rating, areas near glacier margins with a mix of glacial debris and bedrock outcrops were assigned a moderate rating, and areas covered by older, more stable glacial debris (i.e., dating to the Last Glacial Maximum) and talus were assigned a low rating. To assess and map variations in debris-flow hazard around the mountain, the mapping results (combining likelihood and sediment availability) were summarized on a watershed basis. The relative hazard rating incorporated a simple sum of drainage network lengths meeting the slope–drainage area criteria, weighted by sediment availability classified at each drainage network segment. The simple rationale for length is that watersheds with a greater length of gullies exceeding slope–drainage area thresholds in highsediment settings are anticipated to have a higher likelihood of debris-flow production. Specifically, the lengths of network segments exceeding slope–drainage area thresholds and located within areas of “high” sediment availability were weighted 100 percent. Lengths of above-threshold segments with “medium” sediment availability were weighted at 50 percent. Low-sediment-availability segments were weighted at 0 percent (on the basis that these debris flows would be small and unlikely to impact infrastructure, even

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 57–72

Beason, Legg, Kenyon, and Jost

Figure 7. Map showing watersheds of major streams and rivers draining Mount Rainier and their debris-ﬂow hazard ratings resulting from elevation distribution of high-hazard gullies mapped in the hazard assessment. High-hazard gullies are those with high initiation potential within mapped areas of high sediment availability. The indicated elevations were used to divide storm classes. The freezing level shown suggests that rain (as opposed to snow) is likely falling on approximately 95 percent of high-hazard gullies. The freezing level was calculated based on a measured temperature of 4.4°C (40°F) at the SNOTEL station and an assumed 5.5°C per vertical kilometer lapse rate. Weighted summation of high-hazard gullies was identiﬁed based on their slope, drainage area, and sediment availability. The density of hazard features within each designated watershed determined its hazard.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 57–72

Detection of Proglacial Debris Flows at Mount Rainier

if initiated). The weighted sum of lengths resulted in a relative hazard rating to produce watershed-based hazard maps (incorporating both initiation potential and debris-flow volume) for the volcano, as shown in Figure 7. The watersheds with the 10 greatest hazard ratings, starting with the greatest, were the: (1) South Mowich River, (2) Kautz Creek, (3) Nisqually River, (4) Van Trump Creek, (5) Muddy Fork Cowlitz River, (6) South Tahoma Creek, (7) Ohanapecosh River, (8) North Mowich River, (9) White River (draining the Emmons Glacier), and (10) West Fork White River (draining the eastern margin of the Winthrop Glacier). DEBRIS-FLOW HAZARD FORECASTING The impetus for generating the debris-flow hazard forecast was to avoid having park staff and visitors in debris-flow-prone areas when events were likely to occur, like those conditions seen on August 13, 2015. The debris-flow hazard forecasting approach at MORA was based on two separate models combined, which incorporated different variables for dry, warm weather debris flows and cool, wet weather debris flows. The full model is shown in Appendix A. Cool, Wet Weather Debris Flows In recent decades, warm rainstorms occurring with low snowpack have been anecdotally associated with debris flows on MORA. These storm and debrisflow events typically occur in late fall, when atmospheric river storms bring intense tropical moisture from mid-latitudes and drop voluminous rain high on the volcanic flanks (Neiman et al., 2008). Prior to this study, there had been no systematic characterizations of debris-flow occurrence with respect to meteorological and antecedent hydrologic conditions. In practice, such a characterization could be paired with weather forecasts and in situ monitoring to forecast debris-flow hazards. This specific phase of our study focused on characterization of past storms and their associated debris-flow potential to develop a forecasting method for wet weather debris flows. Debris-flow events recorded since 1980 were compiled from multiple sources and included in our analysis if (1) the debris flow’s timing was known within a day, (2) it was associated with measurable precipitation, and (3) it occurred within the monitoring record of the snow telemetry (SNOTEL) station at Paradise (Natural Resources Conservation Service Site 679, elevation 1,640 m [5,381 ft]) on the southern flank of MORA. Debris-flow records included those from Walder and Driedger (1994a, 1994b, 1995), Driedger and Fountain (1989), and Copeland (2009). The SNO-

Figure 8. Elevation distribution of gullies mapped with high debrisflow hazard. High-hazard gullies are those with high initiation potential (based on slope and drainage areas) within mapped areas of high sediment availability. The indicated elevations and corresponding freezing levels were used as classification boundaries in the wet weather debris-flow forecasting method. Freezing levels were calculated based on recorded temperatures at the SNOTEL station and an assumed lapse rate of 5.5°C per vertical kilometer.

TEL station lies at the lower elevation range (see Figure 8) of high-hazard debris-flow gullies mapped in the effort discussed in the “Debris-Flow Hazard Mapping” section above. Data recorded at the station were therefore well suited for characterizing precipitation, temperature, and antecedent snowpack conditions within the elevation range of likely debris-flow initiation. For each debris-flow event, precipitation, temperature, and snowpack measurements were compiled for 1, 3, and 15 day periods on the day of and the day prior to the debris flow. These metrics were also compiled for all monthly maximum precipitation events for the full SNOTEL record to compare the known debris-flow-producing storms to the broader population of storms. This data compilation effort resulted in a total of 11 debris-flow-producing storms that occurred between 1979 and 2014. All 11 storms had daily average temperatures (recorded at the SNOTEL station) above freezing, and all but two events had daily average temperatures above 4.4°C (40°F). Based on the elevation range of mapped high-hazard gullies (Figure 8) and an assumption of vertical lapse rates, these temperatures suggest that all storms were dropping rain (as opposed to snow) on a significant portion of the volcanic flanks most likely to initiate debris flows. A temperature of 4.4°C (40°F) at the SNOTEL, in particular, indicates temperatures above freezing for the full elevation band of high-hazard gullies, suggesting high potential for runoff generation in the zone of likely debris-flow initiation. All debris-flow-producing storms also had limited antecedent snowpack (only two exceeded 2.5 cm [1 in.] snow water equivalent (SWE) on the day of

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 57–72

Beason, Legg, Kenyon, and Jost

the debris flow), therefore suggesting antecedent snowpack inhibits debris-flow generation. Potential mechanisms for this effect may include the ability of the snowpack to inhibit runoff generation and/or stabilize surficial colluvium. Additionally, there were negligible reductions in snowpack in the 3 days leading up to the 11 debris-flow events, suggesting snowmeltderived runoff as an unlikely ingredient for debris-flow generation. Precipitation quantities were further compared to an intensity-duration threshold for the nearby Seattle, WA, area developed by Chleborad et al. (2006), which was based on 3 day and 15 day cumulative precipitation totals in inches. The principals in play for this system are the assumptions that debris flows can be initiated by either heavy short-term storm inputs or longer-term saturation conditions, but also that both conditions can and often do occur in tandem. Initial comparisons to this landslide threshold found that eight of the 11 known debris-flow-producing storms exceeded the Seattle threshold; however, 247 of 376 monthly maximum storms (which had no recorded debris flows) from 1979 to 2014 also plotted above the threshold. The high number of non-debris-flowproducing storms plotting above the threshold suggests that it alone is a poor discriminator of debris-flow potential on MORA. To explore potential refinements to the predictive model, we then filtered out monthly maximum storms from the analysis in a stepwise fashion based on their temperatures and antecedent snowpacks. The first step filtered out storms with greater than 12.7 cm (5 in.) SWE and/or 3 day average temperatures less than freezing. In the remaining group, 33 monthly maximum (non-debris-flow-producing) storms exceeded the Seattle threshold, in addition to the eight debrisflow-producing storms. These numbers indicate approximately 20 percent (8 of 41) of these storms (above-threshold storms with above-freezing temperatures and low snowpacks) generated debris flows. The next filter applied a temperature threshold of 4.4°C (40°F) instead of 0°C (32°F) and revealed that 5 of 14 storms (36 percent) exceeding the Seattle threshold produced debris flows. The increased proportion of debris-flow-producing storms indicates that warm temperatures (i.e., high freezing levels) are indeed a requirement for debris-flow generation. Overall, these results highlight the need for temperature and snowpack information to be coupled with precipitation thresholds in order to increase predictive capability of our wet debris-flow model. The above analysis allowed us to develop a simple decision-tree approach to hazard classification as a planning tool for MORA (Legg, 2015). The approach uses 3 day precipitation and temperature forecasts in

concert with measurements of SWE and 15 day precipitation totals to classify and forecast debris-flow hazards into low, medium, and high hazard categories over a coming 3 day period. More broadly, this effort represents an example of hazard forecasting in an alpine setting where seasonal temperature and snow fluctuations are major drivers of debris-flow potential.

Dry, Warm Weather Debris Flows The method for forecasting dry weather debris flows is an expansion of Legg’s (2015) model, which accounts for weather conditions that produce rapid glacier melt and potential for outburst floods that release and transform to debris flows. In total, 35 debrisflow events that occurred in a dry season (i.e., no rain and relatively warm temperatures, with average high temperature of approximately 18.3°C [65°F]) were compiled from the various sources mentioned in the previous section. From that list, antecedent weather information for the day of the event and the days leading up to the event itself was compiled from the Paradise SNOTEL station and other weather sources in the vicinity. A Monte Carlo analysis was completed on each weather variable to determine its relative importance to the overall detection of a debris flow. Once the relative weighting of each variable was completed, all days in the historic record were run through the model to determine the debris-flow hazard scores on those days (this included the wet weather debris flows) (Table 1). The specific variables of interest for the dry side of the model were: (1) P18 , or 18 day precipitation total at Paradise, which was necessary to determine whether to run the dry side or wet side of the model; (2) Tmax , which is the maximum daily temperature observed at Paradise; (3) Tmax Percentile , which is the maximum temperature expressed as a percentile based on the historic maximum temperatures (1917–2017); (4) DS0SP, which is “days since zero snowpack,” a relative variable used to determine when debris source areas will likely be snow-free; (5) DD32 18 , or the 18 day cumulative degree days above freezing; (6) P3 , or the 3 day precipitation total, a key dry weather variable defined by Walder and Driedger (1994a); and (7) SWE, or snow water equivalent. For the model, DS0SP was set as days since July 11th, which is the average “melt out” date at Paradise in the historic (1917– 2017) record. The model uses known conditions to determine which type of condition is being predicted and therefore which equations to calculate. Each variable is then given a numeric score between 1 and 5 (see Appendix A), and the debris-flow hazard score is calculated (added) by the model.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 57–72

Detection of Proglacial Debris Flows at Mount Rainier Table 1. Performance of current debris-flow hazard model based on all available weather data for the period of 1917–2017 at the Paradise SNOTEL station. Event type categories are split into known debris-flow and outburst flood days from the historic record. The undefined category means that weather conditions were not available to adequately calculate the debris-flow hazard score for that day. This table could not be assessed for relative effectiveness because all included counts are strictly historic events, and no modern events had yet occurred. Event Type

Model Type

Debris ﬂow (N = 42)

Wet Dry Total Wet Dry Total Wet Dry Total Wet Dry Total

Outburst flood (N = 8) Debris flow + outburst flood (N = 50) No debris flow or outburst flood (N = 31,647) Total, no. (%):

Low

Medium

Medium High

High

Very High

Undeﬁned

0 3 3 2 3 5 2 6 8 12,633 11,608 24,241 24,249 (76.50)

0 4 4 0 1 1 0 5 5 980 984 1,964 1,969 (6.21)

1 6 7 0 0 0 1 6 7 539 618 1,157 1,164 (3.67)

12 11 23 0 1 1 12 12 24 1083 1001 2,084 2,108 (6.65)

— 5 5 — 1 1 — 6 6 — 540 540 546 (1.72)

0 0 0 0 0 0 0 0 0 942 719 1,661 1,661 (5.24)

At this time, the method is still being refined as more data are uncovered about the antecedent weather conditions and as more debris flows occur in the park. Additionally, an improved Monte Carlo approach is being undertaken to improve the model. The performance of the model for all available dates between 1917 and 2017 is shown in Table 1. There have been only two debris-flow events since the system was designed and implemented, meaning that all other predictions of moderate to very high have been false positives. In general, those days with a debris flow or outburst flood from the historic record should have a higher score, whereas those days with no event should have a lower score for the model to be considered truly calibrated successfully. Combination Forecast and Data Sources The combination forecast (Appendix A) uses both the wet and dry sides to create a decision tree based on calculated weather factors. Weather information is downloaded every hour from the DarkSky.net Application Programming Interface (API). DarkSky provides a free ensemble forecast for individual locations throughout the park that is easily incorporated into the debris-flow hazard model. It should be noted that free access to this API is ending in 2021, and a new weather source is being sought for this model. Every 4 hours, these weather variables and antecedent weather observations are automatically compiled based on the wet or dry forecast and then run through the decision tree algorithm (Appendix A). A qualitative score (low, medium, medium high, high, or very high) is generated for the day of interest and the next 7 days. This is then reported on a website for monitoring and decisionbased analysis by park staff. Hazard scores are tied to

weather forecasts and will change as forecasts are updated. While this process is automated, park staff still must monitor the model every day to determine the future relative risk for debris-flow activity. REAL-TIME DEBRIS-FLOW MONITORING VIA RSAM SYSTEM The final piece in the debris-flow hazard system at MORA is the ability to detect debris flows as they occur. As shown in the “Debris Flows in 2015” section, debris flows like those in 2015 have a seismic and RSAM signature that is distinctive. With assistance from the University of Washington’s Pacific Northwest Seismic Network (UW PNSN), seismic data are run through the USGS RSAM program and binned into 30 second values. At 5 minute intervals, an automated computer script then downloads the RSAM values and runs through the data file looking for a “debris-flow-like signature.” For the purposes of identification via the RSAM system, a debris-flow signature is roughly defined as an increasing signal above a set point over a set amount of time. If these values are exceeded, an alert is sent out to park staff for analysis and hazard notification via cellphone text messages and emails. As an example, at the RER seismograph, the relevant variables are an RSAM value greater than 500 counts for over 5 minutes with an RSAM value that is increasing (slope > 0.030), on average, over those 5 minutes. Using this definition, three of the four debris flows on August 13, 2015 (2a/2b, 3, and 4), and an additional debris flow that occurred in Tahoma Creek on September 12, 2015 (not discussed in this paper), would have been detected with this system. Additionally, this system would have detected the second debris

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 57–72

Beason, Legg, Kenyon, and Jost

flow on August 13th at roughly 10:20 AM, almost a full 2 hours before park staff were alerted to the event by witnessing it in person and reporting it on the radio. The goal would be to use this advance warning to notify park personnel and initiate emergency procedures, but the system had yet to be in effect prior to 2017 for an event that significantly impacts park infrastructure or use areas. Real-time debris-flow monitoring via the RSAM system is currently being run on the RER seismograph (Puyallup, Tahoma, and South Tahoma Glaciers), Mt. Fremont (FMW) seismograph (Emmons, Inter, and Winthrop Glaciers), and Longmire (LO2/LON) seismograph (Kautz, Nisqually, Pyramid, Success, Van Trump, and Wilson Glaciers). Most of the major glacial streams at MORA now have some sort of seismic monitoring; those without, with the exception of Carbon Glacier, do not have extensive infrastructure development in their watershed boundaries. The overall performance value of the RSAM system in detection of debris flows is being developed and has been improved by debris-flow events that occurred in 2019 (discussed in the next section). There have been several false-positive readings, almost exclusively due to wind noise (especially at RER). Local, regional, and teleseism earthquake events are such short-period and punctuated events that they are excluded in the analysis and rarely generate alerts. When false positives have been detected, staff are able to quickly analyze real-time seismic data to determine if the event is truly a debris flow or some other event. In this sense, the system is semi-automated and still requires human intervention in order to take the step from an alert generation to an alert being broadcast to the field. Last, we are not yet able to collocate exact drainages where a debris flow has occurred due to a paucity of seismic stations. However, a strong signal in one seismograph and relatively weak signals in others (as was the case in the August 2015 event) can help to determine a narrower geographic location for the event. DEBRIS FLOWS IN 2019—MODEL CALIBRATION, VERIFICATION, AND VALIDATION, AND FUTURE WORK A debris flow that occurred on August 5, 2019, in Tahoma Creek, between approximately 6:44 and 8:10 PM PDT (2019-08-06 01:44–03:10 UTC), provides validation information for the debris-flow forecasting and detection systems discussed here. Aerial reconnaissance on August 6 determined that the event originated from a sudden and significant change in the primary outlet stream from the terminus of the South Tahoma Glacier, which resulted in a surge of water within the

Figure 9. Terminus of South Tahoma Glacier on August 6, 2019, at approximately 2:48 PM PDT. The terminus is the darker band horizontally in the upper part of the photo. The outlet stream of the glacier had been on the left-hand side (river right) of the bedrock knob at the center of the photo and switched to the right side (river left) during the event, which incorporated loose, unstable ground moraine material and debris-covered stagnant ice downstream of the glacier and helped in debris-ﬂow generation. (Photo: NPS/Scott Beason).

glacier incorporating a surfeit of sediment in proglacial areas downstream of the glacier (Figure 9). Four separate surges were observed in the RER seismograph, with each event lasting 20.8, 10.4, 15.1, and 39.1 minutes, respectively. The fourth surge had the strongest seismic signal, followed by the second, third, and first surge, respectively. The qualitative debris-flow forecast for the 3 day period between August 4 and 6 was very high, high, and high, respectively. The forecast during this period was using the “dry” side of the algorithm and had inputs and outputs as shown in Table 2a. Unfortunately, the RSAM system did not detect any of the surges in this event; it nearly detected three of the surges, but critical exceedance thresholds were not met. In light of this, the critical threshold for debris-flow detection of RSAM counts from the RER system was re-calibrated from 500 to 400. This calibration proved fortuitous in detecting a second debris-flow sequence about a month later on September 26, 2019. On this day, a single debris-flow sequence occurred between 5:46 and 5:59 PM PDT (2019-09-27 00:46–00:59 UTC). The now-calibrated debris-flow detection system detected the event at 5:49:15 PM PDT (00:49:15 UTC). During the day prior to the event and after this debris flow, numerous RSAM alerts were generated due to wind noise. Upon detection of this event (Figure 10), a field investigation was undertaken within an hour of the event itself, and in-field evidence indicated the presence of a

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 57–72

Detection of Proglacial Debris Flows at Mount Rainier Table 2. Input and output variables for debris-ﬂow forecast day prior to, of, and after each of the August 5 and September 26, 2019, debris ﬂows in Tahoma Creek. See Appendix A for model information and variable names. Tmin P15

Date (a) August 5, 2019, Tahoma Creek debris ﬂow Aug 4

56.15 0.03 Aug 5 60.57 0.21 Aug 6 62.23 0.20 (b) September 26, 2019, Tahoma Creek debris ﬂow Sep 25 37.55 5.76 Sep 26 41.36 5.94 Sep 27 28.10 4.99

Tavg P18

Tmax SD

Tmax perc SWE

DS0SP TA

DD32 TB

DD32 18 HSDry

P1 Forecast Type

P3 Score

65.62 0.23 69.08 0.21 70.16 0.20

75.10 0 77.60 0 78.10 0

0.884 0.00 0.912 0.00 0.936 0.00

23 False 24 False 25 False

43 False 46 False 46 False

583 110.0 612 60.0 637 60.0

0.00 Dry 0.00 Dry 0.00 Dry

0.20 VH 0.00 H 0.00 H

46.06 6.20 44.94 6.76 34.43 5.80

54.58 0 48.52 0 40.76 0

0.433 0.00 0.268 0.00 0.120 0.00

75 True 76 True 77 True

23 False 17 True 9 False

309 40.0 308 60.0 301 52.5

0.00 Wet 0.73 Wet 0.08 Wet

0.44 MH 0.82 H 0.81 M

SD = Snow Depth (inches), HSDry = Hazard Score for the Dry Side of the Model, VH = Very High, H = High and MH = Medium High

debris ﬂow (new boulder levees, signiﬁcant in-stream sediment motion, rapid water rise, rapid increase in turbidity, and a distinctive smell of sediment in the air). Viewing the raw seismic signal along with seismic spectra and RSAM information in a single plot is ex-

tremely useful in visually excluding or including events as possible debris ﬂows. Table 2b shows the forecast inputs and outputs for the event. It should be noted that the September event was using the “wet” side of the forecast algorithm and had the highest possible hazard

Figure 10. (a) Raw RER seismic trace, (b) RER RSAM signature, and (c) RER spectra of the September 26, 2019, debris-ﬂow detection on the Emerald Ridge seismograph, including the time of detection of the event (vertical red line at 17:49:15 [5:49:15 PM] PDT; 00:49:15 UTC). Horizontal red line on RSAM plot (b) is the alert detection threshold (400 counts).

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 57–72

Beason, Legg, Kenyon, and Jost

category. The September 26, 2019, debris-flow event was the first successful automated detection of a debris flow in the park. It is interesting to note that the 2015 and 2019 events had a similar timing and pattern to the signal made by their debris-flow surges (i.e., an August event that had four surges and a September event that had a single surge). Both the debris-flow forecast and RSAM detection systems continue to be refined to improve the overall efficacy of the system. Additional debris-flow activity across the park, especially in those drainages that are monitored by the systems, will only improve the detection of debris flows over time. In-process proposals by the USGS to upgrade the seismic systems in the park, including the installation of infrasound sensors and additional seismometers, will only provide better data to help detect future debris-flow events. CONCLUSIONS Mount Rainier is an environment that is ideally suited for debris-flow genesis and has a rich history of these events. With our work, we have been successful in providing a forecast for debris-flow hazard based on antecedent weather conditions up to 7 days in advance of debris-flow days. We then can then detect individual debris flows using in situ seismometers and the RSAM system. As glaciers continue to retreat, new sediment sources will be exposed to annual storms and occasional outburst floods, all of which will contribute to the threat of debris flows in downstream areas. The forecasting and detection systems we have in place now are in their infancy and will be further refined as more events occur. Additional seismic installations planned in the next decade at MORA will only improve these systems and will provide better warning to park staff and visitors working and recreating at the park. Finally, the insights we are gaining in understanding debris-flow genesis and detecting debris flows at Mount Rainier could prove to be useful for similar efforts in analogous locales in the Pacific Northwest and across the world. ACKNOWLEDGMENTS This work is greatly indebted to the following individuals for their observations, photos, videos, and assistance during and after the 2015 and 2019 debris flows, including, but not limited to: Kate Allstadt, Anthony “Scott” Anderson, Croil Anderson, Jenni Chan, Carolyn Driedger, Maxine Dunkleman, Terry Flower, Jeff Gardner, Sara Hall, Mitch Haynes, Steve Hughes, Zachary Jones, Dave Keltner, Paul Kennard, Glenn Kessler, Rebecca Lofgren, Steve Malone,

Kendra Martinez, Abigail Michel, Seth Moran, Dave Morgan, Casey Overturf, Heather Rogers, Heather Sharp, Kurt Spicer, Trisha Stanfield, Karen Thompson, Claire Todd, Dave Turner, and Yonit Yogev. Finally, we wish to acknowledge the three anonymous reviewers who provided edits and assessment of the manuscript after submission. REFERENCES Anderson, C., 2015, personal communication, Mount Rainier National Park, 55210 238th Avenue E, Ashford, WA 98304. Beason, S. R., 2017, Change in Glacial Extent at Mount Rainier National Park from 1896–2015: National Park Service Natural Resource Report NPS/MORA/NRR—2017/1472, 98 p. Beason, S. R., 2012, Small Glacial Outburst Flood Occurs on Mount Rainier—October 27, 2012: Science Brief, National Park Service, Mount Rainier National Park, Ashford, WA, 3 p. Bullock, A. B.; Bacher, K.; Baum, J.; Bickley, T.; and Taylor, L., 2007, The Flood of 2006: 2007 Update: Unpublished report, Mount Rainier National Park, Ashford, WA, 43 p. Chleborad, A. F.; Baum, R. L.; and Godt, J. W., 2006, Rainfall Thresholds for Forecasting Landslides in Seattle, Washington, Area—Exceedances and Probability: U.S. Geological Survey Open-File Report 2006-1064, 31 p. doi:10.3133/ofr20061064. Copeland, E. A., 2009, Recent Periglacial Debris Flows from Mount Rainier, Washington: Unpublished M.S. Thesis, Water Resources Engineering, Oregon State University, Corvallis, OR, 139 p. Crandell, D. R., 1969, Surficial Geology of Mount Rainier National Park, Washington: U.S. Geological Survey Bulletin 1288, 41 p. doi:10.3133/b1288. Crandell, D. R., 1971, Postglacial lahars from Mount Rainier Volcano, Washington: U.S. Geological Survey Professional Paper 677, 75 p. doi:10.313/pp677. Czuba, J. A.; Magirl, C. S.; Czuba, C. R.; Curran, C. A.; Johnson, K. H.; Olsen, T. D.; Kimball, H. K.; and Gish, C. C., 2012, Geomorphic Analysis of the River Response to Sedimentation Downstream of Mount Rainier, Washington: U.S. Geological Survey Open-File Report 2012-1242, 134 p. doi:10.3133/ofr20121242. Driedger, C. L. and Fountain, A. G., 1989, Glacier outburst floods at Mount Rainier, Washington State, U.S.A.: Annals of Glaciology, Vol. 13, pp. 51–55. doi:10.3189/S0260305500007631. Driedger, C. L. and Kennard, P. M., 1986, Ice Volumes on Cascade Volcanoes: Mount Rainier, Mount Hood, Three Sisters, and Mount Shasta: U.S. Geological Survey Professional Paper 1365, 38 p. doi:10.3133/pp1365 Endo, E. T. and Murry, T. L., 1991, Real-time seismic amplitude measurement (RSAM): A volcano monitoring and prediction tool: Bulletin of Volcanology, Vol. 53, No. 7, pp. 533– 545. doi:10.1007/BF00298154. Ewert, J. W.; Difenbach, A. K.; and Ramsey, D. W., 2018, 2018 Update to the U.S. Geological Survey National Volcanic Threat Assessment: U.S. Geological Survey Scientific Investigations Report 2018-5140, 40 p. doi:10.3133/sir20185140. George, J. L. and Beason, S. R., 2017, Dramatic changes to glacial volume and extent since the late 19th century at Mount Rainier National Park, Washington, USA: Geological Society of America Abstracts with Programs, Vol. 49, No. 6, pp. 299694. doi:10.1130/abs/2017AM-299694.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 57–72

Detection of Proglacial Debris Flows at Mount Rainier Legg, N. T., 2015, An Assessment of Hazards from Rain-Induced Debris Flows on Mount Rainier: Unpublished internal document, Mount Rainier National Park, Ashford, WA, 30 p. Legg, N. T.; Meigs, A. J.; Grant, G. E.; and Kennard, P. M., 2014, Debris flow initiation in proglacial gullies on Mount Rainier, Washington: Geomorphology, Vol. 226, pp. 249–260. doi:10.1016/j.geomorph.2014.08.003. Lescinsky, D. T. and T. W. Sisson, 1998, Ridge-forming, icebounded lava flows at Mount Rainier, Washington: Geology, Vol. 26, No. 4, pp. 351–354. doi:10.1130/00917613(1998)026<0351:RFIBLF>2.3.CO;2. National Park Service, 2018, Mount Rainier Soundscapes: Electronic document, available at https://www.nps.gov/ mora/learn/nature/soundscapes.htm National Park Service, 2015, NPS Dispatch Records for 13 August 2015: Unpublished internal document, Mount Rainier National Park, Ashford, WA. Neiman, P. J.; Ralph, F. M.; Wick, G. A.; Kuo, Y. H.; Wee, T. K.; Ma, Z.; Taylor, G. H.; and Dettinger, M. D., 2008, Diagnosis of an intense atmospheric river impacting the Pacific Northwest: Storm summary and offshore vertical structure observed with COSMIC satellite retrievals: American Meteorological Society Monthly Weather Review, Vol. 136, No. 11, pp. 4398–4420. doi:10.1175/2008MWR2550.1. Pierson, T. C. and Scott, K. M., 1985, Downstream dilution of a lahar: Transition from debris flow to hyperconcentrated streamflow: Water Resources Research, Vol. 21, pp. 1511–1524. doi:10.1029/WR021i010p01511. Richardson, D., 1968, Glacier Outburst Floods in the Pacific Northwest: U.S. Geological Survey Professional Paper 600-D, pp. 79–86. Samora, B. A. and Malver, A., 1996, Inventory of Information on Glaciers in Mount Rainier National Park: Unpublished report, Mount Rainier National Park, Ashford, WA, 417 p. Scott, K. M.; Vallance, J. W.; and Pringle, P. T., 1995, Sedimentology, Behavior, and Hazards of Debris

Flows at Mount Rainier, Washington: U.S. Geological Survey Professional Paper 1547, 56 p. doi:10.3133/ pp1547. Sisson, T. W. and Vallance, J. W., 2009, Frequent eruptions of Mount Rainier over the last ∼2,600 years: Bulletin of Volcanology, Vol. 71, No. 6, p. 595–618. doi:10.1007/s00445-008-02457. Todd, C., 2015, personal communication, Pacific Lutheran University, Department of Geosciences, Rieke Science Center, Room 158, Tacoma, WA 98447. Vallance, J. W., 2005, Volcanic debris flows. In Jakob, M. and Hungr, O. (Editors), Debris-Flow Hazards and Related Phenomena: Springer Praxis Books, Springer, Berlin, pp. 247–274. doi:10.1007/b138657. Vallance, J. W. and Scott, K. M., 1997, The Osceola Mudflow from Mount Rainier: Sedimentology and hazard implications of a huge clay-rich debris flow: Geological Society of America Bulletin, Vol. 109, No. 2, pp. 143–163. doi:10.1130/00167606(1997)109<0143:TOMFMR>2.3.CO;2. Walder, J. S. and Driedger, C. L., 1994a, Geomorphic Change Caused by Outburst Floods and Debris Flows at Mount Rainier, Washington, with Emphasis on Tahoma Creek Valley: U.S. Geological Survey Water-Resources Investigations Report 934093, 100 p. doi:10.3133/wri934093. Walder, J. S. and Driedger, C. L., 1994b, Rapid geomorphic change caused by glacial outburst floods and debris flows along Tahoma Creek, Mount Rainier, Washington, U.S.A.: Arctic and Alpine Research, Vol. 26, No. 4, pp. 319–327. doi:10.2307/1551792. Walder, J. S. and Driedger, C. L., 1995, Frequent outburst floods from South Tahoma Glacier, Mount Rainier, U.S.A.: Relation to debris flows, meteorological origin and implications for subglacial hydrology: Journal of Glaciology, Vol. 41, No. 137, pp. 1–10. doi:10.3189/S0022143000017718. Yogev, Y., 2015, personal communication, Mount Rainier National Park, Ashford, WA 98304.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 57–72

Beason, Legg, Kenyon, and Jost

Appendix A. Current debris-ﬂow hazard forecast model in place at Mount Rainier.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 57–72

Water and Sediment Supply Requirements for Post-Wildﬁre Debris Flows in the Western United States PAUL M. SANTI* Department of Geology and Geological Engineering, Colorado School of Mines, 1500 Illinois Street, Golden, CO 80401

BLAIRE MACAULAY Baseline Water Resource, Inc., #7, 3800-19 Street NE, Calgary, AB T2E 6V2, Canada

Key Terms: Water, Intensity, Sediment, Balance ABSTRACT This work explores two hypotheses related to runoffrelated post-wildfire debris flows: 1) their initiation is limited by rainstorm intensity rather than cumulative rainfall depths and 2) they are not sediment supply limited. The first hypothesis suggests that it is common to generate more than enough rainfall to account for the volume of water in the debris flow, but to actually produce a debris flow, the water must be delivered with sufficient intensity. This is demonstrated by data from 44 debris flows from eight burned areas in California, Colorado, and Utah. Assuming a debris flow comprises 30 percent water and 70 percent solids, these events were generated during rainstorms that produced an average of 17 times as much water as necessary to develop a debris flow. Even accounting for infiltration, the rainstorms still generated an overabundance of water. Intensity dependence is also shown by numerous cases in which the exact timing of debris flows can be pinpointed and is contemporaneous with high-intensity bursts of rainfall. The hypothesis is also supported by rainfall intensityduration thresholds where high-volume storms without high-intensity bursts do not generate debris flows. The second hypothesis of sediment-supply independence for the initiation of debris flows is supported by a significant increase in flow volume occurring directly after wildfire, compared to flows in unburned terrain. Also, repeated flows within short time intervals are only possible with an abundance of channel sediment, dry ravel, and bank failure material that can be mobilized. Field observations confirm these sediment sources, even directly after a debris-flow.

*Corresponding author email: psanti@mines.edu

INTRODUCTION Debris flows in the months following mountain wildfires are a serious and growing problem. For example, multiple debris flows initiated during a 9 January 2018 rainstorm in the Thomas Fire area in California caused 23 fatalities and over 160 injuries and damaged over 400 homes (Kean et al., 2019). The soils in post-wildfire terrain are often more susceptible to erosion and debris flows because of decreased infiltration and corresponding increase in runoff following rainfall due to 1) an increase in the hydrophobicity of soil due the deposition of water-repellent chemicals (DeBano, 1981); 2) infilling of fine ash into soil pore space, thereby reducing porosity and conductivity (Kinner and Moody, 2010); 3) soil vulnerability due to the combustion and removal of vegetation (Robichaud, 2000); and 4) raindrop impact on exposed soil that may compress the soil, further reducing pore space and porosity (Swanson, 1981; Martin and Moody, 2001; Wagenbrenner et al., 2006; and Doerr et al., 2009), among other phenomena (Santi et al., 2013). Because debris flows are usually triggered by rainfall events, there is often an implicit assumption that their generation is rainfall-limited. Similarly, the scouring of debris-flow channels during the event may be interpreted to limit the supply of sediment for future flows in the same channel. The purpose of this study is to demonstrate that neither limitation applies to postwildfire debris flows in particular settings, such as the Western United States. Rainfall is still necessary, and sediment must be available to be mobilized into the flow, but the thresholds can be shown to be quite low. This is why post-wildfire debris flows are common, why they are larger than flows in unburned areas, and why they can occur repeatedly in the same channels. The first hypothesis—that post wildfire debris flows are not rainfall volume limited but rather are rainfall intensity limited—is explored through water balance calculations, to demonstrate how rainstorm volume is partitioned into debris flows, infiltration, and runoff, and through records of rainfall intensity and

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 73–85

Santi and MacAulay

Figure 1. Locations of data sets for rainfall limitation analysis. Forty-four debris-flow basins were evaluated from the eight fires shown. Specific basin information is included in Table 1.

debris-flow timing. While this hypothesis has been suggested by others (e.g., Kean et al., 2011; Staley et al., 2012, 2017), our work uses data from a large number of sites, rather than one or two specific ones, and we demonstrate that the hypothesis is more broadly applicable. The second hypothesis—that these debris flows are not sediment supply limited—is investigated through volume comparisons of burned and unburned source areas, field measurement of sediment sources, and records of repeated flows. The question of the influence of transport or supply limitations on debris flows has also been explored by others, who have generally focused on non–fire related debris flows (e.g., Bovis and Jakob, 1999; Jakob et al., 2005; Bennett et al., 2014; Martin et al., 2017; and Rengers et al., 2020). These hypotheses are expected to apply to runoffgenerated debris flows, which are more common in post-wildfire settings within the first several years after a wildfire, rather than debris flows mobilized from shallow landslides (differences are described in detail in Coe et al. [2008a]). DATA SOURCES Data for rainfall limitation used in this study can be found in Gartner (2005); however, portions of the data set are also available in Cannon et al. (2003, 2008). The

data set contains field measurements from 44 drainage basins with post-wildfire debris flows, including debrisflow volume for the measured storm (may include effects of some pre- and post-storm erosion), basin area, aerial extent of burn severity, total rainfall, rainfall duration, and storm rainfall intensity (Figure 1 and Table 1). The total rainfall amount (mm) for each basin was estimated using inverse distance weighting techniques based on the values of proximal rain gauges, typically located within 2 km of the drainage basin (Gartner, 2005). The basins were burned in eight different wildfires in Colorado, California, and Utah. The average size of the study basins is approximately 1 km2 ; however, basin area ranges from 0.01 km2 to 4.1 km2 . Topography ranges from steep, rugged slopes to more gentle gradients (26–94 percent gradient) (Cannon et al., 2008). Vegetation consists of juniper, shrublands, aspen, fir, grasslands, conifers, and chaparral (Cannon et al., 2008). The climate in the southwestern United States is characterized by very dry, semi-arid periods followed by periods of episodic intense rainfall. The Rocky Mountain sites in Utah and Colorado have warm summers, with monsoonal rains delivering high-intensity rainfall during short-duration convective storms, and the southern California sites have warm dry summers, with winter rain often delivered in long-duration storms with high-intensity periods (Moody and Martin, 2001; Cannon et al., 2008). The

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 73–85

Post-Wildlife Debris-Flow Water and Sediment Supply Requirements Table 1. Summary of data used for water balance compilation (Gartner, 2005; Cannon et al., 2003, 2008). Fire Location Missionary Ridge, CO

Basin Name

Haﬂin Kroeger Woodard Elkhorn Basin 23 Root Creek Mayer Gut Coal Seam, CO Coal Seam A Coal Seam F Coal Seam G Coal Seam H Coal Seam L Coal Seam O Overland, CO Jamestown Porphyry Tower Heil Ranch 2 Mollie, UT Santaquin T2 Santaquin T3 Santaquin T4 Santaquin T5 Santaquin T6 Farmington, UT Compton Bench M Compton Bench S Intake Gaviota, CA Janet Creek J3 Gaviota S Old and Grand Prix, CA Silverwood O Silverwood M Devore - May 04 Lytle W Sweetwater C X XX Cleghorn le Lytle Hourglass Lytle AQ1 Cleghorn Water Tank Waterman N Sawpit Silverwood Pe Silverwood Pw Paradise and Cedar, CA El Capitan I El Capitan II

Total Debris Volume (m3 ) Total Rainfall (mm) Basin Area (km2 ) Burned Area (km2 ) 23,967 8,324 6,541 5,493 4,454 1,476 3,485 818 1,391 250 574 325 257 260 326 174 951 4,119 6,262 9,251 3,061 5,282 1,515 511 3,311 6,293 1,315 6,119 4,510 24,937 10,387 3,802 4,341 1,094 1,528 683 561 2,218 2,279 59,281 31,817 22,826 441 382

Colorado basins consist of sedimentary interbedded sandstone, siltstone, and conglomerate; the California basins consist of coarse crystalline igneous rocks (Cannon et al., 2008); and the Utah basins consist generally of “quartzite, sandstone, siltstone, schist, gneiss, and amphibolite” intruded by dikes and overlain by limestone and shale (McDonald and Giraud, 2002). The data for sediment supply can be found in Santi et al. (2008) and Santi and Morandi (2013), consisting of debris-ﬂow volume measurements with associated measurements of in-channel, sheetwash, and rill contributions for the Santi et al. (2008) data set. The study basins are located primarily in Colorado, California, and Utah.

4.31 4.31 4.32 15.24 2.29 2.29 47.50 14.48 17.02 7.87 7.87 7.87 17.02 17.02 25.40 25.40 25.40 12.19 12.19 12.19 12.19 12.19 22.86 22.86 22.86 34.54 34.54 151.43 139.45 125.21 117.78 84.94 116.35 116.06 121.74 122.00 127.20 121.02 72.08 153.73 125.47 130.39 11.00 11.00

4.11 3.09 2.36 1.19 0.9 1.86 1.47 0.96 2.06 0.07 0.05 0.09 0.21 1.29 0.16 0.18 0.1 1.18 1.83 1.77 1.65 1.32 0.25 0.17 0.42 0.23 0.18 1.29 1.2 0.62 0.82 0.99 0.18 0.11 0.11 0.01 0.01 0.1 0.35 4.19 2.64 1.59 0.37 0.33

3.94 3.06 2.18 1.19 0.89 1.84 1.46 0.95 2.05 0.06 0.04 0.03 0.2 0.15 0.16 0.18 0.1 0.46 0.74 0.77 0.72 0.58 0.19 0.13 0.19 0.23 0.18 1.29 1.16 0.62 0.81 0.99 0.17 0.11 0.11 0.01 0.01 0.1 0.25 4.11 2.64 1.59 0.36 0.33

Support for the hypotheses presented in this article also comes from other previous studies. While these are not strictly results from new data that we generated, they are nevertheless included under the “Results” section below to maintain the ﬂow of the argument for the hypotheses. CONCEPTUAL MODEL The analysis of rainfall limitations is done through a water balance calculation, whereby each drainage basin is idealized as a two-layer system, with ash overlying the soil column (this is a common simpliﬁcation, used, for example, in Moody et al., 2009; Woods

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 73–85

Santi and MacAulay

and Balfour, 2012; and Bodi et al., 2011). Kinner and Moody (2010) suggest the two-layer system exists because of the capillary barrier effect. This effect occurs because ash has much higher hydraulic conductivity and infiltration capacity than does the underlying soil (Moody et al., 2009; Bodi et al., 2011; and Gabet and Bookter, 2011). Kinner and Moody (2010) note this barrier effect is most enhanced under dry conditions. This model was also confirmed by Ebel et al. (2012), who measured soil water content profiles at a recently burned site and found that almost no water infiltrated below the ash layer. Because of the high initial infiltration capacity of the ash, we have assumed that 100 percent of the rainfall will infiltrate the ash layer until it reaches saturation, and any additional rainfall will run off. This is a “fill and spill” infiltration model, in which the capacity is not time variant, rather than a Hortonian model that includes time-variable infiltration (Gabet and Bookter, 2011; Rengers et al., 2016, 2019). The effects of evaporation and transpiration were not included because of the short duration of each rainfall event. We have also assumed that antecedent moisture does not play a role for post-wildfire debris-flow generation (Cannon et al., 2008; Kean et al., 2011), so it is not included as part of the water balance calculation. The independence from antecedent moisture is explored in more detail later in the article. Topography plays a significant role in runoff generation and is undoubtedly a factor for the study basins, which generally have greater than 30–50 percent slope (Gartner, 2005). Steeper topography is expected to be characterized by thinner soils, more exposed bedrock, and faster runoff with less infiltration. As a conservative approach, however, this study does not include the amplifying effects of topography, so the actual runoff is expected to be greater than the calculated runoff. Based on this model, the water balance equation, in which each value is a volume over the entire drainage basin, is P = Vdf + Sash + Ssoil + R

(1)

where P = precipitation (total storm rainfall volume); Vdf = volume of water contained in the debris flow; Sash = volume of water stored through infiltration into ash; Ssoil = volume of water stored through infiltration into unburned soil; and R = volume of runoff in the form of overland flow. Each of these parameters is described in detail below. For some less-constrained input variables, a range of values is given so that both high overland flow and low overland flow end members can be calculated. Rainfall is calculated as the total storm rainfall (measured in proximal rain gauges) multiplied by the drainage basin area.

Debris-flow water content can be calculated under the assumption that a debris flow, by definition, contains approximately 20–40 percent water (Pierson and Costa, 1987; Phillips and Davies, 1991). If the material contains less than 20 percent water, then it is a form of rigid mass wasting; if it is above 40 percent water, it is termed “hyper-concentrated flow” (Phillips and Davies, 1991). Therefore, for the data used in this study, water content of the debris flow is calculated as 20 percent (high overland flow end member) or 40 percent (low overland flow end member) of the total volume of the debris-flow deposit. Ash infiltration is calculated as the product of ash porosity, ash thickness, and burned area within each basin. While ash porosity has been measured as high as 67 percent (Woods and Balfour, 2012) to 83 percent (Cerda and Doerr, 2008), this measurement is calculated immediately post-wildfire, and the ash quickly compresses to a lower porosity. We have used a range of 20–30 percent porosity (representing high and low overland flow end members, respectively), which matches observations of other researchers and is close to and slightly larger than the field capacity of 0.12 to 0.24 measured by Ebel (2013). Ash thickness of 10 mm was assumed, based on similar field thickness measurements by Kinner and Moody (2010) and Ebel et al. (2012). The 10-mm assumption is also in agreement with the values used for controlled ash placement in test plots by Gabet and Sternberg (2008), Woods and Balfour (2012), and Bodi et al. (2011). Ash may be thicker than 10 mm after wildfire in more heavily forested areas, such as the pine and juniper areas in Colorado and Utah, but the low tree density and patchy vegetation may result in a lower average thickness, so the assumption of 10 mm is considered reasonable. The ash infiltration calculation also assumes that ash is present over the entire burned portion of the basin and that ash is not present in the unburned portion. There are possibly burned areas with insignificant ash thickness that are not accounted for in our calculation. Ash infiltration is expected to be lower than the unburned soil infiltration values at those locations before the wildfire; therefore, this component results in an increase in runoff. Unaltered soil infiltration is expected in unburned portions of the basin. We estimated this infiltration using U.S. Department of Agriculture (USDA) Natural Resources Conservation Service runoff curves (USDA, 1986). For this calculation, the following assumptions were made:

r Antecedent Moisture Condition II (this is average soil moisture. AMC I—dry soil—may apply directly

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 73–85

Post-Wildlife Debris-Flow Water and Sediment Supply Requirements

Figure 2. Results of water balance calculation for “Low Overland Flow” assumptions.

following the wildfire, but general conditions were assumed to better match with AMC II); r Hydrologic Soil Groups B, C, and D (representing a range of soil types from fine to sandy textures and slow to high infiltration rates); and r Hydrologic Condition Poor to Fair (poor is <30 percent ground cover and fair is 30–70 percent ground cover). For Colorado and Utah, these conditions produce a range of “Runoff Curve Numbers” (CNs) from 71 to 93. For California the CN range is 67 to 89. Runoff Curve Numbers may be converted to potential maximum retention, S (in inches), with the following equation (from Chow et al., 1988): S = 1000/CN − 10

(2)

S may then be used to calculate expected runoﬀ, Pe , with the equation (Chow et al., 1988) Pe = (P − 0.2S)2 / (P + 0.8S)

(3)

where P is cumulative rainfall in inches. While the range of S values (calculated from CN) is given for state-wide location, the Pe values are calculated for each basin depending on the cumulative rainfall measured in that basin or in nearby rain gauges. Overland flow is calculated from Eq. 1 as the difference in the measured rainfall and the calculated infiltration. A range of values is presented, with both low and high overland flow estimations.

RESULTS Water Balance Calculation Figures 2 and 3 show the final water balance calculation for each basin, in which each component (debris flow, infiltration, and overland flow) is represented as a percentage of the total rainfall on the basin. The “Low Overland Flow Calculation” (Figure 2) assumes the maximum infiltration (storage) values for both burned and unburned areas and assumes debris-flow water content of 40 percent. The “High Overland Flow Calculation” (Figure 3) assumes minimum infiltration values and debris-flow water content of 20 percent. After accounting for the cumulative rainfall of the entire storm and portioning the water into storage in the ash and soil components, we rearranged Eq. 1 to solve for R, the runoff as overland flow. There is excess rainfall in the form of overland flow, R, for all basins except for four in Figure 2 (Haflin, Basin 23, Root Creek, and Coal Seam G) and except for one in Figure 3 (Basin 23). Most basins show substantial overland flow, indicating a significant excess of water in the system. The amount of overland flow is greatest for the California basins (Janet Creek J3 through El Capitan II, which are shown on the right side of the graphs in Figures 2 and 3). The amount of water incorporated into the debris flows is much less than the amount available in each rainstorm. Assuming an average proportion of water in the debris flow (30 percent, which is midrange of the low and high values of 20 percent and 40 percent used in the calculation), only a median of 5.9 percent of the rainstorm water is incorporated into debris flows, with the remainder either infiltrating or

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 73–85

Santi and MacAulay

Figure 3. Results of water balance calculation for “High Overland Flow” assumptions.

exiting as overland flow. This represents 1/17th of the available water, and this small fraction verifies that the debris flows are not rainstorm volume limited. The amount of water in the debris flows is shown in Figure 4. Rainfall Intensity Limitations While rainfall volume can be shown not to limit debris-flow occurrence, a dependence on rainfall intensity can also be demonstrated. Figure 5, from Friedman and Santi (2019) and Friedman (2012), shows the close time proximity between rainfall intensity bursts (zones of increased slope on the rain gauge cumula-

Figure 4. Histogram of percent of storm rainfall incorporated into debris flow for each watershed, assuming the debris flow is composed of 70 percent solids and 30 percent water. In very few cases does the debris flow entrain more than 20 percent of the available water.

tive rainfall curves) and the pressure spikes recorded in the nearby pressure transducers. In this study, pressure transducers were drilled into bedrock in the drainage channel to measure debris flows overriding them, and rain gauges were placed within tens of meters of the pressure transducers (Basin 24) or were in adjacent canyons (Basin 16, located 1.1 km from the Basin 32 pressure transducer). These data show debris-flow response within a few minutes of rainfall intensity bursts. Similar short lag times have been measured by other researchers as well (e.g., Coe et al., 2008b; Kean et al., 2011). Rainfall intensity dependence for debris-flow initiation has also been well established through rainfall intensity-duration threshold graphs, in which local data can be used to construct thresholds dividing storms that produce debris flows, which fall above the threshold line, from those that do not. An example is shown in Figure 6, which summarizes the thresholds from numerous locations (with those from burned areas shown in color). It is possible to have rainstorms that generate large total amounts of water (longduration storms) that do not have sufficient intensity to trigger debris flows. Conversely, short-duration and high-intensity storms can exceed the threshold and cause debris flows. Figure 6 also includes two lines for data from the Great Sand Dunes National Park and Preserve (Santi, 2013) showing the increase in the threshold 2 years after the wildfire (the higher threshold is a conservative estimate based on a series of storms that did not produce debris flows). This demonstrates the acute sensitivity of debris-flow generation to rainfall intensity during the 1- to 2-year period directly following a wildfire.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 73–85

Post-Wildlife Debris-Flow Water and Sediment Supply Requirements

Figure 5. Example of short lag time between rainstorm intensity burst (zones of increased slope of the cumulative rainfall curves shown in blue on the bottom graphs) and debris-ﬂow generation (spikes on the pressure transducers shown in blue on the upper graphs). Note that the rain gauge for Basin 24 was located within meters of the pressure transducer, but the rain gauge in Basin 16 was located 1.1 km from the pressure transducer in Basin 32 (Friedman and Santi, 2019).

Antecedent Moisture Independence In order to assess the importance of antecedent moisture on debris-flow generation, we reviewed records of rainfall prior to the first recorded debris flow in each location for which we completed a water balance, with the exception of the two drainage basins in the Gaviota burn area (Janet Creek J3 and Gaviota S) and the two drainage basins in the Paradise and Cedar burn areas (El Capitan I and El Capitan II). In general, when more than one weather station was included in the analysis, we selected stations at different elevations so that one was below the debris-flow initiation area and one was at or above the initiation area. The analysis is summarized in Table 2. With the exception of the Mollie Fire area in Utah, all other weather stations recorded rainfall of 5.6 mm or less (median, 2.5 mm) during the week preceding the first recorded debris flow (the two weather stations reviewed from the Mollie Fire area showed 17.3 and 30.5 mm of rainfall in the preceding week). Several of the weather stations recorded zero rainfall during that

week. These are very low levels of antecedent moisture and strong evidence that little to no antecedent moisture is required to initiate a debris flow and that the storm that causes the debris flow (generally on the same day as the event) provides sufficient moisture. In addition, the total rainfall over the previous month ranged up to 79.2 mm, with a median value of 21 mm. This indicates that the soil is likely at AMC II, with a low level of field moisture. These previous rainstorms in the area did not cause debris flows, which we surmise is because they lacked the high-intensity bursts identified earlier as a critical component. The above data notwithstanding, it has been noted by others that soil moisture can take up to a month to recover after a rainstorm (Hürlimann et al., 2019), so there may be lingering effects not accounted for in our analysis. Sediment Supply Independence The idea of sediment supply independence is that the probability of debris-flow initiation is not dependent

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 73–85

Santi and MacAulay

Figure 6. Compilation of measured intensity-duration thresholds for debris-flow generation at various unburned (gray lines) and burned (colored lines) locations (from Cannon and DeGraff, 2009). Burned areas have lower threshold storms to trigger debris flows, but for all cases it is possible to have long-duration storms producing large cumulative rainfalls that do not generate debris flows because of low rainfall intensity. Dashed line is for Great Sand Dunes National Park and Preserve, 1 year following the Medano Fire (based on 47 debris flows from eight rainstorms), and dotted line is for the same location 2 years after the fire (31 rainstorms, no debris flows).

Table 2. Analysis of antecedent moisture before ﬁrst debris ﬂow (data from NOAA, 2019).

Fire, Location (dates mo/yr) Old and Grand Prix, CA (10/03–11/03)

Mollie, UT (8/01–9/01)

Farmington, UT (7/03)

Missionary Ridge, CO (6/02–7/02)

Total Rainfall in Previous Week [mm (in.)]

Total Rainfall in Previous Month (mm [in.)]

Weather Station

Drainage Basins

San Bernardino FS 226 Ontario Airport Wrightwood Santaquin Chlorinator Payson Ranger Station Farmington

All

12/25/03

0.0 (0.0)*

All All All

12/25/03 12/25/03 09/12/02

0.3 (0.01) 3.8 (0.15) 17.3 (0.68)

1.5 (0.06) 24.4 (0.96) 17.3 (0.68)

All

09/12/02

30.5 (1.2)

40.6 (1.6)

All

04/06/04

2.5 (0.1)

50.8 (2.0)

All

04/06/04

5.1 (0.2)

43.2 (1.7)

Meyer

07/23/02

1.3 (0.05)

36.6 (1.44)

Root Creek & Basin 23 Meyer Root Creek & Basin 23 Coal Seam A Coal Seams F, G, & H

08/03/02

1.0 (0.04)

79.2 (3.12)

07/23/02 08/03/02

5.6 (0.22) 3.0 (0.12)

16.8 (0.66) 21.3 (0.84)

08/05/02 09/07/02

1.0 (0.0) 0.0 (0.0)

11.4 (0.45) 18.5 (0.73)

Farmington Lower Lemon Dam

Durango

Coal Seam, CO (6/02–7/02)

Date of First Post-Fire Debris Flow (mm/dd/yr)

Glenwood Springs #2

*No record from 12/16/03 through 12/25/03, but no precipitation recorded on other dates after 11/1/03

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 73–85

Post-Wildlife Debris-Flow Water and Sediment Supply Requirements

Figure 7. Measurement of incremental debris production using multiple channel cross sections (from Santi et al., 2008). Channel yield rate, calculated as the volume of debris produced per unit channel length, is the slope of the graph at any point.

on the debris volume but rather on other factors, such as rainfall amounts. Furthermore, the magnitude, or volume, of the debris flow is also not limited by sediment supply, at least during the period directly following a wildfire. Three lines of evidence demonstrate that postwildfire debris-flow generation is not sediment supply limited, with the first two dealing explicitly with magnitude and the third dealing with debris-flow probability. The first is the multi-fold increase in volume of debris flows following wildfire. Santi and Morandi (2013) compared the volumes of debris produced from a large data set of western U.S. debris flows, including 274 events from recently burned areas (within 1 year), 162 events from recovering basins (1 to 10 years after wildfire), and 216 events from areas that are unburned or fully recovered (10 years after wildfire). They showed that the area yield rate (debris-flow volume divided by basin area) was doubled for burned areas. When they used cluster analysis to subdivide the data into groups with similar basin size, channel length, and channel gradient, they showed that burned areas had an even higher difference in debris-flow volume, ranging from 2.7 to 5.4 times the volumes produced by unburned areas. These are significant increases, and while they do not specifically measure the availability of post–debris flow sediment, they demonstrate that there is a substantial amount of sediment in post-wildfire drainage basin and channel systems. Any limitations to pre-fire debris flows are radically altered following wildfire. Second, the amount of sediment produced by these debris flows is substantial, and it has been shown that

the majority comes from channel scour as water moves down-canyon. Santi et al. (2008) measured incremental debris production from the channel and surrounding hillside for sections of the drainage channel extending from zero-order channels near the top of the drainage basin to the canyon mouth at the bottom of the basin. An example of their data is shown in Figure 7. Based on data from 46 debris flows, they showed that hillslope and rill erosion accounted for an average of only 3 percent of the final debris volume, but channel scour accounted for nearly the entire remainder. Sediment in the channel accumulates through normal weathering and sedimentation processes, strongly supplemented during and after the fire by dry ravel (at locations where dry ravel is common, such as southern California) (Swanson, 1981; Wells, 1987; Florsheim et al., 1991; and Schmidt et al., 2011). This produces a sediment-filled channel with ample material to be incorporated into a debris flow by channel scour. In some cases, a debris flow may scour to bedrock, but at many locations sediment remains in the channel (Figure 8) and may be incorporated into subsequent flows. Furthermore, post–debris flow channel banks are over-steepened from scour, and these banks frequently fail, recharging the sediment supply for future flows (Figure 9). Tang et al. (2019) demonstrates this concept, showing that over a series of seven storms, sediment came from the channel during the early storms and from hillslopes in later storms. Finally, multiple debris-flow events have been observed in the same canyon over short time frames,

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 73–85

Santi and MacAulay Table 3. Multiple post-wildfire debris flows emanating from the same drainage basins following the 2002 Coal Seam Fires in Colorado (data from Cannon et al., 2003). This is strong evidence of sedimentsupply independence for this type of debris flow. Dates are all for 2002. Coal Seam Fire Basins A B C D F G H I Red Mountain Figure 8. Channel scour from a debris flow, with remaining sediment that can be incorporated into successive debris flows.

indicating that the supply of sediment is not easily exhausted. For example, Gartner et al. (2004) provide a database of post-wildfire debris-flow and flood events in the Western United States, noting at least eight locations where repeated debris flows occurred

8/5 X X X X X X X X X

9/7

9/11

9/12

9/17

10/2–3 X X

X X X X

in the same drainage basin within days to months of each other. Cannon et al. (2003) include records of debris flows and hyperconcentrated sediment flows following the 2002 Coal Seam and Missionary Ridge Fires in Colorado. A summary of these records is included in Tables 3 and 4, which show multiple flows in the majority of the drainage basins (7 of 9 in the Coal Seam Fire area and 15 of 16 in the Missionary Ridge Fire area), further supporting the interpretation of sediment-supply independence. Cannon and Gartner (2005) come to a similar conclusion, noting that “basins with thin colluvial covers and minimal channel-fill deposits generally produce debris flows only in response to the first significant rainfall of the season. Basins with thick channel-fill deposits … frequently produce numerous debris flows throughout the rainy season.” DISCUSSION

Figure 9. Failure of channel banks that have been over-steepened by recent debris-ﬂow scour. This process quickly adds new sediment to the channel that can be incorporated into successive debris ﬂows (photograph by Rich Giraud, Utah Geological Survey).

Wildfire creates hillslope and mountain drainage settings that have a delicate sensitivity to rainfall. Even accounting for a thin layer of porous ash, burned areas have lower capacity to absorb rainfall—because of lower infiltration capacity—resulting in more runoff (Moody et al., 2009; Kinner and Moody, 2010; Bodi et al., 2011; and Gabet and Bookter, 2011). All of these factors create a setting in which the threshold of rainfall intensity and duration is lower for initiation of debris flows and in which rainfall intensity bursts are more important than total rainfall. The response time, seen in Figure 5, is on the order of minutes. The observation that antecedent moisture is not required is an important component in a debris-flow initiation model that ties in closely with the rainfall intensity sensitivity. Whereas an implicit assumption might be that post-wildfire hillsides are composed of dry material that needs to reach a requisite saturation before a debris flow can occur, instead we offer support for

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 73–85

Post-Wildlife Debris-Flow Water and Sediment Supply Requirements Table 4. Multiple post-wildfire debris flows emanating from the same drainage basins following the Missionary Ridge Fires in Colorado (data from Cannon et al., 2003). This is strong evidence of sediment-supply independence for this type of debris flow. Dates are all for 2002. Missionary Ridge Fire Basins Unnamed 1 Day Shearer True Red Freeman Unnamed 2 Root Unnamed 3 Elkhorn Haflin Kroeger Coon Stevens Freed Woodard

7/23

8/3

X X X X

8/5

8/8

8/20

8/21

8/29

9/7

9/10–12

9/20

10/2

X X X X X

X X X

X X X X X

X X

X X X

X X X X X X

a model in which there is ample water in the system throughout the rainstorm, and that debris-flow generation does not require substantial water but simply water delivered quickly enough to overload the reduced infiltration capacity. In post-wildfire settings, soils are more easily eroded (Moody et al., 2005; Santi et al., 2013), and more soil is available to be mobilized because of channel scour, bank failure, and dry ravel during and immediately following the fire (Swanson, 1981; Wells, 1987; Florsheim et al., 1991; Santi et al., 2008, 2013; and Schmidt et al., 2011). It has been documented that debris flows grow substantially in volume in transit (e.g., Santi et al., 2008) and that there can be multiple flows in the same drainage basin during the first year or two following a wildfire (e.g., Cannon et al., 2003; Gartner et al., 2004). These observations also support a different model for debris-flow generation in post-wildfire settings. The implicit model has held that the majority of debris is fully discharged following a debris flow and that substantial time may be required before there is sufficient sediment to generate another debris-flow event (e.g., Theule et al., 2015). Instead, the canyon is better envisioned as a trench through loose material like sand, which may be scoured at the base of the trench, and even if scour progresses to bedrock, there is still material being delivered to the channel through bank erosion. As with the rainfall intensity sensitivity, the delivery of sediment is expected to reduce over time as the area recovers from the wildfire. Recovery includes vegetation regrowth, return to pre-fire soil permeability and erosion resistance, flushing of excess sediment through the system by fluvial processes, and winnowing of fines and development of cobble and boulder armor of stream channels (Santi, 2013; Santi et al., 2013). It has been shown that this recovery is typi-

X X

cally on the order of 1 to 3 years, returning the region to its pre-fire debris-flow sensitivity and potential (Martin and Moody, 2001; Moody and Martin, 2001; and Santi and Morandi, 2013). CONCLUSIONS Using conservative assumptions for infiltration and debris-flow water content, there is excess water from rainfall in nearly every analyzed drainage basin, which produces significant overland flow runoff during debris-flow–generating storms. This means that for post-wildfire settings, at least in the Western United States and perhaps in other semi-arid mountainous or Mediterranean climates, debris flows are not rainfall volume limited. The model for debris-flow generation then becomes a system in which there is ample surface water flow both before and after the debris flow, and the debris flow is triggered not by reaching a threshold total water volume or antecedent saturation but rather by reaching a threshold rainfall intensity. The limiting factor for triggering debris-flow behavior of the fluid runoff is the dynamics of the pulse of water and entrained sediment. Furthermore, the supply of sediment in drainage channels is substantial, producing much larger debris flows than are produced prefire, and the supply of sediment is capable of producing repeat events in the same channel, at least until vegetation recovers enough to temper the overland flow or until smaller rainstorms move sediment through the system by fluvial transport. REFERENCES Bennett, G.; Molnar, P.; McArdell, B.; and Burlando, P., 2014, A probabilistic sediment cascade model of sediment

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 73–85

Santi and MacAulay transfer in the Illgraben: Water Resources Research, Vol. 50, pp. 1225–1244. Bodi, M. B.; Mataix-Solera, J.; Doerr, S. H.; and Cerda, A., 2011, The wettability of ash from burned vegetation and its relationship to Mediterranean plant species type, burn severity and total organic carbon content: Geoderma, Vol. 160, pp. 599–607. Bovis, M. J. and Jakob, M., 1999, The role of debris supply conditions in predicting debris flow activity: Earth Surface Processes Landforms, Vol. 24, pp. 1039–1054. Cannon, S. H. and DeGraff, J., 2009, The increasing wildfire and post-fire debris-flow threat in Western USA, and implications for consequences of climate change. In Sassa, K. and Canuti, P. (Editors), Landslides—Disaster Risk Reduction: SpringerVerlag, Berlin, Germany, pp. 177–190. Cannon, S. H. and Gartner, J. E., 2005, Wildfire-related debris flow from a hazards perspective. In Jakob, M. and Hungr, O. (Editors), Debris-Flow Hazards and Related Phenomena: Praxis, Springer, Berlin, Germany, pp. 363–385. Cannon, S. H.; Gartner, J. E.; Holland-Sears, A.; Thurston, B. M.; and Gleason, J. A., 2003, Debris-flow response of basins burned by the 2002 Coal Seam and Missionary Ridge fires, Colorado. In Boyer, D. D.; Santi, P. M.; and Rogers, W. P. (Editors), Engineering Geology in Colorado—Contributions, Trends, and Case Histories: Association of Engineering Geologists Special Publication 14, Colorado Geological Survey Special Publication 55, on CD-ROM. Cannon, S. H.; Gartner, J. E.; Wilson, R. C.; Bowers, J. C.; and Laber, J. L., 2008, Storm rainfall conditions for floods and debris flows from recently burned areas in southwestern Colorado and southern California: Geomorphology, Vol. 96, pp. 250–269. Cerda, A. and Doerr, S. H., 2008, The effect of ash and needle cover on surface runoff and erosion in the immediate post-fire period: Catena, Vol. 74, pp. 256–263. Chow, V. T.; Maidment, D. R.; and Mays, L. W., 1988, Applied Hydrology: McGraw-Hill, Inc., New York. 572 p. Coe, J. A.; Cannon, S. H.; and Santi, P. M., 2008a, Introduction to the special issue on debris flows initiated by runoff, erosion, and sediment entrainment in western North America: Geomorphology, Vol. 96, No. 3–4, pp. 247–249. Coe, J. A.; Kinner, D. A.; and Godt, J. W., 2008b, Initiation conditions for debris flows generated by runoff at Chalk Cliffs, central Colorado: Geomorphology, Vol. 96, pp. 270–297. DeBano, L. F., 1981, Water Repellent Soils: A State-of-the-Art: General Technical Report PSW-46, U.S. Department of Agriculture Forest Service, Pacific Southwest Forest and Range Experiment Station. pp. 1–21. Doerr, S. H.; Shakesby, R. A.; and MacDonald, L. H., 2009, Soil water repellency—A key factor in post-fire erosion. In Cerda, A. and Robichaud, P. R. (Editors), Restoration Strategies after Forest Fires: Science Publishers, Enfield, NH, pp. 197–223. Ebel, B. A., 2013, Wildfire and aspect effects on hydrologic states after the 2010 Fourmile Canyon Fire: Vadose Zone Journal, Vol. 12, doi:10.2136/vzj2012.0089. Ebel, B. A.; Moody, J. A.; and Martin, D. A., 2012, Hydrologic conditions controlling runoff generation immediately after wildfire: Water Resources Research, Vol. 48, W03529. Florsheim, J. L.; Keller, E. A.; and Best, D. W., 1991, Fluvial sediment transport following chaparral wildfires, Ventura County, southern California: Geology Society America Bulletin, Vol. 103, pp. 504–511. Friedman, E. Q., 2012, Debris-Flow Hazard Assessment and Monitoring Within the 2010 Medano Fire Burn Area, Great Sand

Dunes National Park and Preserve, Colorado: Unpublished M.S. Thesis, Colorado School of Mines, 104 p. Friedman, E. Q. and Santi P. M., 2019, Relationship between rainfall intensity and debris-flow initiation in a southern Colorado burned area. In Kean, J. W.; Coe, J. A.; Santi, P. M.; and Guillen, B. K. (Editors), Debris Flow Hazards Mitigation: Mechanics, Monitoring, Modeling, and Assessment, Proceedings of the Seventh International Conference on Debris-Flow Hazards Mitigation: AEG Special Publication 28, pp. 484–491. Gabet, E. J. and Bookter, A., 2011, Physical, chemical and hydrological properties of ponderosa pine ash: International Journal Wildland Fire, Vol. 20, pp. 443–452. Gabet, E. J. and Sternberg, P., 2008, The effects of vegetative ash on infiltration capacity, sediment transport, and the generation of progressively bulked debris flows: Geomorphology, Vol. 101, pp. 666–673. Gartner, J. E., 2005, Relations Between Wildfire Related DebrisFlow Volumes and Basin Morphology, Burn Severity, Material Properties and Triggering Storm Rainfall: M.A. Thesis, University of Colorado. Gartner, J. E.; Bigio, E. R.; and Cannon, S. H., 2004, Compilation of Post Wildfire Runoff-Event Data from the Western United States: USGS OFR, 2004-1085. Hürlimann, M.; Oorthuis, R.; Abancó, C.; Carleo, L.; and Moya, J., 2019, Monitoring of rainfall and soil moisture at the Rebaixader catchment (Central Pyrenees). In Kean, J. W.; Coe, J. A.; Santi, P. M.; and Giuillen, B. K., (Editors), DebrisFlow Hazards Mitigation: Mechanics, Monitoring, Modeling, and Assessment, Proceedings of the 7th International Conference on Debris-Flow Hazards Mitigation 2019: AEG Special Publication 28, pp. 131–137. Jakob, M.; Bovis, M.; and Oden, M., 2005, The significance of channel recharge rates for estimating debris-flow magnitude and frequency: Earth Surface Processes Landforms, Vol. 30, pp. 755–766. Kean, J. W.; Staley, D. M.; and Cannon, S. H., 2011, In situ measurements of post-fire debris flows in southern California: Comparisons of the timing and magnitude of 24 debris-flow events with rainfall and soil moisture conditions: Journal Geophysical Research Earth Surface, doi.org/10.1029/2011JF002005. Kean, J. W.; Staley, D. M.; Lancaster, J. T.; Rengers, F. K.; Swanson, B. J.; Coe, J. A.; Hernandez, J. L.; Sigman, A. J.; Allstadt, K. E.; and Lindsay, D. N., 2019, Inundation, flow dynamics, and damage in the 9 January 2018 Montecito debris-flow event, California, USA: Opportunities and challenges for post-wildfire risk assessment: Geosphere, Vol. 15, pp. 1140–1163. Kinner, D. A. and Moody, J. A., 2010, Spatial variability of steady-state infiltration into a two-layer soil system on burned hillslopes: Journal Hydrology, Vol. 381, pp. 322–332. Martin, D. A. and Moody, J. A., 2001, Comparison of soil infiltration rates in burned and unburned mountainous watersheds: Hydrological Processes, Vol. 15, pp. 2893–2903. Martin, Y. E.; Johnson, E.; and Chaikina, O., 2017, Gully recharge rates and debris flows: A combined numerical modeling and field-based investigation, Haida Gwaii, British Columbia: Geomorphology, Vol. 278, pp. 252–268. McDonald, G. N. and Giraud, R. E., 2002, September 12, 2002, Fire-Related Debris Flows East of Santaquin and Spring Lake, Utah. Utah Geological Survey: Electronic document, available at http://geology.utah.gov/online/techrpt/ santaquin0902.pdf

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 73–85

Post-Wildlife Debris-Flow Water and Sediment Supply Requirements Moody, J. A.; Kinner, D. A.; and Ubeda, X., 2009, Linking hydraulic properties of fire-affected soils to infiltration and water repellency: Journal Hydrology, Vol. 379, pp. 291–303. Moody, J. A. and Martin, D. A., 2001, Initial hydrologic and geomorphic response following a wildfire in the Colorado front range: Earth Surface Processes Landforms, Vol. 26, pp. 1049–1070. Moody, J. A.; Smith, J. D.; and Ragan, B. W., 2005, Critical shear stress for erosion of cohesive soils subjected to temperatures typical of wildfires: Journal Geophysical Research Earth Surface, Vol. 110, pp. 1–13. NOAA, 2019, Climate Data Online, https://www.ncdc.noaa.gov/ cdo-web/, accessed January 7, 2019. Phillips, C. J. and Davies, T. R. H., 1991, Determining rheological parameters of debris flow material: Geomorphology, Vol. 4, pp. 101–110. Pierson, T. C. and Costa, J. E., 1987, A rheologic classification of subaerial sediment-water flows: Geological Society America Reviews Engineering Geology, Vol. 7, pp. 1–12. Rengers, F. K.; Kean, J. W.; Reitman, N. G.; Smith, J. B.; Coe, J. A.; and McGuire, L. A., 2020, The influence of frost weathering on debris flow sediment supply in an alpine basin: Journal Geophysical Research Earth Surface, Vol. 125, https://doi.org/10.1029/2019JF005369. Rengers, F. K.; McGuire, L. A.; Kean, J. W.; Staley, D. M.; and Hobley, D. E. J., 2016, Model simulations of flood and debris flow timing in steep catchments after wildfire: Water Resources Research, Vol. 52, doi:10.1002/2015WR018176. Rengers, F. K.; McGuire, L. A.; Kean, J. W.; Staley, D. M.; and Youberg, A. M., 2019, Progress in simplifying hydrologic model parameterization for broad applications to postwildfire flooding and debris-flow hazards: Earth Surface Processes Landforms, Vol. 44, pp. 3078–3092. Robichaud, P. R., 2000, Fire effects on infiltration rates after prescribed fire in Northern Rocky Mountain forests, USA: Journal Hydrology, Vol. 231–232, pp. 220–229. Santi, P. M., 2013, Timing of landscape recovery after wildfire based on debris-flow parameters: In Geological Society of America Annual Meeting Abstracts with Programs, Geological Society of America, Boulder, CO. Santi, P. M.; Cannon, S.; and DeGraff, J., 2013, Wildfire and landscape change. In Shroder, J.; James, L. A.; Harden, C. P.; and Clague, J. J. (Editors), Treatise on Geomorphology.Vol. 13, Geomorphology of Human Disturbances, Climate Change, and Natural Hazards: Academic Press, San Diego, CA, pp. 262–287. Santi, P. M.; deWolfe, V. G.; Higgins, J. D.; Cannon, S. H.; and Gartner, J. E., 2008, Sources of debris flow material in burned areas: Geomorphology, Vol. 96, pp. 310–321.

Santi, P. M. and Morandi, L., 2013, Comparison of debris-flow volumes from burned and unburned areas: Landslides, Vol. 10, pp. 757–769. Schmidt, K. M.; Hanshaw, M. N.; Howle, J. F.; Staley, D. M.; Stock, J. D.; and Bawden, G. W., 2011, Hydrologic conditions and terrestrial laser scanning of post-fire debris flows in the San Gabriel Mountains, CA, USA. In Proceedings, 5th International Conference on Debris-flow Hazards Mitigation: Mechanics, Prediction and Assessment, June 14–17, 2011, Padua, Italy. Staley, D.; Kean, J.; Cannon, S.; Schmidt, K.; and Laber, J., 2012, Objective definition of rainfall intensity–duration thresholds for the initiation of post-fire debris flows in southern California: Landslides, Vol. 10, doi:10.1007/s10346-0120341-9. Staley, D. M.; Negri, J. A.; Kean, J. W.; Laber, J. L.; Tillery, A. C.; and Youberg, A. M., 2017, Prediction of spatially explicit rainfall intensity–duration thresholds for post-fire debris-flow generation in the western United States: Geomorphology, Vol. 278, pp. 149–162. Swanson, F. J., 1981, Fire and geomorphic processes. In Mooney, H. A.; Bonniksen, T. H.; Christensen, N. L.; Lotan, J. E.; and Reiners, W. A. (Editors), Fire Regimes and Ecosystem Properties: U.S. Department of Agriculture, Forest Service General Technical Report WO-26, 401–420. Tang, H.; McGuire, L. A.; Rengers, F. K.; Kean, J. W.; Staley, D. M.; and Smith, J. B., 2019, Evolution of debris-flow initiation mechanisms and sediment sources during a sequence of postwildfire rainstorms: Journal Geophysical Research Earth Surface, Vol. 124, pp. 1572–1595. Theule, J.; Liébault, F.; Laigle, D.; Loye, A.; and Jaboyedoff, M., 2015, Channel scour and fill by debris flows and bedload transport: Geomorphology, Vol. 243, pp. 92–105. United States Department of Agriculture (USDA), 1986, Urban Hydrology for Small Watersheds (PDF): Technical Release 55 (TR-55) (2nd ed.). Natural Resources Conservation Service, Conservation Engineering Division. Wagenbrenner, J. W.; MacDonald, L. H.; and Rough, D., 2006, Effectiveness of three post-fire treatments in the Colorado Front Range: Hydrological Processes, Vol. 20, pp. 2989–3006. Wells, W. G., 1987, The effect of fire on the generation of debris flows in southern California. In Costa, J. E. and Wieczorek, G. F. (Editors), Debris flows/avalanches: Process, recognition, and mitigation: GSA Reviews Engineering Geology, Vol. 7, pp. 105–113. Woods, S. W. and Balfour, V. N., 2012, The effects of soil texture and ash thickness on the post-wildfire hydrogeological response from ash-covered soils: Journal Hydrology, Vol. 393, pp. 274–286.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 73–85

Measurements of Velocity Profiles in Natural Debris Flows: A View behind the Muddy Curtain GEORG NAGL* JOHANNES HÜBL ROLAND KAITNA Department of Civil Engineering and Natural Hazards, University of Natural Resources and Life Sciences, Peter Jordan Straße 82, 1190 Vienna, Austria

Key Terms: Debris Flow, Monitoring, Velocity Profile ABSTRACT The internal deformation behavior of natural debris flows is of interest for model development and model testing for debris-flow hazard mitigation. Up to now, only a few attempts have been made to measure velocity profiles in natural debris flows due to the low predictability and high destructive power of these flows. In this contribution, we present recent advances to measure in-situ velocity profiles together with flow parameters like flow height, basal normal stress, and pore fluid pressure. This was accomplished by constructing a fin-shaped monitoring barrier with an array of paired conductivity sensors in the middle of Gadria Creek, Italy. We present results from two natural debris-flow events. Compared to the first event on July 10, 2017, the second event on August 19, 2017, was visually more liquid. Both debris flows exhibited significant longitudinal changes of flow properties like flow height and density. The liquefaction ratios reached values up to unity in some sections of the flows. Velocity profiles for the July event were mostly concaveup, while the profiles for the more liquid event in August were linear to convex. These measurements provide new insights into the dynamics of real-scale debris flows. INTRODUCTION Debris flows are gravitational mass flows that occur in steep channels, which characterize mountain landscapes. The high volumetric content of sediment together with grain sizes ranging over several orders of magnitudes, and velocities sometimes exceeding 15 m/s, make measuring velocity profiles in natural debris flows challenging. However, observations under natural conditions avoid scaling effects and provide some indication of the constitutive flow behavior of the mixture, both of which are useful for model development and testing. The aim of this study is to provide

*Corresponding author email: georg.nagl@boku.ac.at

data on the internal deformation behavior of natural debris flows. Measurements of velocity profiles in natural sediment-water mixtures are rare, but they mostly show a strong dependence on material composition (Arai and Takahashi, 1983; Mainali and Rajaratnam, 1994; Johnson et al., 2012; and Kaitna et al., 2014), which has also been observed in artificial solid-fluid mixtures (Sanvitale et al., 2011; Chen et al., 2017). For natural flows, measurements of mean velocity and surface velocity are available (Berti et al., 1999; Genevois et al., 2000; Marchi et al., 2002; and Theule et al., 2018). For example, internal deformation behavior was derived from paired shear force measurements on a vertical side wall at the Illgraben test site in Switzerland (Walter and McArdell, 2015). The importance of non-hydrostatic fluid pressure, which reduces the shear resistance in debris-flow mixtures, has been shown by different authors (e.g., Pierson, 1986; Iverson and Lahusen, 1989; Iverson, 1997; Major, 2000; and Kaitna et al., 2014, 2016), and it has also been measured in the field (McArdell et al., 2007; McCoy et al., 2010, 2013). Additionally, the runout length can be increased by the remobilization and deposit behavior of a residual layer due the pulsing nature of debris flows (Davies, 1990; Hu et al., 2011). Herein, we present results of our efforts to measure the internal deformation behavior in natural debris flows at a monitoring station on Gadria Creek, Italy. We first give an overview of the test site and the installed setup. Subsequently, we show measurements of velocity profiles, normal stresses, flow heights, and basal pore fluid pressure for two debris flows observed in 2017. METHODS Field Site The catchment for Gadria Creek is located in the Vinschgau valley in South Tyrol, Italy, and it occupies an area of 6.3 km² (Figure 1a). The highest point of the catchment is at 2,945 m above sea level (a.s.l.), and the

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 87–94

Nagl, Hübl, and Kaitna

Figure 1. (a) Study site of the Gadria Creek in South Tyrol, Italy. (b) Monitoring barrier with measurement system.

confluence of the receiving river Etsch is at 807 m a.s.l. With one to two debris flows per year in recent years, the area was considered to be well suited for debrisflow monitoring (Comiti et al., 2014; Coviello et al., 2019a, 2019b). The steep terrain and frequent thunderstorm events, as well as metamorphic rock and thick glacial deposits, ensure a sufficient quantity of material available to be mobilized and transported. A grain size distribution of deposited debris-flow material carried out in autumn 2017 showed a wide range of grain sizes. A rigid combination of pebble counts on levées and sieving analysis of collected material less than 63 mm showed in the

cumulative curve a median diameter (d50 ) of 150 mm, a d10 of 6.3 mm, and a d90 of 420 mm (Figure 2). Less than 2 percent of the material was clay and silt. Since the last ice age, Gadria Creek has developed a large fan, which is mainly used for agriculture and settlement (Brardinoni et al., 2018). At the apex of the fan, at 1,390 m a.s.l., a slit check dam was built, providing a retention capacity of around 40,000 to 60,000 m3 . Just upstream of the retention area, a monitoring station was installed by the Torrent Control Service of South Tyrol in cooperation with the Free University of Bozen-Bolzano in 2011, including two radar sensors to measure ﬂow height, rain gauges, geophones, and three

Figure 2. Rigid combination of sieving analysis and pebble count of debris-ﬂow deposit provided by Bunte and Abt (2001).

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 87–94

Velocity Proﬁles in Natural Debris Flows

cameras (Comiti et al., 2014) (Figure 1a). In 2016, the test site was extended with a sensor-equipped debrisflow breaker (“monitoring barrier”) to measure impact pressures and investigate the process/barrier/ground interaction. In the course of the construction, force plates, fluid pressure sensors, and a velocity profiler were also installed. Barrier The monitoring barrier is located 200 m upstream of the retention basin at an altitude of 1,400 m a.s.l. The mean channel slope is 6° at the position of the barrier, and it is protected against erosion. The construction consists of two concrete parts, the barrier itself and an unconnected traverse check dam in front of the barrier flush to the ground. For measuring normal stress and shear stress, two force plates were installed on the transverse check dam, one in front of the barrier and the second one 2 m to the side, both set to a sampling frequency of 2,400 Hz. The barrier was combined into a single concrete fin-shaped structure in the middle of the channel and connected to a foundation plate (Figure 1b). Monitoring System Two quadratic force plates of 1 m2 were attached to the transverse check dam. Each force plate is supported by four load pins with a maximum capacity of 10 kN each. In the middle of each force plate, a fluid pressure sensor was installed. Each sensor consists of a pressure transducer connected to a reservoir filled with hydraulic oil. The top of the sensor (flush with the force plate) is sealed with a thin silicone membrane and protected with two steel meshes of 0.5 and 2 mm grid sizes, similar to those used in rotating drum experiments by Kaitna et al. (2014). Two ultra-sonic sensors for flow height measurement were installed above each force plate. The sampling frequency of the ultra-sonic sensors and the pressure transducer was set to 100 Hz. The sensor data recorder is a MGCplus HBM data acquisition system with a sampling rate set to 2,400 Hz. The velocity profiler is situated on the orographically left side of the barrier 3 m behind the front to minimize the disturbance of the passing material, but it is still capable of capturing a maximum flow height of 1.8 m. The profiler consists of 11 sensors at different heights (levels). The first level is located at 18 cm above the concrete bed; the next levels are equally stepped at 15 cm intervals. Each velocity sensor consists of a pair of conductivity sensors at a distance of 6 cm apart. All signals were filtered with a Butterworth lowband-pass 500 Hz filter. The normalized sensor signals were cross-correlated to determine the velocity of passing debris (Nagl and Hübl, 2017). We set the size of the

correlation window to 1 second (2,400 data points) and moved the window with a step size of 24 data points to derive continuous velocity estimates over time for each level. Results with a correlation coefficient <0.8 and unrealistic accelerations from adjacent values were excluded from further analysis, following the argumentation of Kern et al. (2010) and Kaitna et al. (2014). Finally, a digital video system equipped with an infrared spot was installed on the orographic left side of the channel, which enabled us to assess the surface velocity near the profiler by particle tracking. RESULTS Debris Flow of July 10, 2017 On July 10, 2017, a debris flow was triggered by intense rainfall. The front velocity was about 1 m/s, and the maximum flow height was around 1 m (Figure 3d). Video recordings revealed that the flow had a steep front with rocks around 0.5 m in diameter, followed by a mud-rich tail with some boulders immersed in the flow (see video snapshots in Figure 3a). The main surge was followed by small waves. The complete event lasted around 288 seconds (4.8 minutes) and had a total volume below 1,000 m3 . The normal stress, σN , reached values up to 19,000 N/m2 , and the basal pore fluid (P) pressure peaked only slightly lower (Figure 4). The liquefaction ratio (LR = P/σN ) was therefore very high throughout the flow and reached values up to 0.9 at the tail. Hence, except for the very front of the flow, excess pore water pressure was observed during the whole event duration. For the duration of the first surge, the median of the velocity profiles from the profiler exhibited a concave-up form. The numbers beside the boxes (Figure 3e–g) are the number of successful correlations (see Methods). The independently derived surface velocity (red box) is in the same range but slightly higher than the uppermost velocity of the profiler. This might be connected to the non-existent effect of wall friction, as surface velocities were derived at some distance from the barrier. The 10/90 percentiles of the box-whisker plot shows the highest variability on the upper levels. A closer look into the small waves shows a convex form of the velocity profile (Figure 3f). Taking all velocity profiles into consideration, a concave-up form dominates. During the July 10th event, no deposition of sediment was observed at the sensor location. Debris Flow of August 19, 2017 The second event on August 19, 2017, again followed a heavy rainfall event, and it began with

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 87–94

Nagl, Hübl, and Kaitna

Figure 3. Data of the debris flow from July 10, 2017. (a-c) Picture series. (d) Flow height (m) of the ultra-sonic sensor of force plate 1 beside the barrier. (b) Velocity profile of the first surge (360–410 seconds). (c) Velocity profile of a small wave (460–480 seconds). (d) Collective velocity profile of the complete debris flow. The gray boxes represent the 10th and 90th percentiles, and the whiskers are the minimum and maximum values. The points in the boxes stand for the median values. Red color box on top presents the surface velocity.

a sediment-laden flood that included woody debris, which later transformed into a debris flow with a less pronounced front and a maximum flow height of 1.8 m. The maximum surface velocities up to 4 m/s (Figure 5d) were much higher compared to the event in July 2017. Additionally, the hydrograph differed significantly. The August 2017 event consisted of two main

surges with no characteristic bouldery front; the second surge included six small waves. A mud-rich tail with no visible boulders finalized the debris flow. The complete event lasted 1,200 seconds (∼20 minutes). As mentioned earlier, woody debris was transported during the precursory sediment-laden flood as well as during the debris flow. Video recordings revealed that

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 87–94

Velocity Proﬁles in Natural Debris Flows

Figure 4. Normal stress (red line), basal pore ﬂuid pressure (blue line), and liquefaction ratio (black line) in running average values of 1 Hz.

a log got caught for 164 seconds at the front of the barrier, causing (1) sediment to deposit in the vicinity of the force plates, and (2) diversion of the flow to some extent in the cross-channel direction. This affected the force plate in front of the barrier significantly and probably also the second plate at the side of the barrier. Here, we measured normal stresses up to 36,000 N/m2 during the first surge. The corresponding basal pore fluid pressure achieved values to 20,000 N/m2 , and as a result, the liquefaction ratio reached values of 0.5 to 0.6 (Figure 6). For the second surge, an excessive pore pressure was observed, but it did not reach values as measured for the flow on July 10th. We found a linear to slightly convex velocity profile for the first surge and for the complete event from the profiler (Figure 5e and g). For the debris-flow event of August 19, 2017, some fluvially transported channel sediment was deposited up to a height of 0.2 m before the first surge arrived at the barrier. The first incoming surge eroded the deposited layer at the front of the surge at the measurement time of 150 seconds, as shown in Figure 5e with the black bars. The second surge (700–1,200 seconds) showed no mobilization of the first sensor level. For the periods in which the debris was stationary between the surges, no velocities were measured with the velocity profiler. DISCUSSION The velocity profiles shown here represent the first results from our monitoring site at Gadria Creek. Despite the fact that differences in the median velocity values over the height of the flow are larger than the 10 and 90 percentile data, the derived data are sub-

ject to some uncertainties that must be taken into account. First, there are uncertainties that are connected to shortcomings of the experimental field setup. It is likely that we measured velocities of particles passing and probably sliding along a rigid wall; i.e., there is the effect of wall friction (cf. Jop et al., 2005; Kaitna et al., 2014). Additionally, we measured only particle velocity and not fluid velocity, and the geometry of the paired conductivity sensors captured only flow variations in the flow direction. Second, there are uncertainties associated with the data analysis. For example, the choice of a threshold for the correlation coefficient is to some extent arbitrary. We tried to avoid misleading correlation results by defining a high correlation coefficient of 0.8. Another source of error arises from the comparison of velocities derived from the profiler with a surface velocity derived from video recordings. Due to the resolution of the camera, we could not derive surface velocities at the boundary of the barrier, but only in a region some 5–20 cm distant. Additionally, we found that for natural flows including large boulders and woody debris, the deposition pattern may influence the flow along the barrier, as can be seen for the second event in August 2017. This seems unavoidable for a field study, as we cannot regulate the flow hydrograph or the flow composition. Despite these limitations, the monitoring site provided detailed information on the deformation behavior, erosion, and deposition pattern of the natural debris flows. For example, the first debris flow (July 10, 2017) showed concave-up profiles and velocities at the first level and then changed to a convex profile. Instead, the front of the second debris flow (August 19, 2017) showed some erosion on the deposited first level during the very first part of the front and showed a

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 87–94

Nagl, Hübl, and Kaitna

Figure 5. Data of the debris flow from August 19, 2017. (a-c) Picture series. (d) Flow height (m) of the ultra-sonic sensor of force plate 1 beside the barrier. (b) Velocity profile of the first surge (200–300 seconds). (c) Velocity profile of the second surge (700–1,000 seconds). (d) Collective velocity profile of the complete debris flow. The gray box represents the 10the and 90th percentiles, and the whiskers are the extreme values. The points in the boxes stand for the median values.

linear velocity profile with a lower liquefaction ratio than the first debris flow. For the fast-flowing and rather fluid middle part of the second event (Figure 5f), the derived velocity profiles were convex, with very low velocities at the base, indicating that material that was deposited earlier was overridden by a surge from be-

hind, similar as for the small waves during the second event. We note that the average proﬁles for the total duration of the event contain very diﬀerent velocity values. This is due to the fact that at some levels, positive correlations of conductivity signals were only possible for

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 87–94

Velocity Proﬁles in Natural Debris Flows

Figure 6. Normal stress (red line), basal pore ﬂuid pressure (blue line), and liquefaction ratio (black line) in running average values of 1 Hz.

a limited duration. The physical interpretation is that either no debris-flow material passed the sensor (this happened typically in the uppermost layers), or there was material touching the sensor, but it did not move. The latter occurred for the debris flow in August 2017 (Figure 5), where we detected no movement at the lowermost layer during most of the flow. CONCLUSIONS Two debris flows occurred at Gadria Creek in South Tyrol in the year 2017, and they provided a successful test of the recently installed monitoring site. At this “monitoring barrier,” we measured the vertical velocity profiles when the debris flows passed the concrete structure. Close to the structure, basal normal stress, pore fluid pressure, and flow height were recorded. The minimum temporal resolution for the velocity profiles at this stage of our analysis is around 1 second. Our measurements demonstrate that natural debris flows undergo different states of deformation during the flow and indicate no constant velocity profile throughout the flow. Velocity profiles are strongly affected by surges and deposition of material between surges. We assume that the general shape of the derived profiles may be representative for the respective section of the flow. The connection between excess pore fluid pressure and the velocity profiles needs to be further investigated. ACKNOWLEDGMENTS Friedrich Zott deserves all the credit for the installation of the monitoring system. We thank Pierrepaolo Macconi and Lea Gasser for logistic and field support in all situations and all other parties that helped in this

project. The monitoring barrier has been conducted through a cooperative agreement between the Department of Civil Protection of the Autonomous Province of Bozen-Bolzano in Italy and the University of Natural Resources and Life Science, Vienna. REFERENCES Arai, M. and Takahashi, T., 1983, A method for measuring velocity profiles in mud flows. In Proceedings 20th International Congress, Vol. 3: International Association for Hydraulic Research (IAHR), pp. 279–286. Berti, M.; Genevois, R.; Simoni, A.; and Tecca, P. R., 1999, Field observations of a debris flow event in the Dolomites: Geomorphology, Vol. 29, No. 3–4, pp. 265–274. doi:10.1016/S0169555X(99)00018-5. Brardinoni, F.; Picotti, V.; Maraio, S.; Bruno, P. P.; Cucato, M.; Morelli, C.; and Mair, V., 2018, Postglacial evolution of a formerly glaciated valley: Reconstructing sediment supply, fan building, and confluence effects at the millennial time scale. GSA Bulletin, Vol. 130, No. 9–10, pp. 1457–1473. Bunte, K. and Abt, S. R., 2001, Sampling Surface and Subsurface Particle-Size Distributions in Wadable Gravel- and Cobble-Bed Streams for Analyses in Sediment Transport, Hydraulics, and Streambed Monitoring: U.S. Department of Agriculture, Forest Service, Rocky Mountain Research Station, Fort Collins, CO, General Technical Report RMRS-GTR-74, 428 p. Chen, H.; Hu, K.; Cui, P.; and Chen, X., 2017, Investigation of vertical velocity distribution in debris flows by PIV measurement: Geomatics, Natural Hazards and Risk, Vol. 8, No. 2, pp. 1631–1642. doi:10.1080/19475705.2017.1366955. Comiti, F.; Marchi, L.; Macconi, P.; Arattano, M.; Bertoldi, G.; Borga, M.; Brardinoni, F.; Cavalli, M.; D’Agostino, V.; Penna, D.; and Theule, J., 2014, A new monitoring station for debris flows in the European Alps. First observations in the Gadria basin: Natural Hazards, Vol. 73, No. 3, pp. 1175– 1198. doi:10.1007/s11069-014-1088-5. Coviello, V.; Arattano, M.; Comiti, F.; Macconi, P.; and Marchi, L., 2019a, Seismic characterization of debris flows: Insights into energy radiation and implications for warning:

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 87–94

Nagl, Hübl, and Kaitna Journal of Geophysical Research: Earth Surface, Vol. 124, No. 6, pp. 1440–1463. Coviello, V.; Theule, J. I.; Marchi, L.; Comiti, F.; Crema, S.; Cavalli, M.; Arattano, M.; Lucía, A.; and Macconi, P., 2019b, Deciphering sediment dynamics in a debris flow catchment: Insights from instrumental monitoring and highresolution topography. In Kean, J. W.; Coe, J. A.; Santi, P. M.; and Guillen, B. K. (Editors), Proceedings Seventh International Conference on Debris-flow Hazards Mitigation: Association of Environmental and Engineering Geologists Special Publication 28. Colorado School of Mines, Arthur Lakes Library, pp 1–8; http://dx.doi.org/10.25676/11124/173231. Davies, T. R. H., 1990, Debris-flow surges—Experimental simulation: Journal of Hydrology (New Zealand), Vol. 29, No. 1, pp. 18–46. Genevois, R.; Tecca, P. R.; Berti, M.; and Simoni, A., 2000, Debris-flows in the dolomites: Experimental data from a monitoring system. In Wieczorek, G. F. and Naeser, N. D. (Editors), Debris Flow Hazard Mitigation: Mechanics, Prediction and Assessment: Balkema, Rotterdam, Netherlands, pp. 283–292. Hu, K.; Hu, C.; Li, Y.; and Cui, P., 2011, Characteristics and mechanism of debris-flow surges at Jiangjia Ravine. In Proceedings 5th International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction and Assessment: Casa Editrice Università La Sapienza, Roma, Italy, pp. 211–217. Iverson, R. M., 1997, The physics of debris flows: Reviews of Geophysics, Vol. 35, No. 3, pp. 245–296. Iverson, R. M. and Lahusen, R. G., 1989, Dynamic porepressure fluctuations in rapidly shearing granular materials: Science, Vol. 246, No. 4931, pp. 796–799. doi:10.1126/ science.246.4931.796. Johnson, C. G.; Kokelaar, B. P.; Iverson, R. M.; Logan, M.; Lahusen, R. G.; and Gray, J. M. N. T., 2012, Grain-size segregation and levee formation in geophysical mass flows: Journal Geophysical Research, Vol. 117, No. F1. pp. F01032. doi:10.1029/2011JF002185. Jop, P.; Forterre, Y.; and Pouliquen, O., 2005, Crucial role of sidewalls in granular surface flows: Consequences for the rheology: Journal Fluid Mechanics, Vol. 541, No. 1, pp. 167–192. doi:10.1017/S0022112005005987. Kaitna, R.; Dietrich, W. E.; and Hsu, L., 2014, Surface slopes, velocity profiles and fluid pressure in coarse-grained debris flows saturated with water and mud: Journal Fluid Mechanics, Vol. 741, pp. 377–403. doi:10.1017/jfm.2013.675. Kaitna, R.; Palucis, M. C.; Yohannes, B.; Hill, K. M.; and Dietrich, W. E., 2016, Effects of coarse grain size distribution and fine particle content on pore fluid pressure and shear behavior in experimental debris flows: Journal Geophysical Research: Earth Surface, Vol. 121, No. 2, pp. 415–441. doi:10.1002/2015JF003725.

Kern, M. A.; Bartelt, P. A.; and Sovilla, B., 2010, Velocity profile inversion in dense avalanche flow: Annales Glaciology, Vol. 51, No. 54, pp. 27–31. doi:10.3189/172756410791386643. Mainali, A. and Rajaratnam, N., 1994, Experimental study of debris flows: Journal Hydraulic Engineering, Vol. 120, pp. 104–123. Major, J. J., 2000, Gravity-driven consolidation of granular slurries: Implications for debris-flow deposition and deposit characteristics: Journal of Sedimentary Research, Vol. 70, No. 1, pp. 64–83. Marchi, L.; Arattano, M.; and Deganutti, A. M., 2002, Ten years of debris-flow monitoring in the Moscardo Torrent (Italian Alps): Geomorphology, Vol. 46, No. 1–2, pp. 1–17. doi:10.1016/S0169-555X(01)00162-3. McArdell, B. W.; Bartelt, P.; and Kowalski, J., 2007, Field observations of basal forces and fluid pore pressure in a debris flow: Geophysical Research Letters, Vol. 34, No. 7, pp. 171. doi:10.1029/2006GL029183. McCoy, S. W.; Kean, J. W.; Coe, J. A.; Staley, D. M.; Wasklewicz, T. A.; and Tucker, G. E., 2010, Evolution of a natural debris flow. In situ measurements of flow dynamics, video imagery, and terrestrial laser scanning: Geology, Vol. 38, No. 8, pp. 735–738. doi:10.1130/G30928.1. McCoy, S. W.; Tucker, G. E.; Kean, J. W.; and Coe, J. A., 2013, Field measurement of basal forces generated by erosive debris flows: Journal Geophysical Research: Earth Surface, Vol. 118, No. 2, pp. 589–602. doi:10.1002/jgrf.20041. Nagl, G. and Hübl, J., 2017, A check-dam to measure debris flow–structure interactions in the Gadria Torrent. In Matjaž, M.; Vít, V.; Yueping, Y.; and Kyoji, S. (Editors), Advancing Culture of Living with Landslides: Springer International Publishing, Cham, Switzerland, pp. 465–471. Pierson, T. C., 1986, Flow behaviour of channelized debris flows, Mount St. Helens, Washington. In Abrahams, A. D. (Editor), Hillslope Processes: Allen & Unwin, Boston, MA, pp. 269–296. Sanvitale, N.; Bowman, E. T.; and Genevois, R., 2011, Experimental measurements of velocity through granular-liquid flows. In Genevois, R.; Hamilton, D. L.; and Prestininzi, A. (Editors), Proceedings of the 5th International Conference on Debris flow Hazards Mitigation: Mechanics, Prediction and Assessment. Casa Editrice Università La Sapienza, Rome, Italy, pp 375–384. Theule, J. I.; Crema, S.; Marchi, L.; Cavalli, M.; and Comiti, F., 2018, Exploiting LSPIV to assess debris-flow velocities in the field: Natural Hazards and Earth System Sciences, Vol. 18, pp. 1–13. doi:10.5194/nhess-18-1-2018. Walter, F. and McArdell, B., 2015, What is the velocity profile of debris flows?: EGU General Assembly Conference Abstracts, Vol. 17, pp. EGU 2015-12815.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 87–94

Combining Instrumental Monitoring and High-Resolution Topography for Estimating Sediment Yield in a Debris-Flow Catchment VELIO COVIELLO* Free University of Bozen-Bolzano, Faculty of Science and Technology, piazza Università 5, 39100 Bolzano, Italy

JOSHUA I. THEULE TerrAlp Consulting, Chemin du Grand Pré 100, 38410 St Martin d’Uriage, France

STEFANO CREMA Research Institute for Geo-hydrological Protection, National Research Council, Corso Stati Uniti 4, 35127 Padova, Italy

MASSIMO ARATTANO Research Institute for Geo-hydrological Protection, National Research Council, Strada delle Cacce 73, 10135 Torino, Italy

FRANCESCO COMITI Free University of Bozen-Bolzano, Faculty of Science and Technology, piazza Università 5, 39100 Bolzano, Italy

MARCO CAVALLI Research Institute for Geo-hydrological Protection, National Research Council, Corso Stati Uniti 4, 35127 Padova, Italy

ANA LUCÍA Center for Applied Geoscience, Eberhard Karls Universität Tübingen, Schnarrenbergstrasse 94-96, D-72074 Tübingen, Germany

PIERPAOLO MACCONI Civil Protection Agency, Autonomous Province of Bozen-Bolzano, via Cesare Battisti 23, 39100 Bolzano, Italy

LORENZO MARCHI Research Institute for Geo-hydrological Protection, National Research Council, Corso Stati Uniti 4, 35127 Padova, Italy

Key Terms: Bedload Transport, Debris Flows, Monitoring, Rainfall, Topographic Survey, Gadria Catchment ABSTRACT In mountain basins, long-term instrumental monitoring coupled with high-resolution topographic surveys can provide important information on sediment yield. The *Corresponding author email: velio.coviello@unibz.it

Gadria catchment, located in the eastern Italian Alps, typically features several low-magnitude flood episodes and a few debris-flow events per year, from late spring to late summer. Beginning in 2011, sensors devoted to debris-flow detection (geophones, video cameras, flow stage sensors) were installed along the main channel, upstream of a retention basin. In case of debris flows, high-resolution topographical surveys of the retention basin are carried out multiple times per year. Rainfall is measured in the lower part of the catchment and at the

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 95–111

Coviello, Theule, Crema, Arattano, Comiti, Cavalli, Lucía, Macconi, and Marchi

headwaters, while passive integrated transponder tracing of bedload was performed in the main channel during spring and summer 2014. In this work, we present the reconstruction of the sediment dynamics at the catchment scale from 2011 to 2017. Results show that (i) coarse sediment yield is dominated by the few debris flows occurring per year; (ii) debris-flow volume estimations may be significantly different—up to 30 percent lower—when performed through a digital elevation model of difference analysis, compared to the timeintegration of the debris-flow discharge estimates; (iii) using this latter method, the volumes are affected by significant uncertainties, particularly for small values of flow depth; and (iv) rainfall analysis permits us to characterize debris-flow initiation but also highlights difficulties in discriminating triggering from non-triggering rainstorms if based on rainfall duration and intensity only.

INTRODUCTION Massive, abrupt sediment inputs—typical of debris flows and large floods occurring in steep channels— dramatically transform the morphology of mountain areas. The upstream edge of the alluvial fans may act as a bedload trap, creating longitudinal discontinuities in sediment transport and causing large-scale aggradation of sediment (Hoffman and Gabet, 2007). The sediment supply to the channel network is typically episodic, controlled by the interaction between geomorphic conditions and hydrological processes (Benda and Dunne, 1997). In the Alps, one of the largest, most recent perturbations of sediment supply has been the Last Glacial Maximum, which dramatically altered landslide activity and sediment yields during the postglacial period (Cossart et al., 2008; Savi et al., 2014). In the Vinschgau-Venosta Valley (eastern Italian Alps), intense debris-flow activity altered sediment continuity along the Adige River during the last deglaciation, imposing large-scale bed aggradation, confluence migration, and channel obstructions, with the formation of temporary lakes and fan-delta systems (Brardinoni et al., 2018). Understanding the effect of debris flows and bedload-transporting events on channel topography, as well as quantifying sediment yield, is of paramount relevance for hazard assessment and design of mitigation measures. Climatic change represents a further challenge for recognizing current and future trends in sediment transport and, thus, design of adequate control structures. However, the quantification of the sediment cascade associated with both debris flows and bedload events, and their relative yields at the catchment scale, has been rarely addressed in an integrated way.

In recent years, the increase in topographic instrument automation and resolution has significantly improved the cost-effectiveness of multi-temporal analysis of digital elevation models (DEMs). The geomorphic changes associated with erosion/deposition processes can be quantified through “DEM of difference” (DoD) grids, in which the elevation difference between new and old surfaces permits assessment of erosion and deposition patterns within the catchment along with the sediment output from the catchment in the time interval between two DEMs (Schürch et al., 2011; Theule et al., 2015; Cavalli et al., 2017; and Papa et al., 2018). However, discriminating both the contribution of the different sediment transport processes (namely debris flows vs. bedload) and the effects of multiple flow events is often challenging. Thus, long-term instrumental monitoring of sediment fluxes through catchment-scale sensor networks can provide precious information, especially if coupled with highresolution topographical surveys (McCoy et al., 2010; Comiti et al., 2014). The measurement of rainfall permits linking the occurrence of the sediment fluxes to their most common triggering factor. In catchments instrumented for debris-flow monitoring, the identification of critical rainfall thresholds offers insights into the interactions among climate, sediment supply processes, and debris-flow occurrence (Hürlimann et al., 2019). The Gadria catchment (eastern Italian Alps) offers the opportunity to understand the main processes driving sediment supply at the catchment scale thanks to the detailed, ongoing monitoring activities occurring in this area. In this article, we present the catalog of the flow events recorded in the Gadria between 2011 and 2017, when frequent field surveys were carried out. Specific objectives of the study are (i) the analysis of rainfall generating different processes in the channel (bedload transport and debris flows), (ii) the assessment of the differences in distance travelled by particles transported by floods with bedload and by debris flows, and (iii) the assessment of debris-flow volumes both using DoD and monitoring data. STUDY SITE The Gadria catchment is located in the VinschgauVenosta Valley, eastern Italian Alps, and belongs to the Adige River basin (Figure 1a). At the fan apex, the catchment has a drainage area of 6.3 km2 and ranges in elevation from 1,394 to 2,945 m a.s.l. (Figure 1b). The Gadria is underlain by paragneiss and orthogneiss lithologies from the Permian and Cretaceous metamorphisms. Sediment produced by the weathering of these highly fractured rocks and thick Quaternary deposits fills the channel network, generally through shallow debris slides, rockfalls, and dry

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 95–111

Estimation of Sediment Yield in a Debris-Flow Catchment

Figure 1. (a) Location of the Gadria instrumented catchment in the Vinschgau-Venosta Valley, Province of Bozen-Bolzano, Italy; (b) map of the catchment and of the alluvial fan; and (c) details of the instruments installed at the lower monitoring station.

raveling on the steep slopes (Figure 2a). These colluvial processes that dominate the upper and intermediate sections of the basin, along with the presence of steep channels, create the perfect conditions for chronic and frequent debris-ﬂow activity (Figure 2b). The Gadria catchment is characterized by dry innerAlpine climate, with mean annual precipitation of 480 mm in the Venosta Valley ﬂoor (weather sta-

tion Laas-Lasa, 863 m a.s.l., 1989–2012 period), due to the sheltering effect of the mountainous ranges to southerly and northerly winds. Mean annual precipitation increases with altitude, with 662 mm measured at a rain gauge located at 1,754 m a.s.l. (1993–2012 period). Annually, the Gadria features on average from one to two en masse processes (i.e., debris flows) and several small floods causing bedload transport.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 95–111

Coviello, Theule, Crema, Arattano, Comiti, Cavalli, Lucía, Macconi, and Marchi

METHODS Instrumental Monitoring

Figure 2. (a) View of one of the main sediment source areas of the Gadria catchment (2,200 m a.s.l.); (b) the main channel in the intermediate sector of the catchment (1,500 m a.s.l.); (c) lower monitoring station; and (d) the retention basin located at the catchment outlet.

Several consolidation check-dams have been built along the Gadria, and a retention basin was constructed in the 1970s to protect settlements located on the Gadria fan (Figure 2d). This basin is periodically emptied, but some residual risk still exists as extremely large debris ﬂows could exceed its capacity.

In the Gadria basin, the first downstream monitoring station (“lower station”) was equipped in 2011 close to the fan apex, at an elevation of about 1,400 m a.s.l. (Comiti et al., 2014). This installation was designed for the measurement of basic debris-flow variables, the characterization of flow dynamics, and the development of early warning systems. The station is composed of three video cameras framing the channel and the retention basin, four vertical geophones (10 Hz) placed along the left channel bank (two in the ground and two on the wing of a consolidation check-dam), and two radar sensors for the measurement of the flow stage at the same cross sections where the geophones are installed (Figures 1c and 2c). The distance between the cross sections monitored with the stage sensors is 76 m. In 2013, the geophone array was extended in the upstream direction with three additional geophones installed along the main channel (Coviello et al., 2015). This latter geophone network (Figure 1c) recorded the seismic data that are analyzed in the present work. The geophone signal is acquired by a 24-bit precision analog-to-digital (AD) converter unit with a sampling rate of 128 Hz. Each AD converter is wired to the recording unit, and power is supplied by solar panels. In normal water flow conditions, the monitoring system records a limited number of variables per second (amplitude envelope, maximum and minimum value of the raw signal, main frequency, band width). Automatic debris-flow detection is performed using a detection algorithm based on the ratio of the short time average (STA) over the long time average (LTA), calculated in real time on the seismic signal gathered by the linear array of geophones previously mentioned. Both the geophone network and the STA/LTA detection algorithm are described in detail in Coviello et al. (2019). The average slope of the channel reach monitored by the stage sensors and the geophones at the lower station is 16 percent. A rain gauge and a further stage sensor (D4 and R3 in Figure 1c) are located about 500 m upstream, at an elevation of 1,500 m a.s.l. The video footage has been used to assess the surface velocity of debris flows through the application of a large-scale particle image velocimetry technique (Theule et al., 2018). The upstream monitored areas are located at an elevation of about 2,200 m a.s.l., with the objective of monitoring initiation conditions through the use of rainfall measurements (rain gauges R1 and R2, Figure 1b). In 2018, a monitoring network composed of rain-triggered video cameras, geophones, and soil moisture probes was installed on a ridge separating

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 95–111

Estimation of Sediment Yield in a Debris-Flow Catchment

two steep channels to detect triggering processes related to debris-ﬂow initiation (Coviello et al., 2020).

RESULTS AND DISCUSSION Catalog of Flow Events

Topographical Surveying and Passive Integrated Transponder (PIT) Tracing Repeated topographic surveys of the retention basin at the beginning of the debris-flow season and after each debris flow were carried out by terrestrial laser scanning (TLS) from 2011 to 2013. After this period, Structure from Motion (SfM) photogrammetry was adopted because this method is more efficient, resulting in point clouds with better spatial coverage. Photos were processed using Agisoft Photoscan Professional. From 2013 to 2015, an operator took photos using a Canon EOS 700D (18-mm focal length) from a 6-m extendable pole at one frame per second while walking along the channel banks (Figure 3a). The operator would stop every 3 to 6 m, turning the pole to capture approximately five photos to cover all view angles at each stop. From 2015 on, photos were taken from a helicopter (using a Canon EOS 700D, Nikon D4S, and Canon 5D Mark IV) flying approximately 100–600 m over the retention basin, channel, and source areas, and also from a drone (DJI sensor FC300S) for just the retention basin. Painted reference points were appropriately distributed around the retention basin and channels and were surveyed with the total station and differential GPS (dGPS). For DoD analysis, CloudCompare freeware was used to further align both TLS and SfM point clouds using the iterative closest point algorithm. This was applied to unchanged permanent features, resulting in root mean square errors of 21 cm for older TLS comparisons (volume uncertainty for the retention basin of ±740 m3 ), 2 cm for extendable pole SfM (±100 m3 ), and 9 cm for helicopter SfM (±450 m3 ) for the retention basin area. Ten-centimeter DEMs were developed, and their differences were used for measuring the volumes in the retention basin (Figure 3b). In the main channel, 280 PITs were installed in 2014 and their positions measured using dGPS to contrast incipient motion and transport distances between debris flows and bedload events. The horizontal accuracy in the position of PIT-tagged clasts is of the same order of magnitude as their diameter (small cobbles to small boulders; i.e., 0.05 to 1 m). The grain size of PIT-tagged clasts and their degree of embeddedness (percent) were estimated in the field. They were distributed throughout the channel at upstream distances of 1,253–1,439 m, 675–1,061 m, and 178–236 m from the retention basin outlet. Field checks and antennae surveys took place after floods with bedload transport and after debris flows.

Different sediment transport processes occurred at Gadria from the beginning of the monitoring activities in 2011; while debris flows were documented throughout the whole monitoring period, floods featuring bedload transport were only systematically detected from 2014 to 2016 (Table 1). We constructed the event catalog of debris flows and floods featuring bedload transport based on the analysis of the geophone data set gathered at the lower station. Complementary data (rainfall, flow stage measurements, video recordings) ensured the complete event characterization. Compared to debris flows, seismic signals produced by bedload-transporting events present significantly longer durations and lower amplitude peaks (Figure 4) (Mao et al., 2009; Coviello et al., 2015; and Bel, 2017). Flood events with bedload transport that occurred in 2014 and 2015 were identified by analyzing 140 days of continuous seismic recordings from the geophone network installed at the lower station. Over the course of 2 years, nine bedload events were detected using an intensity-duration threshold (amplitude above the long-time-average of the seismic signal for at least 10 minutes). The event detection was validated through the manual inspection of the video frames, when available, and by the analysis of rainfall events recorded at the upper station. An image every 5 minutes was recorded by the video camera framing the channel in the upstream direction (video frames available from 4 June to 18 July 2014, from 1 May to 23 October 2015, and from 14 April to 28 June 2016). In addition, seven additional bedload events were identified by inspecting the images recorded in the periods of time lacking seismic records. Finally, two snowmelt-induced bedload events (20 and 21 May 2015) were directly observed in the field, with bedload rates measured by portable traps (Bunte et al., 2004). During most of the field season in 2016, very few seismic data were recorded because of technical issues. However, the STA/LTA detection algorithm was working in real time, and the video images gave a visual validation of the warning outcome (Figure 5). The duration and the main-front velocity of the debrisflow event that occurred on 12 July 2016 were inferred from the video footages (Tables 1 and 2). The alarm activated 40 seconds before the first surge by a precursory, liquid surge and lasted for the whole duration of the first surge. Subsequently, it switched off and reactivated in correspondence with the passage of a secondary surge. The debris flow of 26 July 2016 occurred during the maintenance of the video system. In this case, videos of the debris flow were

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 95–111

Coviello, Theule, Crema, Arattano, Comiti, Cavalli, Lucía, Macconi, and Marchi

Figure 3. Photogrammetric surveying of (a) the main channel with the 6-m extendable pole; (b) deposition in the retention basin retrieved from the diﬀerence of DEMs (DoD) of two photogrammetric surveys carried out in April and August 2014; and (c) three-dimensional view of the August 2017 DoD with the post-event point cloud.

100

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 95–111

Estimation of Sediment Yield in a Debris-Flow Catchment Table 1. Debris flows and flood events with bedload transport detected at the lower station of Gadria. Dates of topographic and passive integrated transponder (PIT) surveys carried out from 2014 to 2017 are in italics. Peak times correspond to the amplitude peaks of the seismic signal. Flow events characterized only with images or direct observations are marked with *. For 2017, only the information on debris-flow occurrence is available.

Date

Peak Time (hh.mm UTC+2)

Peak Amplitude at Geo3 (µm/s)

Duration (min)

Bedload Bedload Bedload

05:00 05:00 18:00

— — 2

>120 >120 360

Bedload Bedload Debris ﬂow

14:55 20:30 17:13

4 — 200

360 >180 26

Bedload Bedload Bedload

07:45 13:32 12:00

3 7 2

360 120 180

Bedload Bedload Bedload Debris ﬂow

13:00 14:00 15:23 17:16

12 173

>600 600 >240 50

Bedload Bedload Bedload Bedload Bedload Bedload

8:30 15:00 18:40 21:20 15:52 14:46

— 16 13 14 — —

>240 180 >300 >180 >180 >180

Bedload Debris ﬂow Debris ﬂow

10:00 19:00 12:48

— — 50

180 30 15

Debris ﬂow

19:25

Debris ﬂow Debris ﬂow

21:05 02:30

— —

20 20

Typology

10 Apr 2014 topographic survey 9 May 2014 PIT survey 5 Jun 2014* 9 Jun 2014* 29 Jun 2014 2 Jul 2014 PIT survey 8 Jul 2014 13 Jul 2014* 15 Jul 2014 19 Jul 2014 PIT survey 21 Jul 2014 24 Jul 2014 13 Aug 2014 18 Aug 2014 topographic survey 18 Apr 2015 topographic survey 20 May 2015* 21 May 2015* 6 Jun 2015 8 Jun 2015 10 Jun 2015 topographic survey 16 Jun 2015* 29 Jul 2015 4 Aug 2015 7 Aug 2015 2 Jun 2016* 3 Jun 2016* 10 Jun 2016 topographic survey 15 Jun 2016* 12 Jul 2016* 26 Jul 2016 21 Oct 2016 topographic survey 10 Jul 2017 13 Jul 2017 topographic survey 2 Aug 2017 topographic survey 8 Aug 2017 19 Aug 2017* 31 Aug 2017 topographic survey

taken directly by some of the authors. In 2017, instrumental bias compromised geophone recordings and storage, making data analysis complicated and misleading. Table 3 provides a summary of the available instrumental data set in the investigated time windows (2011–2017). Rainfall Analysis Data recorded at the rain gauges R1 and R2, located in the upper sector of the catchment (Figure 1b), were analyzed to identify the characteristics of the rainstorms that triggered the debris ﬂows and to compare them with non-triggering rainfall. We computed the mean intensity, I (mm/hr), and the total duration,

D (in hours), of the triggering rainstorms (from the onset of the rainstorm to the passage of the debris ﬂow at the lower station) from 2011 to 2017. When data from both rain gauges were available, only the one that corresponded to the most severe conditions for debris-ﬂow triggering was retained (Figure 6a). The minimum empirical I-D threshold for triggering (Hürlimann et al., 2019) is I = 8(D)−0.96 .

(1)

The rainfall intensity recorded at R2 for the debris ﬂow of 26 July 2016 plots well below the I-D threshold, while R1 was not functioning. R2 was likely not able

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 95–111

101

Coviello, Theule, Crema, Arattano, Comiti, Cavalli, Lucía, Macconi, and Marchi

Figure 4. Geophone signals of the (a) debris flow of 15 July 2014 and (b) flood with bedload transport of 8 July 2014. The peak amplitudes of the debris flow are about one order of magnitude greater than the flows induced by the bedload event, while its duration is significantly lower (30 minutes vs. 6 hours).

to catch the rainstorm that triggered this debris flow, as the storm was rather small in terms of spatial extent. To investigate both triggering and non-triggering events, an objective procedure for extracting rainfall events was defined. For the computation of event duration and rainfall depth, start and end time of the rainfall events have been identified by requiring a hia-

tus of at least 6 hours, with maximum rain of 0.2 mm, to separate two consecutive rainfall events. In the case of debris-flow triggering rainstorms, this approach to the assessment of rainfall event recognition leads to longer duration than is associated with those from the start of the precipitation to the occurrence of the debris flow, which was considered for the identification of the

Figure 5. Debris flow of 12 July 2016. Video frames are of the precursory surge (a), the main front (b), and of the tail of the first surge (c), the inter-surge (d), the second surge (e), and the end of the flow (f).

102

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 95–111

Estimation of Sediment Yield in a Debris-Flow Catchment

rainfall threshold (Eq. 1). Only the months from May to September, which include the typical debris-flow season, have been analyzed. This approach permitted the isolation of 228 rainfall events for the R1 rain gauge for the June 2013 through June 2016 period and 184 rainfall events for the R2 rain gauge for the May 2014 through July 2016 period (2017 was not considered because there were too many gaps in data at both rain gauges). We decided to focus on the R1 rain gauge (Figure 6b) because it is closer to the main debris-flow initiation areas. The boxplots reported in Figure 7a and b compare mean intensity and duration for debris flow, bedload, and non-triggering rainfall recorded at the R1 rain gauge. Despite the small number of debris flows for which continuous rainfall data are available, a general pattern in terms of rainfall event typology can be identified. In particular, debris-flow events are characterized by significantly higher mean intensity than are the other two classes. Bedload events are characterized by longer duration (∼2×) with respect to the average duration of non-triggering rainfall events, which is almost coincident with the median duration of debris-flow triggering rainstorms. Long rainfall duration seems to be the dominant predisposing factor for the occurrence of bedload events. Looking at the whole point clouds distribution, the same patterns can be noticed in Figure 6b.

load transport. The log-log plot of travel distance versus grain-size diameter shows an inverse correlation for the particles transported during the first, bedloadonly period (Figure 8). On the contrary, no clear relationship can be detected for the second period when a debris flow occurred, as it could be anticipated based on the transport en masse of sediment by debris flows (Theule et al., 2015). Thus, in the second time interval, the longer displacement lengths are most likely produced by the debris flow that occurred on 15 July 2014. Indeed, PIT-tag measurements in debris-flow channels are mainly useful for analyzing the variability in clast entrainment (e.g., based on clast position along cross sections and thus on the imparted shear stress) rather than transport distances. In addition, PIT-tracing installation is very resource-intensive, and manual surveying of the instruments is time-consuming; thus, the recovery rate can be very low. The 2014 debris flow had a 29 percent recovery rate due to the depth of the pit-tag deposits, which we assume are mostly buried in the sediment trap. However, more significant results could be achieved if the travel distance of the tagged particles is measured in a debris-flow channel, which, differently from the Gadria, is not trapped by a sediment retention basin (Bel, 2016). In fact, the maximum travel distance (slightly smaller than 1,400 m) is close to the distance between the most upstream transect equipped with PIT tags and the retention basin (Figure 8).

Particle Motion In 2014, three surveys of PIT positions were carried out on 9 May, 2 July, and 19 July. In the time interval from 9 May through 2 July, three flood events featuring bedload transport were observed, while in the following time interval (2 through 19 July), two bedload events and one debris flow occurred (Table 1). Consequently, we can attribute the displacement lengths measured after the first time interval to bed-

Debris-Flow Volumes Table 2 reports the volume of sediment deposited by debris flows in the retention basin. Video recordings and post-event observations show that the trap efficiency of the retention basin (ratio of sediment volume retained to the total incoming sediment) is significantly below 100 percent (Comiti et al., 2014). Thus, we estimate that from 10 percent to 20 percent of sediment

Table 2. Debris-ﬂow events that occurred at Gadria from 2011 to 2017; volumes computed using stage sensor measurements integrated in time are compared with digital elevation model of diﬀerence (DoD) analysis of the retention basin.

Date

Debris-Flow Peak Discharge (m3 /s)

Debris-Flow Volume (m3 )

5 August 2011 18 July 2013

11 80

2,400 7,550 (disturbed)

15 July 2014 8 June 2015 12 July 2016 26 July 2016 10 July 2017 8 August 2017 19 August 2017

26 27 18 3.2 1.1 — 9.7

11,600 12,600 — — 220 — 860

Time Interval DoD June–September 2011 June 2011 (empty trap)–August 2013 April–August 2014 April–June 2015 June–July 2016 July 2016 October 2016–July 2017 Beginning–end of August 2017

DoD (m3 )

Reference

2,000 8,100

Comiti et al., 2014 Arattano et al., 2015

10,400 9,850 2,400 1,000 >700 (disturbed) 4,600

This study This study This study This study This study This study

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 95–111

103

104

ࣹ ࣹ • ࣹ • • • ࣹ x x x x

x indicates entire or most part of the month; ࣹ = few days (no debris-ﬂow events); and • = few days (including debris-ﬂow events). *

x x x ࣹ

x ࣹ

x x x x

x • x • x x

x ࣹ x x x x x x x • x x x ࣹ x x x

x ࣹ x x x x x x x x x x • x ࣹ x x x x

ࣹ x x x • x x • ࣹ Flow stage Ground vibration Rainfall (R1) Video images

Apr May Jun Jul Aug Sep Oct Apr May Jun Jul Aug Sep Oct Apr May Jun Jul Aug Sep Oct Apr May Jun Jul Aug Sep Oct Apr May Jun Jul Aug Sep Oct

2017 2016 2015 2014 2013

Table 3. Available data set and analysis performed in the investigated time window (2011–2017). For 2011, we only dispose of video images and stage measurements of the debris ﬂow of 5 August. No debris ﬂows occurred in 2012.*

Coviello, Theule, Crema, Arattano, Comiti, Cavalli, Lucía, Macconi, and Marchi

Figure 6. Debris-flow triggering rainfall characterized from the onset of the rainstorm to the passage of the debris flow at the lower station (a) and (b) automatically extracted including the whole event, processing the June 2013 through June 2016 R1 continuous data series. The empirical intensity-duration (I-D) rainfall threshold for debris-flow triggering in the Gadria catchment is reported in (a), where data from the rain gauge featuring the most severe I-D conditions for debris-flow triggering were selected (see legend).

flowed through the slit opening of the retention check dam. In addition, bedload transport contributes to the sediment yield, as well as to the erosion of deposits, in both the retention basin and the channel network (Figure 9). Specifically, eight bedload events occurred in the Gadria catchment in the period between the two surveys in 2014—three in 2015 and one in 2016. In 2016, one event (2 June) transported up to 500 m3 of fine sediment and wood, based on video images and post-event DEM (Figure 10). The volume of debris flows was calculated using data from the monitoring station. The velocity of each debris-flow surge (Table 4) was estimated considering the mean propagation velocity of each front as the ratio of the distance between two equipped cross sections (stage sensors D2 and D3, see Figure 2) to the time interval between the arrival of the debris-flow surge at the two stations (Arattano et al., 2015). This velocity estimation neglects that the velocity profiles of

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 95–111

Estimation of Sediment Yield in a Debris-Flow Catchment

Figure 8. Passive integrated transponder (PIT) tracing at Gadria in 2014, from 9 May to 2 July (only bedload transport observed) and from 2 July to 19 July (one debris ﬂow). The upper travel-distance limit (dashed line) is the distance between the most upstream transect equipped with PIT tags and the retention basin.

Figure 7. Boxplot of (a) mean intensity values and (b) duration values for debris ﬂow, bedload and non-triggering rainfall events at the R1 rain gauge. On each box, the central mark indicates the median, and the bottom and top edges of the box indicate the 25th and 75th percentiles, respectively. The whiskers extend to the most extreme data points not considered outliers, with the outliers plotted individually. Outliers are identiﬁed as values more than 1.5 times the interquartile range away from 25th or 75th percentile.

high-density flows are generally not constant along the flow depth (Lanzoni et al., 2017; Sarno et al., 2018; and Nagl et al., 2020). The discharge of each debris flow is computed as the product of surge velocity by the flow cross-sectional area, estimated using the flow stage measurement. Freezing the flow cross-section area during each event is the main possible source of uncertainty of this method, as the cross-sectional area

Figure 9. Video frames of the channel and retention basin on 1 July (a, b), on 15 July (c, d) after three bedload events, and on 16 July 2014 (e, f) after a debris-flow event. Erosion of fine to medium-size material due to bedload events and coarse deposits produced by the debris flow in the main channel are highlighted in (c) and (e), respectively.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 95–111

105

Coviello, Theule, Crema, Arattano, Comiti, Cavalli, Lucía, Macconi, and Marchi

Figure 10. The channel and the retention basin on 2 June at the beginning (a, b) and at the peak (c, d) of the bedload event and on 3 June 2016 (e, f), with sediment and wood deposits.

Table 4. Peak velocity and volume of debris-ﬂow events, separated in surges, occurring at Gadria from 2011 to 2016. Some of the surges were aggregated for the volume estimation: in such cases, the peak velocities of the aggregated surges are shown as used in the volume estimations. For some events that occurred in 2016 and 2017, surge velocity and volume were not calculated because of lack of ﬂow stage data.

Date 5 August 2011

18 July 2013

15 July 2014

8 June 2015

12 July 2016 26 July 2016 10 July 2017 8 August 2017 19 August 2017

106

Debris-Flow Surges

Surge Velocity (m/s)

Surge Volume (m3 )

Surge I Surge II Surge III Main front

2.6 1.7 1 5.7

1,749 249 418 3,600

Surge II Surge III Surge IV Surge V Pre-surge Main front + III Surge IV + V Surge VI Surge VII Pre-surge Main front + III Surge IV Surge V Surge VI Surge VII + VIII Surge I Surge II Surge I Surge I — Surge I + II

4.4 5.4 3.6 1.2

850 1,000 1,800 300 — 3,400 6,000 1,750 500 — 3,500 2,650 850 850 4,750 — — — 220 — 860

1.4, 5.0 3.1, 5.0 2.9 1.5 2.4, 3.3 3.3 1.5 1.2 3.0, 1.9 — — — 1.14 — 1.74

Reference Comiti et al., 2014

Arattano et al., 2015; Coviello et al., 2019; This study

Coviello et al., 2019; This study

This study This study This study This study This study

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 95–111

Estimation of Sediment Yield in a Debris-Flow Catchment

may change during a single debris ﬂow as a result of erosion/deposition processes. The instrumented channel reach of the Gadria, however, is rather stable: the consolidation check dams (Figures 2c and 3a) prevent channel bed incision, while deposition is limited to the formation of lateral levees that show small to moderate topographic diﬀerences from event to event. The method is also sensitive to the choice of the hydrometric datum adopted to perform the calculations. Finally, the bulked volume carried by each surge (Table 4) was calculated as Volume = v ×

A (h (t)) ,

(2)

where A(t) is the cross-sectional area at the time t; v is flow velocity of the surge; and t0 and te represent the initial and final time of the surge, respectively. We calculate the volumes of the surges that follow the main front for the debris flows that occurred in 2013, 2014, and 2015 (Table 4). The calculations are based on flow stage measurements gathered at station D3 (Figure 1b). There is some subjectivity in the definition of a surge and in the choice of the beginning and end of each surge. In some cases, we observe multiple little waves, one following the other, and some of them are very small (e.g., surge IV in 2013 event, surge V in 2014 event, and surge VII in 2015; see Figure 11). In other cases, the hydrograph shows a rising limb that increases its height slowly before reaching the peak (e.g., surge V in 2013, surge IV in 2014, and surge VIII in 2015; see Figure 11). This increases the uncertainty in the estimation of the surge volume because the peak velocity of the slow-rising limb has to be assigned as well. For these reasons, a surge-by-surge volume estimation would be quite uncertain, and we deem it more appropriate to lump multiple surges, differently from what we did in previous analysis (Coviello et al., 2019). In Table 4 the peak velocities used for volume estimations are shown: in some cases, even though some surges were lumped into a single wave, we report the peak velocities of the main aggregated surges used to estimate the debris-flow volume. In Figure 11 we report the partition into surges of these three events and the aggregation we used for volume estimation. For the following years, we could calculate the volume for only two events, 10 July and 19 August 2017. DoD and Hydrograph-Based Estimates of Sediment Yield Volume estimations may significantly differ if performed through the DoD of the retention basin or with instrumental measurements carried out with the use of flow stage data (Table 2). Most estimates

Table 5. Sensitivity analysis on the volume estimation based on the product of surge velocity by the cross-sectional area of the flow, calculated using the flow stage measurements. Three parameters control the calculation: the flow height, the velocity, and the cross-sectional area. The three parameters were perturbed by 5%, 10%, and 15%, respectively (positively for flow height and velocity, negatively for the cross-sectional area). The volumes of the debris flows of 18 July 2013 (medium-size event) and 19 August 2017 (small event) were recalculated. Results are presented in percentage of variation compared to the original value.

Parameter Perturbation 5% Volume variation, 18 July 2013 Volume variation, 19 August 2017 Perturbation 10% Volume variation, 18 July 2013 Volume variation, 19 August 2017 Perturbation 15% Volume variation, 18 July 2013 Volume variation, 19 August 2017

Velocity (%)

Flow Height (%)

Area (%)

5.0 15.2

7.7 9.7

−31.1 −43.2

10.0 31.7

15.5 19.7

−41.7 −55.3

15.0 49.5

23.5 30.1

−43.7 −57.1

deriving from topographic surveys resulted in consistently lower volumes, up to 30 percent, especially for larger and long-lasting debris flows. This is consistent with the observed outflow of the fine sediment through the check dam during the tail of each debris-flow event. In addition, volume estimations performed using Eq. 2 represent a wet volume that includes the water volume as well, whereas the DoD yields the bulk volume of the deposits. For the DoD of the 18 July 2013 debris flow, the empty retention-basin surface surveyed in June 2011 through TLS had to be used for the 2013 pre-event surface as a result of the sediment removal carried out in summer 2013, before the event. Based on field observations, we estimate that the amount of sediment removed was approximately equal to the deposit of the 2011 event (i.e., 2,000 m³). However, for this event we consider an uncertainty of about 1,000 m³ for the DoD volume estimation. For the event of 19 August 2017, the calculated volume (860 m³) results are much smaller than the volume estimated through the DoD (2,300 m³). This effect is due to the simplification of using a stable cross-sectional area, which produces a large error for small values of the flow height. By means of a sensitivity analysis of the parameters used for the volume estimation, we show that the volume estimation is very sensitive to small variation of the cross-sectional area (Table 5). The three parameters controlling Eq. 2 (flow height, flow velocity, and the cross-sectional area) were perturbed by 5 percent, 10 percent, and 15 percent, respectively (positively for flow height and velocity, negatively for the cross-sectional area), and the volume of the debris flow of 18 July 2013 was re-calculated. In 2013, the measured cross section before and after the debris flow

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 95–111

107

Coviello, Theule, Crema, Arattano, Comiti, Cavalli, Lucía, Macconi, and Marchi

Figure 11. Partition in surges of the debris-ﬂow hydrographs recorded at station D3 and aggregation used for the volume estimations.

differed by about 0.15 m on the right side of the channel, resulting in a possible reduction of the crosssectional area of about 0.5 m2 . For the perturbation of the cross-sectional area, we hypothesized that a deposit larger than the latter one (+5 percent, +10 percent, and +15 percent, respectively) occupied part of the cross-sectional area. Another possible source of uncertainty is related to the actual behavior of the debris flow along the channel reach between the two sensors D2 and D3. In case of local flow deceleration, the velocity based on travel time (celerity) is highly affected, thereby leading to a possible underestimation of debris-flow volumes. However, such a behavior should affect the hydrograph shape at the two stations D3 and D2. Indeed, the recorded hydrographs show a debris-flow front at station D2 that is significantly higher (by almost 20 percent) than the front height at station D3 (Figure 12). This contrasts with the hydrograph deformation normally occurring between two sections

108

according to the kinematic behavior of debris-flow waves (Arattano and Savage, 1994), as also observed in the Gadria during 2011 and 2013 debris-flow events (Comiti et al., 2014). When the volume of debris flows is calculated using data from a pair of stage sensors, applying the formula provided by Eq. 2, a check of the hydrographs at the two stations should always be carried out to verify possible hydrograph deformations due to flow deceleration along the channel reach between the stations. Such a possibility is more common for debris flows of small magnitude that may not have enough energy to flow regularly along the channel. In the neighboring Strimm catchment, a bedload yield of 200 m3 yr−1 was observed during 2 years of PIT tracing carried out from 2011 to 2013 (Dell’Agnese et al., 2015), which also contributes to the filling-up of the retention basin. Neglecting the contribution of the Strimm, a sediment yield of about 1,000 m3 km−2 yr−1 was calculated based on DoD

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 95–111

Estimation of Sediment Yield in a Debris-Flow Catchment

Figure 12. Hydrographs of the debris ﬂow that occurred on 19 August 2017.

analysis in the period of time from 2011 through 2017, which is largely dominated by debris-flow processes, considering both preliminary estimates from bedload virtual velocities and debris-flow volumes stemming from Eq. 2. The sediment yield value is significantly lower than the estimation made for the time period 2005–2011 (5,200 m3 km−2 yr−1 ), during which much larger debris-flow events (up to 35,000 m3 ) were observed (Cavalli et al., 2017). When comparing the specific sediment yields of the two periods, it is worth mentioning the different approaches adopted in computing the DoDs. For the 2005–2011 time window, a single DoD was obtained by subtracting two LiDAR DTMs covering the whole catchment. DoD 2015–2011 was thus able to account for sediment output due to intra–debris-flows event sediment transport (Cavalli et al., 2017). Since 2011, a DoD was computed for each individual debris-flow event, hence focusing the analysis on sediment transported during single events. Although much lower than the sediment yield computed for the 2005–2011 period, the 2011–2017 sediment yield estimation confirms that the Gadria catchment currently is one of the most active debris-flow catchments at the regional level, as noted by Brardinoni et al. (2012) from the analysis of a historical database of debris flows that occurred in the 1998– 2009 time interval.

CONCLUSIONS We constructed a catalog of debris ﬂows and ﬂoods featuring bedload transport by analyzing the monitoring data set (geophone data, video images) gathered at the experimental Gadria catchment, located in the eastern Italian Alps. Gaps in data, caused by both sensor malfunctions and problems with data storage, limited some analyses. These problems became especially

critical in 2017 and urged a revision of the system, which was implemented in 2018. The analysis of rainfall recorded at the catchment headwaters has allowed an empirical threshold for debris-flow initiation to be defined. Although debrisflow triggering rainstorms show higher mean intensity, the available data do not allow us to clearly differentiate rainstorm-determining bedload transport from those without bedload. This result confirms that rainfall analysis alone does not explain the occurrence and the type of sediment fluxes at the catchment scale. We argue that even in a catchment characterized by unlimited sediment supply, as in the Gadria, short-period variations in the amounts and location of sediment sources in the headwater channels influence debrisflow triggering and bedload initiation. In most cases, topographic surveys of debris-flow deposits provide volume estimations that were significantly lower—up to 30 percent lower—than those obtained through instrumental measurements carried out with the use of hydrograph data (product of each surge velocity by the cross-sectional flow area integrated over time). Volume estimations of small events (i.e., small values of flow height) are affected by significant uncertainties—by more than 50 percent. Sediment yield of about 1,000 m3 km−2 yr−1 is estimated for the investigated period (2011–2017), a value significantly lower than the one estimated in a previous time interval affected by larger debris flows. Sediment yield is dominated by debris flows, whereas the contribution of flood events featuring bedload transport is apparently very minor and mostly relative to the gravel and small cobble fractions only, based on preliminary data from travel distances of PIT-tagged clasts. The results achieved in the first years of monitoring in the Gadria catchment demonstrate the capability of a system based on multiple sensors installed at various

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 95–111

109

Coviello, Theule, Crema, Arattano, Comiti, Cavalli, Lucía, Macconi, and Marchi

stations in deciphering sediment dynamics at catchment scale. A longer period of observation is nonetheless needed to identify the factors (e.g., climate) controlling sediment yield variability over time. ACKNOWLEDGMENTS The Gadria monitoring station is managed by the Civil Protection Agency of the Autonomous Province of Bozen-Bolzano. The current research at Gadria is supported by the SEDIPLAN-r project (Sediment Budgeting and Planning for Rivers in South-Tyrol: From Hazard Mitigation to Environmental Restoration), funded by the European Regional Development Fund. This research was also supported by the Kinoflow project, funded by the Autonomous Province of Bozen-Bolzano. REFERENCES Arattano, M.; Bertoldi, G.; Cavalli, M.; Comiti, F.; D’Agostino, V.; and Theule, J., 2015, Comparison of methods and procedures for debris-flow volume estimation. In Lollino, G.; Arattano, M.; Rinaldi, M.; Giustolisi, O.; Marechal, J.-C.; and Grant, G. E. (Editors), Engineering Geology for Society and Territory: Springer International Publishing, Cham, pp. 115–119, doi:10.1007/978-3-319-09054-2_22. Arattano, M. and Savage, W. Z., 1994, Modelling debris flows as kinematic waves: Bulletin International Association Engineering Geology-Bulletin de l’Association Internationale Géologie l’Ingénieur, Vol. 49, No. 1, pp. 3–13, doi:10.1007/BF02594995. Benda, L. and Dunne, T., 1997, Stochastic forcing of sediment routing and storage in channel networks: Water Resources, Vol. 33, pp. 2865–2880, doi:10.1016/j.rmed.2004.03.018. Brardinoni, F.; Church, M.; Simoni, A.; and Macconi, P., 2012, Lithologic and glacially conditioned controls on regional debris-flow sediment dynamics: Geology, Vol. 40, pp. 455–458, doi:10.1130/G33106.1. Brardinoni, F.; Picotti, V.; Maraio, S.; Paolo Bruno, P.; Cucato, M.; Morelli, C.; and Mair, V., 2018, Postglacial evolution of a formerly glaciated valley: Reconstructing sediment supply, fan building, and confluence effects at the millennial time scale: GSA Bulletin, Vol. 130, pp. 1457–1473, doi:10.1130/B31924.1. Bunte, K.; Abt, S. R.; Potyondy, J. P.; and Ryan, S. E., 2004, Measurement of coarse gravel and cobble transport using portable bedload traps: Journal Hydraulic Engineering, Vol. 130, pp. 879–893, doi:10.1061/(ASCE)0733-9429(2004)130. Cavalli, M.; Goldin, B.; Comiti, F.; Brardinoni, F.; and Marchi, L., 2017, Assessment of erosion and deposition in steep mountain basins by differencing sequential digital terrain models: Geomorphology, Vol. 291, pp. 4–16, doi:10.1016/j.geomorph.2016.04.009. Comiti, F.; Marchi, L.; Macconi, P.; Arattano, M.; Bertoldi, G.; Borga, M.; Brardinoni, F.; Cavalli, M.; D’Agostino, V.; Penna, D.; and Theule, J., 2014, A new monitoring station for debris flows in the European Alps: First observations in the Gadria basin: Natural Hazards, Vol. 73, pp. 1175–1198, doi:10.1007/s11069-014-1088-5. Cossart, E.; Braucher, R.; Fort, M.; Bourlès, D. L.; and Carcaillet, J., 2008, Slope instability in relation to glacial de-

110

buttressing in alpine areas (Upper Durance catchment, southeastern France): Evidence from field data and 10Be cosmic ray exposure ages: Geomorphology, Vol. 95, pp. 3–26, doi:10.1016/j.geomorph.2006.12.022. Coviello, V.; Arattano, M.; Comiti, F.; Macconi, P.; and Marchi, L., 2019, Seismic characterization of debris flows: Insights into energy radiation and implications for warning: Journal Geophysical Research: Earth Surface, Vol. 124, pp. 1– 24, doi:10.1029/2018jf004683. Coviello, V.; Arattano, M.; and Turconi, L., 2015, Detecting torrential processes from a distance with a seismic monitoring network: Natural Hazards, Vol. 78, pp. 2055–2080, doi:10.1007/s11069-015-1819-2. Coviello, V.; Berti, M.; Marchi, L.; Comiti, F.; Marchetti, G.; Miyata, S.; and Macconi, P., 2020, Multi-parametric observations of debris-flow initiation at the headwaters of the Gadria catchment (eastern Italian Alps). In EGU General Assembly 2020, doi:10.5194/egusphere-egu2020-8720. Dell’Agnese, A.; Brardinoni, F.; Toro, M.; Mao, L.; Engel, M.; and Comiti, F., 2015, Bedload transport in a formerly glaciated mountain catchment constrained by particle tracking: Earth Surface Dynamics, Vol. 3, pp. 527–542, doi:10.5194/esurf-3-527-2015. Hoffman, D. F. and Gabet, E. J., 2007, Effects of sediment pulses on channel morphology in a gravel-bed river: Bulletin Geological Society America, Vol. 119, pp. 116–125, doi:10.1130/B25982.1. Hürlimann, M.; Coviello, V.; Bel, C.; Guo, X.; Berti, M.; Graf, C.; Hübl, J.; Miyata, S.; Smith, J. B.; and Yin, H. Y., 2019, Debris-flow monitoring and warning: Review and examples: Earth-Science Reviews, Vol. 199, pp. 102981, doi:10.1016/j.earscirev.2019.102981. Lanzoni, S.; Gregoretti, C.; and Stancanelli, L. M., 2017, Coarse-grained debris flow dynamics on erodible beds: Journal Geophysical Research: Earth Surface, Vol. 122, pp. 592–614, doi:10.1002/2016JF004046. Mao, L.; Cavalli, M.; Comiti, F.; Marchi, L.; Lenzi, M. A.; and Arattano, M., 2009, Sediment transfer processes in two Alpine catchments of contrasting morphological settings: Journal Hydrology, Vol. 364, pp. 88–98, doi:10.1016/j.jhydrol.2008.10.021. McCoy, S. W.; Kean, J. W.; Coe, J. A.; Staley, D. M.; Wasklewicz, T. A.; and Tucker, G. E., 2010, Evolution of a natural debris flow: In situ measurements of flow dynamics, video imagery, and terrestrial laser scanning: Geology, Vol. 38, pp. 735–738, doi:10.1130/G30928.1. Nagl, G.; Hübl, J.; and Kaitna, R., 2020, Velocity profiles and basal stresses in natural debris flows: Earth Surface Processes Landforms, Vol. 45, pp. 1764–1776, doi:10.1002/esp.4844. Papa, M. N.; Sarno, L.; Vitiello, F. S.; and Medina, V., 2018, Application of the 2D depth-averaged model, FLATModel, to pumiceous debris flows in the Amalfi Coast: Water, Vol. 10, doi:10.3390/w10091159. Sarno, L.; Carleo, L.; Papa, M. N.; and Villani, P., 2018, Experimental investigation on the effects of the fixed boundaries in channelized dry granular flows: Rock Mechanics Rock Engineering, Vol. 51, pp. 203–225, doi:10.1007/s00603-017-1311-2. Savi, S.; Norton, K. P.; Picotti, V.; Akçar, N.; Delunel, R.; Brardinoni, F.; Kubik, P.; and Schlunegger, F., 2014, Quantifying sediment supply at the end of the last glaciation: Dynamic reconstruction of an alpine debris-flow fan: Bulletin Geological Society America, Vol. 126, pp. 773–790, doi:10.1130/B30849.1. Schürch, P.; Densmore, A. L.; Rosser, N. J.; Lim, M.; and McArdell, B. W., 2011, Detection of surface change in

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 95–111

Estimation of Sediment Yield in a Debris-Flow Catchment complex topography using terrestrial laser scanning: Application to the Illgraben debris-ﬂow channel: Earth Surface Processes Landforms, Vol. 36, pp. 1847–1859, doi:10.1002/ esp.2206. Theule, J. I.; Crema, S.; Marchi, L.; Cavalli, M.; and Comiti, F., 2018, Exploiting LSPIV to assess debris-ﬂow velocities in

the field: Natural Hazards Earth System Sciences, Vol. 18, pp. 1–13, doi:10.5194/nhess-18-1-2018. Theule, J. I.; Liébault, F.; Laigle, D.; Loye, A.; and Jaboyedoff, M., 2015, Channel scour and fill by debris flows and bedload transport: Geomorphology, Vol. 243, pp. 92–105, doi:10.1016/j.geomorph.2015.05.003.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 95–111

111

Using High Sample Rate Lidar to Measure Debris-Flow Velocity and Surface Geometry FRANCIS K. RENGERS* U.S. Geological Survey, 1711 Illinois Street, Golden, CO 80401

THOMAS D. RAPSTINE U.S. Geological Survey, 1711 Illinois Street, Golden, CO 80401 Colorado School of Mines, 1500 Illinois Street, Golden, CO 80401

MICHAEL OLSEN Oregon State University, 140 Peavy Hall, 3100 SW Jeﬀerson Way, Corvallis, OR 97331

KATE E. ALLSTADT U.S. Geological Survey, 1711 Illinois Street, Golden, CO 80401

RICHARD M. IVERSON U.S. Geological Survey, 1300 SE Cardinal Court #100, Vancouver, WA 98683

BEN LESHCHINSKY Oregon State University, 140 Peavy Hall, 3100 SW Jeﬀerson Way, Corvallis, OR 97331

MACIEJ OBRYK U.S. Geological Survey, 1300 SE Cardinal Court #100, Vancouver, WA 98683

JOEL B. SMITH U.S. Geological Survey, 1711 Illinois Street, Golden, CO 80401

Key Terms: Debris-Flow, Lidar, Flume ABSTRACT Debris flows evolve in both time and space in complex ways, commonly starting as coherent failures but then quickly developing structures such as roll waves and surges. These processes are readily observed but difficult to study or quantify because of the speed at which they evolve. Many methods for studying debris flows consist of point measurements (e.g., flow height or basal stresses), which are inherently limited in spatial coverage and cannot fully characterize the spatiotemporal evolution of a flow. In this study, we use terrestrial lidar to measure debris-flow profiles at high sampling rates to examine debris-flow movement with high temporal and spatial precision and accuracy. We acquired measurements during gate-release experiments at the U.S.

*Corresponding author email: frengers@usgs.gov

Geological Survey debris-flow flume, a unique experimental facility where debris flows can be artificially generated at a large scale. A lidar scanner was used to record repeat topographic profiles of the moving debris flows along the length of the flume with a narrow swath width (∼1 mm) at a rate of 60 Hz. The highresolution lidar profiles enabled us to quantify flow front velocity of the debris flows and provided an unprecedented record of the development and evolution of the flow structure with a sub-second time resolution. The findings of this study demonstrate how to obtain quantitative measurements of debris-flow movement. In addition, the data help us to quantitatively define the development of a saltating debris-flow front and roll waves behind the debris-flow front. Such measurements may help constrain future modeling efforts. INTRODUCTION Debris flows are rapidly deforming mixtures of sediment and water that can lead to dramatic geomorphic

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 113–126

113

Rengers, Rapstine, Olsen, Allstadt, Iverson, Leshchinsky, Obryk, and Smith

change (e.g., Stiny, 1910; Anderson et al., 2015), as well as the destruction of life and infrastructure (Costa, 1984). Because these hazardous phenomena can be so deadly, it is important to understand the fundamental physical processes that govern debris-flow initiation, movement, and deposition. Much progress has been made toward understanding debris-flow physics through numerous experiments at the U.S. Geological Survey (USGS) debris-flow flume over several decades (Iverson et al., 2010). Debris-flow experiments in a large-scale flume can reproduce much of the behavior of natural debris flows (Iverson et al., 2010), which allow for the quantification of physical parameters that are difficult to measure in natural events. Such measurements are crucial for informing debris-flow models (George and Iverson, 2014; Iverson and George, 2014). However, even in a controlled flume setting, monitoring the continuous evolution of debris-flow movement is a challenge. In past experiments at the USGS debrisflow flume, mounted sensors at cross sections (as many as six in some experiments) were used to measure flow depths, pore pressures, and basal stresses, but these sensors gave only point measurements of flow properties. Recordings by video and time-lapse photography were used to measure flow-front velocities and understand how a debris flow evolves, but topographic information was limited to fixed points. In order to continue to make progress in understanding temporally variable debris-flow geometries, we designed a novel research experiment wherein we continually tracked the surface of a debris flow along its entire flow path, revealing the time–space evolution of the flow surface in high resolution. The ability to measure a near-continuous record of a debris-flow front has important implications for understanding debris-flow processes. For example, prior research has suggested that a debris-flow front can reach a steady state over a long, nearly constant slope (Hungr, 2000); however, surges are known to influence the shape and velocity at the front (Costa, 1984). A near-continuous dataset of debris-flow geometry could provide useful data to explore these ideas in detail. Moreover, understanding the evolving velocity and depth of debris flows has practical implications for understanding debris-flow runup on structures (Iverson et al., 2016) and debris-flow mitigation for structural designs (Prochaska et al., 2008). Finally, detailed surface topography at rapid time sequences could help to explain the depositional history of a debris-flow fan, which can be difficult to interpret because multiple surges can result in a complex fan stratigraphy where the layering does not correspond to individual surges (McCoy et al., 2010). In this study, we used a unique application of a terrestrial lidar scanner to measure the surface displace-

114

ment of two experimental debris flows at the USGS debris-flow flume. In recent years, terrestrial lidar has been shown to be a useful tool for monitoring landscape change with high precision in rapidly eroding landscapes (e.g., Staley et al., 2014; Rengers et al., 2016; and DeLong et al., 2018). These units are designed to obtain the three-dimensional (3D) geometry of all objects that are located within a panoramic view of a fixed position. Multi-temporal lidar studies, where researchers study a deforming surface with repeat static surveys, are typically on the order of multiple years (O’Neal and Pizzuto, 2011; Neugirg et al., 2016) or months (Rabatel et al., 2008; Rengers and Tucker, 2015; and Longoni et al., 2016). Recent advances have shown that valuable surface change information can be derived from repeat surveys at higher temporal resolutions using permanently mounted lidar units. For example, cliff retreat has been monitored with lidar to track rockfall at intervals as short as 1 hour (Williams et al., 2018, 2019). Moreover, swath lasers (Jacquemart et al., 2017; Takahashi et al., 2019) have been implemented to show topographic change through a portion of a channel cross section as debris flows move past. However, heretofore lidar has not been used to track a mass failure that is rapidly moving at speeds greater than 1 m/s. The most analogous prior applications were the use of lidar to track the shape and speed of sea waves (Brodie et al., 2015; Martins et al., 2017) and fully aerated, stepped water flows (Kramer et al., 2020). For this study, we modified a typical terrestrial lidar scanner to record two experimental debris flows. Instead of using the instrument to complete a panoramic scan from a fixed point, which requires several seconds to minutes per scan, we fixed its azimuth to capture high-frequency longitudinal profiles of the flume. This approach allowed us to capture the geometry of experimental debris flow every 0.017 seconds (∼60 Hz) as the flow traveled from the release point to the deposition point at the base of the flume. This time series of topographic profiles records the continuous evolution of the debris-flow surface as it moves downstream. Herein we describe how our new lidar acquisition technique can be used to gain increased understanding of fundamental debris-flow processes. By imaging an entire debris-flow surface compared with a few discrete locations, we show that it is possible to document the flow position and speed, to identify changes in flow characteristics (e.g., the development of a saltating front of coarse grains at the debris-flow front), and to characterize depositional fan evolution. This study is a demonstration of a new technical approach that can be used for monitoring the highly transient nature of debris flows. Consequently, it may become a

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 113–126

Debris-Flow Movement from Lidar

key tool to help researchers better constrain the evolution of flow behavior. METHODS Experimental Setup Experimental debris flows were created on May 24, 2017, and May 25, 2017, at the USGS debris-flow flume in the H. J. Andrews Experimental Forest near Blue River, Oregon. The flume is 95 m long, 2 m wide, and has a bed slope of 31° throughout most of its length. The experimental debris flows were mixtures of equal parts (50/50) sand and gravel, with initial volumes of 8 m3 and 10 m3 , respectively. In both scenarios, the sediment mixture was placed uphill of a gate and saturated with water. The experimental flows were generated when the gate was opened, allowing the saturated mixture of sediment to flow downhill. More details about how gate release experiments are conducted at the facility can be found in Iverson et al. (2010) and Iverson and George (2019). Videos of all past experiments, including the two we focus on in this study, are available online (Logan et al., 2018), and associated experimental data are also available online (Obryk and Iverson, 2019). For the May 24, 2017 (8 m3 ) experiment, the bed of the flume was partially covered in non-cohesive sediment on the surface in order to simulate debris flow movement over a deformable substrate. By contrast, the experimental debris flow on May 25, 2017 (10 m3 ) was released over bare, bumpy concrete tiles with a bump height of 1.6 cm (Iverson et al., 2010). Lidar data were obtained during the two flume experiments (Rengers and Olsen, 2020) using a Riegl VZ400 terrestrial lidar scanner (Figure 1). A custom C++ program, Drive VZ-400 (Olsen et al., 2012) was modified to operate the scanner and to fix a longitudinal profile with a vertical field of view of 44° in order to measure the entire flume and runout path. Horizontal angles along that profile were limited to 0.13° in order to focus on a narrow width of the flume. The scanner was manually leveled, and we improved the leveling quality of the data to ± 0.008° by applying corrections from the scanner inclination sensors. Using these settings, the lidar scanner was able to resolve the flume from top to bottom every 0.017 seconds. We began surveying the flume prior to the gate release. Therefore, the first lidar swaths of the flume were measuring the flume surface prior to initiation of the flow. A general 3D survey of the flume with multiple scan positions was conducted before the first flume experiment (on a clean flume without any sediment) and was georeferenced into Universal Transverse Mercator (UTM) coordinates using survey control targets positioned along the flume (Figure 1). From the raw scan

data stream, we extracted the time stamp with X-, Y-, and Z-coordinates of each lidar observation at a desired time range, time interval, and vertical angle resolution for analysis. Finally, the lidar points were organized into swaths that were uniquely numbered, with each swath defined as one full lidar capture of the entire flume profile. In our analyses, we established a local coordinate system to simplify the analyses. The X-coordinate of the flow position was defined as the horizontal distance from the gate (meters) as shown in Figure 2a. We note that for calculations of flow depth and velocity we omit portions of the point cloud where the gate is opening (a horizontal distance of 0.7 m) because the lidar unit records the gate opening, making it difficult to differentiate the flow from the gate geometry. The Z-coordinate, oriented perpendicular to the X-coordinate, represents the height (m) above the lowest point of the flume (Figure 2a). The Y-coordinate was oriented at right angles to both X and Z. To determine positions along the flume bed and bed-normal flow depths, we used the rotated X’-Z’ coordinate system shown in Figure 2a. We also note that a slope break occurs at the bottom of the flume (X = 67.7 m or X’ = 74). At that point the slope decreases as a catenary curve over a 2.2-m vertical drop, transitioning from a slope of 31° to 4° (Iverson et al., 2010). The lidar unit used on the flume collected topographic data at a variety of ranges (distances from the unit) and incidence angles. In order to estimate the accuracy of detecting change with the lidar unit, we examined how repeat lidar swaths of the flume bed changed before any flow moved down the flume. Ideally, during this period we would observe no change in the flume bed prior to the arrival of flow. However, differences will be observed due to ranging and angular encoder errors of the scanner. Note that our accuracy evaluation is focused on the accuracy of detecting change rather than accuracy associated with obtaining the absolute position or geometry. As a result, given a consistent setup for the test, systematic errors affecting positioning or geometry capture would be consistent during repeat profiles and thus would not affect change results comparing differences between profiles. Moreover, we examined the accuracy as a function of distance by comparing the lidar at three different locations on the flume. In addition, we examined the precision of the lidar data by comparing the lidar at the three flume locations with three fixed lasers mounted on the flume. Tracking Debris-Flow Velocity and Flow Behavior with Lidar We used the lidar profiles to track the debrisflow surface geometry and flow velocity as it moved

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 113–126

115

Rengers, Rapstine, Olsen, Allstadt, Iverson, Leshchinsky, Obryk, and Smith

Figure 1. (a) Photograph showing the position of the lidar scanner at the base of the ﬂume. (b) Oblique view of lidar data showing the ﬂume. Colors denote the elevation in meters. The lidar unit was located in the light-colored half circle in the foreground of (b).

down the flume. The lidar unit had sufficient precision (±3 mm) to detect both the main body of the flow and bouncing gravel above and in front of the main flow body. We used this unfiltered data to explore the formation of the saltating front, when larger rocks move to the front of the flow due to grain-size segregation and eventually form a cloud of bouncing rocks ahead of the main flow front. This saltating front has been observed previously in videos (Iverson et al., 2010). However, lidar offers a new approach for measuring the evolution of the saltating front from the gate release. Flow velocity was determined by tracking the change in flow-front distance between lidar timesteps. We tracked changes in relative height of each lidar swath compared to the initial flume elevation using: H = zi − zbed ,

(1)

where i is the line swath iteration that represents all of the points obtained during a 0.017-second lidar scan, H is the flow depth (H = 0 at the elevation of the flume surface before the gate is opened), zbed is the elevation of the flume bed surface before the gate is opened, and zi is the vertical elevation of the flow surface (note this is not a bed-normal depth). The flume surface included sediment for the May 24 (8 m3 ) experiment; therefore, H is calculated relative to the surface from a lidar swath prior to the gate opening rather than from a bare bed. Because lidar points are captured at even angular increments, they are not evenly spaced at

116

fixed intervals on the object, particularly when the surface is oblique to the scanner. Hence, we interpolate zi and zbed at evenly spaced points every 5 cm along the horizontal flume axis before using Eq. 1 to calculate H. In order to focus our analysis on the main flow body and not saltating rocks, we filtered the H values from each lidar swath using a spatial median filter. This approach effectively reduced the points representing the rocks that were detached from the main flow body and allowed us to better image the flow front. After applying the median filter, we defined the position of the flow front Fx numerically. Fx was identified as the maximum distance along the flume bed using the X’ downslope coordinate direction (Figure 2a), where the upslope flow depth minus the downslope flow depth is larger than 0.03 m: Fx = max(H j − H j+1 > 0.03),

(2)

where j is an index of the position along the flume bed surface. Here we note that because flow is moving down an inclined plane, we use the distance along the slope for j (as opposed to the horizontal distance). We used a threshold flow depth of 0.03 m instead of 0 m (1) in order to exclude any rolling particles preceding the flow front and (2) because the flume bed is covered in a bumpy substrate (Iverson et al., 2010). Finally, we calculated the flume bed-parallel flow-front velocity as the change in the position of the flow front along the

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 113–126

Debris-Flow Movement from Lidar

Figure 2. (a) Lidar swaths over a 35-second interval of the May 24 (8 m3 ) experiment. Each individual color represents a single lidar swath, which are numbered from 1 to 2,050. Cool colors represent the position of the flow at earlier times, and warm colors show later times (note the swaths are subsampled by 10). (b) Flow depth (i.e., the flow elevation subtracted from the flume surface elevation) with respect to distance along the flume. The saltating front first appears at X = 18 m downstream of the flume gate, indicated by dots showing bouncing particles. (c) Video still image shows a sharp front before saltating front formation and splashing in the flow body upstream from the flow front. (d) Video still image shows saltating front formation.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 113–126

117

Rengers, Rapstine, Olsen, Allstadt, Iverson, Leshchinsky, Obryk, and Smith Table 1. Mean and standard deviation of measurements of the fixed laser and lidar unit at three locations prior to any flow on the flume surface to estimate the repeatability. Top 31.7 m from Gate

Fixed laser Lidar

Middle 65.4 m from Gate

Bottom 80 m from Gate

Mean (mm)

SD (mm)

Mean (mm)

SD (mm)

Mean (mm)

SD (mm)

2.80E-02 –2.00E+00

9.40E-02 1.70E+00

–7.00E-02 –2.70E+00

1.10E-01 1.90E+00

–9.80E-02 2.20E-01

2.40E-01 8.50E-01

SD = standard deviation.

slope distance over the time between lidar swaths: V =

Fx i+1 − Fx i , (ti+1 ) − ti

tained with the ﬁxed lasers, but we do expect a correspondence in the timing of ﬂow passage. (3)

where ti is the maximum time recorded for each lidar swath i. Because the lidar unit takes approximately 0.017 seconds to acquire the full debris-flow profile geometry, the time associated with each point in a single lidar swath differs. Thus, we use the maximum time recorded for the swath to calculate Eq. 3, acknowledging that there could be an uncertainty up to 0.017 seconds in the velocity estimate. Comparing Lidar Timing with Fixed Instrumentation Data We examined the flow position in the lidar data as a function of time and depth using an approach similar to a seismic gather; herein we use the term “flow gather” to describe this visual analysis technique. The flow gather is a 3D matrix of lidar flow depth, distance along the flume, and time, where the rows represent the average time of a lidar swath, the columns represent the distance along the flume, and a color displays the lidar-measured flow depth at all flume distances for the lidar swath time. Using this approach, we can visualize the entire flow in a single image. Moreover, we can compare the timing of the lidar-observed flow front to direct sensor measurements of the flow front at three fixed points along the flume. The instrumented points have force plates to measure stress (normal and shear), as well as a laser probe that measures the flow depth normal to the bed surface. We also compared the lidar-measured flow depths to time-series measurements obtained from three fixed lasers positioned at slope distances of 31.7 m, 65.4 m, and 80 m measured from the gate. We note that the lidar scanning path along the flume is offset laterally from the position of the fixed-mounted lasers, and thus the measurements of flow depth gauge different depths in the cross-flume direction (e.g., perpendicular to flow). As a result, due to differences in crossflume flow depth, we do not expect an exact match between lidar flow-depth measurements and those ob-

118

RESULTS AND DISCUSSION Lidar Accuracy and Precision Our lidar unit proved to provide consistent and repeatable results during the experiment. As an estimate of measurement accuracy, the mean difference between the profiles measured in the flume bed before the arrival of flow was –2.0 mm, –2.7 mm, and 2.2 mm at distances of 31.7 m, 65.4 m, and 80.0 m from the gate, respectively (Table 1). The standard deviation of those measurements ranged from 0.85 mm to 1.9 mm (Table 1). In addition, the results do not appear to strongly differ as a function of range. Although the lidar is less accurate and precise than the three fixed lasers, it still provided data of sufficient quality (millimeter level) for the needed flow observations given that the changes during the flows were centimeters to several decimeters in magnitude. Lidar-Observed Debris-Flow Behavior The unfiltered lidar swath data recorded the surface of the moving debris flow with high temporal resolution and provided a detailed view of the flow behavior. Video recordings of the May 24, 2017 (8 m3 ) experiment show watery splashing of the flow as it moved downslope from the headgate and proceeded down the flume (Figure 2). As the flow approached X = 17 m, larger particles started to appear in the data at the front of the flow, showing the formation of a saltating flow front (Figure 2). The debris in this experiment stopped shortly after the slope break at the bottom of the flume (X = 67.7 m), developing a fan at the flume base with a steep front at X = 71.6 m (Figure 2). Similar behavior was observed during the experiment on May 25, 2017 (10 m3 ), where a saltating front appeared to begin around X = 17 m from the debris-flow release gate (Figure 3). At X = ∼18 m the flow front in the unfiltered lidar data was obscured by bouncing particles (Figure 3b). The larger mass involved in the

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 113–126

Debris-Flow Movement from Lidar

Figure 3. (a) Lidar swaths over a 25-second interval of the May 25 (10 m3 ) experiment. Each individual color represents a single lidar swath, which are numbered from 1 to 1,480. Cool colors represent the position of the flow at earlier times, and warm colors show later times (note the swaths are subsampled by 10). (b) Flow depth (i.e., the flow elevation subtracted from the flume surface elevation) with respect to distance along the flume. The saltating front becomes evident at X = ∼16 m downstream of the flume gate, indicated by dots showing bouncing particles. Note the much longer travel distance of this flow due to the higher initial volume of sediment released than the May 24 experiment (Figure 2b). (c) Video still image shows a sharp front before saltating front formation. (d) Video still image shows saltating front formation.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 113–126

119

Rengers, Rapstine, Olsen, Allstadt, Iverson, Leshchinsky, Obryk, and Smith

Figure 4. Data from the experiment on May 25, 2017 (10 m3 ). (a) Lidar difference between the debris height and the flume bed height (time = 4.6 seconds or lidar swath 300); (b) Lidar difference between the debris height and the flume bed height (time = 12.3 seconds or lidar swath 800).

experiment on May 25, 2017 (10 m3 ) resulted in a longer runout to X = 84.2 m with a thinner fan deposit, and the final depositional front was similarly steep (Figure 3). Select time steps show initial and final phases of the flow (for simplicity we use only the May 25 experiment to demonstrate the behavior) (Figure 4). Initially the debris flow had a steep front, and during the early stages of motion the front was not the deepest portion of the flow (Figure 4). Downstream, the larger particles moved to the front of the flow as the grains self-sorted according to size. At the end of the flume, the lidar swath showed leading large particles moving ahead of a steep front (Figure 4), which was followed by a tail that stretched behind it for tens of meters. Lidar-Observed Flow Velocity When the spatial median filter was applied to the lidar data, it removed most of the signal from larger bouncing particles downstream of the main flow front (Figure 5). Once the airborne particles were removed from the lidar data, Eq. 2 was used to locate the flow front at discrete time steps (Figure 6). This enabled us to subsequently apply Eq. 3 to estimate the velocity

120

of the flow front for the duration of the experiment. The down-slope velocities determined for the two experiments showed similar patterns (Figure 7). Both experiments showed an initial acceleration downstream of the headgate, up to a peak velocity. The May 24 (8 m3 ) experiment reached an initial peak near 19.5 m (horizontal position), and the May 25 (10 m3 ) experiment attained an initial peak velocity at 32 m (horizontal position). Both experiments showed periods of flow acceleration following the initial peak velocity, coincident with roll waves overtaking the main flow front (Figure 7). Roll waves are small surges with a gravel front that form spontaneously behind the main flow front (Iverson et al., 2010). The May 24 experiment slowed to 0.53 m/s at X = 68.3 m horizontal distance, near the rapid change in the flume slope, whereas the May 25 experimental flow was still moving at 1.5 m/s at 80 m from the headgate (see Figure 7 for velocities and Figure 2a for distances). Comparing Lidar Timing with Fixed Instrumentation Data The flow gather representation of the lidar data provides a concise method for viewing the lidar flow depth

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 113–126

Debris-Flow Movement from Lidar

Figure 5. (a) A single lidar swath showing the debris flow surface depth (H) as well as airborne particles in the flow that are fully detached from the main body of the flow. (b) The same lidar swath as in (a) showing the debris flow surface depth after applying a median filter, making the flow front easier to identify. Note: This process was applied to the experimental debris flows on both May 24 and 25, 2017, but for simplicity we only show the results for May 24, 2017.

as a function of time and distance along the flume (Figure 8). The flow gather enables us to evaluate agreement among the lidar-observed front arrival time, flow thickness, and basal normal and shear stresses records of the flow at fixed points (Figure 8). For example, the measured stress and thickness have a value of zero for times before the flow front is observed, and they abruptly increase at the time when the flow front crosses the sensor (Figure 8). A direct comparison of flow depths measured with the lidar versus the fixed laser probes further illustrates the correspondence in time between the lidar and instrumentation (Figure 9). The correlation coefficients between the lidar and the fixed lasers are 0.97, 0.94, and 0.96 at 31.7 m, 65.4 m, and 80 m downslope of the headgate, respectively. This parity in time between the measured flow front at three fixed sensors and the flow gather shows the utility of using lidar to image the full profile of the moving debris flow. In addition, the flow gather also reveals flow behavior; for example, the period of deep flow (yellow streaks, Figure 8) that follows the initial flow front (black line, Figure 8) reveals the secondary surges and roll waves that follow the main flow front (Figure 8).

Implications for Understanding Debris Flows In order to design appropriate mitigation measures to address debris-flow hazards, appropriate data must be collected to understand the debris-flow processes. In particular, we must understand the evolution of debrisflow velocities (Arattano and Marchi, 2000; Hungr, 2000; Kaitna et al., 2014), the development of the dense granular saltating front (a key factor in erosion) (McCoy et al., 2012, 2013), and the runout and depositional behavior of debris flows (Major and Iverson, 1999; Rickenmann, 2005). This study makes a step toward addressing those needs by demonstrating the capability of a terrestrial lidar scanner to rapidly measure debris-flow geometry and velocity. The velocity measurements obtained during this study show that a steady-state debris-flow velocity assumption (Hungr, 2000) does not align with our observed velocity. We show that once a peak debris-flow velocity is obtained, there is a decrease in velocity despite a constant slope, punctuated by surges due to roll waves. We also confirm the observations of Costa (1984) and Iverson et al. (2010) showing that roll waves influence the shape of the debris flow (Figure 7c) and its velocity

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 113–126

121

Rengers, Rapstine, Olsen, Allstadt, Iverson, Leshchinsky, Obryk, and Smith

Figure 6. Debris flow height (H) calculated using Eq. 1, as a function of distance down the flume, shown at five discrete time steps (a–e). Vertical red line indicates the position of the flow front determined using Eq. 2. Note: This process was applied to the experimental debris flows on both May 24 and 25, 2017, but for simplicity we only show the results for May 24, 2017.

122

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 113–126

Debris-Flow Movement from Lidar

Figure 7. Flow front velocity derived from the lidar for both the May 24, 2017 (8 m3 ) and May 25, 2017 (10 m3 ), experiments. Note a moving average was applied to the velocity using a moving window size of 10 points.

(yellow streaks in Figure 8 show roll waves, and their steeper slopes in time–distance space indicate higher velocities). The lidar observations are also useful in showing how debris-flow behavoir changes over time. We observe a distinct saltating front that is formed abruptly and maintained for most of the flow (Figures 2 and 3). Time stamps of the flow also show that the watery tail following the front enlongates with distance downstream (Figure 6) rather than reaching a steady length. Finally, the lidar shows the time sequence of deposition, revealing how later portions of the flow tend to override early deposits, extending the depositional fan (Figures 2 and 3). This style of measurement has not previously been documented for debris-flow studies and therefore has potential for future targeted investigations of specific flow phenomena. The experimental design used in this study will be most easily applied in controlled settings such as large-scale flumes, but as the technology for field-mounted lidar units improves, it may be possible to apply this approach to field-based monitoring in the future. Moreover, lidar data from experimental debris-flow observations could be used in the fu-

ture as a validation technique to compare with lowercost instruments such as video cameras, which would be more easily deployed in remote field investigations. Finally, the data obtained from lidar observations are of sufficient temporal and spatial quality to be informative for numerical model testing (George and Iverson, 2014; Iverson and George, 2014), which could further refine efforts to develop debris-flow mitigation approaches. CONCLUSIONS This study demonstrates a novel use of a terrestrial lidar unit to continuously monitor surface profiles of rapidly deforming debris flows. We used a terrestrial lidar scanner to track experimental debris flows during a flume experiment at a time interval of 0.017 seconds. The data acquired in this study enable us to quantify aspects of the flow through time, such as the velocity of the flow front, the position and time when coarse and fine grains begin to segregate, and the size and movement of roll waves. Moreover, the lidar data agree well with experimental data at fixed positions on the flume, further validating the accuracy of the technique. These

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 113–126

123

Rengers, Rapstine, Olsen, Allstadt, Iverson, Leshchinsky, Obryk, and Smith

Figure 8. This plot is a 3D matrix of lidar flow depth, distance along the flume, and time, herein referred to as a flow gather. Flow depth is measured from lidar swaths for the May 25, 2017 (10 m3 ) experiment plotted as a function of time and distance along the flume axis. The flow depth displayed is equal to the flow height above the flume bed, and for clarity the color scale is saturated at 0.3 m (median flow depth is 0.013 m with a standard deviation of 0.265 m). Black arrows indicate roll waves. The black line with white trim approximately indicates the flow front. Basal normal stress, basal shear stress, and laser-based flow thickness data are overlaid at the fixed cross sections where they were measured. Note, for display purposes, stress curves increase to the right and the thickness curves increase to the left. These measurements follow the methods described in Iverson et al. (2010).

Figure 9. Comparison of data from the static mounted lasers at the flume versus the lidar data at the same points. (a) Depth normal to the flume surface at 31.7-m distance from the gate along the flume surface. (b) Depth normal to the flume surface at 65.4-m distance from the gate along the flume surface. (c) Depth normal to the flume surface at 80.0-m distance from the gate along the flume surface. Note the fixed laser and lidar are not measuring flow at exactly the same position, so variations in depth measurements are expected.

124

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 113–126

Debris-Flow Movement from Lidar

findings highlight an overarching approach for using the high temporal and spatial resolution capabilities of terrestrial lidar instruments to track debris-flow movement. This approach shows promise for future use in experimental studies and provides unique data that can be used to enhance our validation of debris-flow modeling. ACKNOWLEDGMENTS All lidar data used in this study are available via Sciencebase (Rengers and Olsen, 2020). Videos of the experiments are available in Logan et al. (2018), and further data from this experiment are available in Obryk and Iverson (2019). The authors particularly acknowledge the help of Jason Kean, who helped to survey key points of infrastructure around the flume. Leica Geosystems and Maptek I-Site provided software that was used in this study. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government. REFERENCES Anderson, S. W.; Anderson, S. P.; and Anderson, R. S., 2015, Exhumation by debris flows in the 2013 Colorado Front Range storm: Geology, Vol. 5, No. 43, pp. 391–394. doi:10.1130/G36507.1. Arattano, M. and Marchi, L., 2000, Video-derived velocity distribution along a debris flow surge: Physics Chemistry Earth, Part B: Hydrology, Oceans Atmosphere, Vol. 25, No. 9, pp. 781–784. doi:10.1016/S1464-1909(00)00101-5. Brodie, K. L.; Raubenheimer, B.; Elgar, S.; Slocum, R. K.; and McNinch, J. E., 2015, Lidar and pressure measurements of inner-surfzone waves and setup: Journal Atmospheric Oceanic Technology, Vol. 32, No. 10, pp.1945–1959. doi:10.1175/JTECH-D-14-00222.1. Costa, J. E., 1984, Physical geomorphology of debris flows. In Costa, J. E. and Fleisher, P. J. (Editors), Developments and Applications of Geomorphology: Springer Praxis Books, Berlin, Germany, pp. 268–317. DeLong, S. B.; Youberg, A. M.; DeLong, W. M.; and Murphy, B. P., 2018, Post-wildfire landscape change and erosional processes from repeat terrestrial lidar in a steep headwater catchment, Chiricahua Mountains, Arizona, USA: Geomorphology, Vol. 300, pp. 13–30. doi:10.1016/j.geomorph.2017.09.028. George, D. L. and Iverson, R. M., 2014, A depth-averaged debris-flow model that includes the effects of evolving dilatancy. II. Numerical predictions and experimental tests: Proceedings Royal Society A, Vol. 2170, No. 470, pp. 20130820. doi:10.1098/rspa.2013.0820. Hungr, O., 2000, Analysis of debris flow surges using the theory of uniformly progressive flow: Earth Surface Processes Landforms, Vol. 25, No. 5, pp. 483–495. doi: 10.1002/(SICI)10969837(200005)25:5<483::AID-ESP76>3.0.CO;2-Z. Iverson, R. M., and George, D. L., 2014, A depth-averaged debris-flow model that includes the effects of evolving dilatancy. I: Physical basis: Proceedings Royal Society A, Vol. 2170, No. 470, p. 20130819. doi:10.1098/rspa.2013.0819. Iverson, R. M. and George, D. L., 2019, Valid debris-flow models must avoid hot starts, 2019. In Kean, J. W.; Coe, J. A.;

Santi, P. M.; and Guillen, B. K. (Editors), Debris-Flow Hazards Mitigation: Mechanics, Monitoring, Modeling, and Assessment: Association of Environmental and Engineering Geologists, Golden, Colorado, USA, pp. 25–32. Iverson, R. M.; George, D. L.; and Logan, M., 2016, Debris flow runup on vertical barriers and adverse slopes: Journal Geophysical Research: Earth Surface, Vol. 121, No. 12, pp. 2333– 2357. doi:10.1002/2016JF003933. Iverson, R. M.; Logan, M.; LaHusen, R. G.; and Berti, M., 2010, The perfect debris flow? Aggregated results from 28 largescale experiments: Journal Geophysical Research: Earth Surface, Vol. F3, No. 115, doi:10.1029/2009JF001514. Jacquemart, M.; Meier, L.; Graf, C.; and Morsdorf, F., 2017, 3D dynamics of debris flows quantified at sub-second intervals from laser profiles: Natural Hazards, Vol. 89, No. 2, pp. 785– 800. doi:10.1007/s11069-017-2993-1. Kaitna, R.; Dietrich, W.; and Hsu, L., 2014, Surface slopes, velocity profiles and fluid pressure in coarse-grained debris flows saturated with water and mud: Journal Fluid Mechanics, Vol. 741, pp. 377–403. doi:10.1017/jfm.2013.675. Kramer, M.; Chanson, H.; and Felder, S., 2020, Can we improve the non-intrusive characterization of high-velocity air– water flows? Application of LIDAR technology to stepped spillways: Journal Hydraulic Research, Vol. 58, No. 2, pp. 350–362. doi:10.1080/00221686.2019.1581670. Logan, M.; Iverson, R. M.; and Obryk, M. K., 2018, Video Documentation of Experiments at the USGS Debris-Flow Flume 1992–2017 (ver 1.4, January 2018): U.S. Geological Survey Open-File Report 2007–1315. doi:10.3133/ofr20071315. Longoni, L.; Papini, M.; Brambilla, D.; Barazzetti, L.; Roncoroni, F.; Scaioni, M.; and Ivanov, V. I., 2016, Monitoring riverbank erosion in mountain catchments using terrestrial laser scanning: Remote Sensing, Vol. 8, No. 3, pp. 241. doi:10.3390/rs8030241. Major, J. J. and Iverson, R. M., 1999, Debris-flow deposition: Effects of pore-fluid pressure and friction concentrated at flow margins: Geological Society America Bulletin, Vol. 111, No. 10, pp. 1424–1434. doi:10.1130/ 0016-7606(1999)111<1424:DFDEOP>2.3.CO;2. Martins, K.; Blenkinsopp, C. E.; Power, H. E.; Bruder, B., Puleo, J. A.; and Bergsma, E. W., 2017, High-resolution monitoring of wave transformation in the surf zone using a LiDAR scanner array: Coastal Engineering, Vol. 128, pp. 37–43. doi:10.1016/j.coastaleng.2017.07.007. McCoy, S. W.; Kean, J. W.; Coe, J. A.; Staley, D. M.; Wasklewicz, T. A.; and Tucker, G. E., 2010, Evolution of a natural debris flow: In situ measurements of flow dynamics, video imagery, and terrestrial laser scanning: Geology, Vol. 38, No. 8, pp. 735–738. doi:10.1130/G30928.1. McCoy, S. W.; Kean, J.; Coe, J.; Tucker, G.; Staley, D.; and Wasklewicz, T., 2012, Sediment entrainment by debris flows: In situ measurements from the headwaters of a steep catchment: Journal Geophysical Research: Earth Surface, Vol. 117, No. F3. doi:10.1029/2011JF002278. McCoy, S. W.; Tucker, G.; Kean, J.; and Coe, J., 2013, Field measurement of basal forces generated by erosive debris flows: Journal Geophysical Research: Earth Surface, Vol. 118, No. 2, pp. 589–602. doi:10.1002/jgrf.20041. Neugirg, F.; Stark, M.; Kaiser, A.; Vlacilova, M.; Della Seta, M.; Vergari, F.; Schmidt, J.; Becht, M.; and Haas, F., 2016, Erosion processes in calanchi in the Upper Orcia Valley, Southern Tuscany, Italy based on multitemporal high-resolution terrestrial LiDAR and UAV surveys: Geomorphology, Vol. 269, No. 15, pp. 8–22. doi:10.1016/j.geomorph.2016.06.027.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 113–126

125

Rengers, Rapstine, Olsen, Allstadt, Iverson, Leshchinsky, Obryk, and Smith Obryk, M. K., and Iverson, R. M., 2019, Sensor Data from Natural-Release Experiments Conducted in May, 2017, at the USGS Debris-Flow Flume, HJ Andrews Experimental Forest, Blue River, Oregon: U.S. Geological Survey Data Release, doi:10.5066/P9QSZ4XO. Olsen, M. J.; Butcher, S.; and Silvia, E. P., 2012, RealTime Change and Damage Detection of Landslides and Other Earth Movements Threatening Public Infrastructure: OTRECRR-11-23: Transportation Research and Education Center (TREC), Portland, OR, 80 p. doi:0.15760/trec.47. O’Neal, M. A. and Pizzuto, J. E., 2011, The rates and spatial patterns of annual riverbank erosion revealed through terrestrial laser-scanner surveys of the South River, Virginia: Earth Surface Processes Landforms, Vol. 36, No. 5, p. 695. doi:10.1002/esp.2098. Prochaska, A. B.; Santi, P. M.; and Higgins, J. D., 2008, Debris basin and deflection berm design for fire-related debrisflow mitigation: Environmental Engineering Geoscience, Vol. 14, No. 4, pp. 297–313. doi:10.2113/gseegeosci.14.4.297. Rabatel, A.; Deline, P.; Jaillet, S.; and Ravanel, L., 2008, Rock falls in high-alpine rock walls quantified by terrestrial lidar measurements: A case study in the Mont Blanc area: Geophysical Research Letters, Vol. 35, No. 10, doi:10.1029/2008GL033424. Rengers, F. and Olsen, M., 2020, Lidar Data for Gate Release Experiment at the USGS Debris Flow Flume 24 and 25 May 2017: U.S. Geological Survey Data Release, doi:10.5066/P9OU3U4P. Rengers, F. and Tucker, G., 2015, The evolution of gully headcut morphology: A case study using terrestrial laser scanning and hydrological monitoring: Earth Surface Processes Landforms. Vol. 40, No. 10, pp. 1304–1317. doi:10.1002/esp.3721. Rengers, F.; Tucker, G. E.; Moody, J. A.; and Ebel, B. A., 2016, Illuminating wildfire erosion and deposition patterns

126

with repeat terrestrial lidar: Journal Geophysical Research: Earth Surface, Vol. 121, No. 3, pp. 588–608. doi:10.1002/ 2015JF003600. Rickenmann, D., 2005, Runout prediction methods. In Matthias, J. and Hungr, O. (Editors), Debris-Flow Hazards and Related Phenomena: Springer Praxis Books, Berlin, Germany, pp. 305– 324. doi:10.1007/3-540-27129-5_13. Staley, D. M.; Wasklewicz, T. A.; and Kean, J. W., 2014, Characterizing the primary material sources and dominant erosional processes for post-fire debris-flow initiation in a headwater basin using multi-temporal terrestrial laser scanning data: Geomorphology, Vol. 214, pp. 324–338. doi:10.1016/j.geomorph.2014.02.015. Stiny, J., 1910, Die Muren: Verlag der Wagner’schen Universitätsbuchhandlung, Innsbruck, Austria. 139 p. Takahashi, Y.; Fujimura, N.; Akita, H.; and Mizuno, M., 2019, Dynamic characteristics of extreme superelevation of debris flows observed by laser profile scanners in Sakura-jima volcano, Japan. In Kean, J. W.; Coe, J. A.; Santi, P. M.; and Guillen, B. K. (Editors), Debris-Flow Hazards Mitigation: Mechanics, Monitoring, Modeling, and Assessment: Special Publication 28, Association of Environmental and Engineering Geologists, Golden, Colorado, USA. doi: 10.25676/11124/ 173217. Williams, J. G.; Rosser, N. J.; Hardy, R. J.; and Brain, M. J., 2019, The importance of monitoring interval for rockfall magnitude-frequency estimation: Journal Geophysical Research: Earth Surface, Vol. 124, No. 12, pp. 2841–2853. doi:10.1029/2019JF005225. Williams, J. G.; Rosser, N. J.; Hardy, R. J.; Brain, M. J.; and Afana, A. A., 2018, Optimising 4-D surface change detection: An approach for capturing rockfall magnitude frequency: Earth Surface Dynamics, Vol. 6, No. 1. pp. 101–119. doi:10.5194/esurf-6-101-2018.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 113–126

Experimental Investigation on the Impact Dynamics of Saturated Granular Flows on Rigid Barriers NICOLETTA SANVITALE* ELISABETH BOWMAN Dept. Civil and Structural Engineering University of Sheffield, Sheffield Sir Frederick Mappin Building Mappin St, Sheffield S1 3JD United Kingdom

MIGUEL ANGEL CABRERA Departamento de Ingeniería Civil y Ambiental Universidad de los Andes, Bogotá Carrera 1 Este No. 19a –40 111711 Colombia

Key Terms: Debris Flows, Engineering Geology, Geotechnical, Physical Modelling, Landslides ABSTRACT Debris flows involve the high-speed downslope motion of rocks, soil, and water. Their high flow velocity and high potential for impact loading make them one of the most hazardous types of gravitational mass flows. This study focused on the roles of particle size grading and degree of fluid saturation on impact behavior of fluidsaturated granular flows on a model rigid barrier in a small-scale flume. The use of a transparent debris-flow model and plane laser-induced fluorescence allowed the motion of particles and fluid within the medium to be examined and tracked using image processing. In this study, experiments were conducted on flows consisting of two uniform and one well-graded particle size gradings at three different fluid contents. The evolution of the velocity profiles, impact load, bed normal pressure, and fluid pore pressure for the different flows were measured and analyzed in order to gain a quantitative comparison of their behavior before, during, and after impact. INTRODUCTION A debris flow is a rapid surging mass of non-plastic soil, rock, and water in a steep channel that may present high impact load and long runout (Iverson, 1997; Takahashi, 2007; and Hungr et al., 2013). A common approach to prevent these flows from reaching vulnerable areas is by obstructing their channelized paths with engineered barriers, which trap most of the transported debris, dampening the overall flow inertia and, therefore, decreasing their expected runout. These barriers can be rigid walls or flexible nets, with their main goal being to withstand the impact forces from the transported debris and suspended (fluidized) ma*Corresponding author email: nicoletta.sanvitale@gmail.com

terial. Rigid barriers, also called check dams or sometimes catching dams, are the most common mitigation structure against debris flows, due to the minimal technical skills required in their construction and relative ease of obtaining building materials for reinforced concrete (Hübl et al., 2009). The mechanics of debris flows depend on the interactions between the solid and fluid phases, which involve frictional, collisional, and viscous stress transfer between particles and fluid, as well as flow-bed interactions for both particles and fluid. While the estimation of the pressures generated by the impact of debris flows on civil engineering structures has been widely investigated (Moriguchi et al., 2009; Armanini et al., 2011; Bugnion et al., 2011; Hu et al., 2011; Scheidl et al., 2012; Cui et al., 2015; and Zhou et al., 2018), the state of knowledge is still insufficient to accurately understand the effect of solid-fluid interactions on the dynamics and load evolution of the impact process. As a result, design approaches tend to be semi-empirical (Zhang, 1993; Van Dine, 1996; Armanini, 1997; Arattano and Franzi, 2003). The current paper presents the results of experiments using transparent analogue debris flows in a small-scale flume, aimed at investigating the bulk impact forces on rigid barriers. Granular flows with different particle size distributions and fluid contents were adopted for the tests. The dynamics of the impact against the rigid barrier normal to the flow direction were observed via planar laser-induced fluorescence, PLIF (Stöhr et al., 2003; Sanvitale and Bowman, 2012). Impact forces against the obstacle, the basal total and fluid pressures, the flow height, and the midcross-sectional flow dynamics at impact were recorded and are discussed here. EXPERIMENTAL SETUP Before testing, the fluid-saturated granular material was stored in a rectangular sealed tank at the

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 127–138

127

Sanvitale, Bowman, and Cabrera

Figure 1. Apparatus employed in the tests. (Inset) PLIF setup.

top of the channel. The material was gently agitated by hand within the tank to ensure consolidation was avoided, prior to manually releasing a sluice gate. The material flowed down the 2.57-m-long and 150-mmwide rectangular flume, which had an adjustable angle of inclination. The angle was set at 20° for the tests described here (see Figure 1). The barrier model was made of a 10-mm-thick, 145-mm-wide, and 190mm-tall polymethyl methacrylate (i.e., acrylic) plate mounted perpendicularly to the base (see inset in Figure 1). This enabled the barrier to effectively cross the full width of the flume but be unaffected by wall interactions. The plate was centrally connected via an aluminum support at the base to an axial load cell (U9C, HBM, Germany) and fixed to a linear bearing (LZMHS12-37T2P1, SKF, Sweden). The barrier model was fixed to the flume bed at 2.25 m from the gate release. The side walls of the channel were made of borosilicate glass, and the bottom of the flume was roughened with 3D-printed polylactic acid plates with a hexagonal packing of 3 mm semi-spheres. The roughened bed was instrumented along its base with three pore-pressure transducers, denoted PPT2, 3,

128

and 4 (PDCR 810, Druck, UK), and a load cell (LUXBID, Kyowa, Japan) with a top circular sensing plate of 23 mm diameter. PPT2 was located 350 mm upslope from the end of the flume. Pore-pressure sensor PPT2 and the load cell were located closest to the barrier at 75 mm distance and 30 mm to either side of the centerline. The transducers PPT3 and PPT4 were located 175 mm and 350 mm further upslope from PPT2, respectively. All basal sensors had 3D-printed disk headings, equivalent to the roughness of the rest of the base, with the top of the heading flush with the base. A 0.5-mm-thick, 532 nm laser light sheet was allowed to pass through a slit cut in the roughened bed and barrier model base, illuminating the flowing material along the flume centerline. The laser used was a Laser Quantum Opus (UK) 532 producing continuous illumination at a power of approximately 1.5 W. The laser beam was positioned perpendicular to the bottom of the flume via a mirror and then sent through three uncoated plano-convex cylindrical lenses (purchased from www.thorlabs.com) that spread the beam into a light sheet. The length of the illuminated flume

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 127–138

Impact Dynamics of Saturated Granular Flows

section was approximately 130–150 mm. A high-speed camera (Miro M310, Phantom, USA) located close to the end of the flume recorded video of the illuminated cross section at 2000 frames per second with a resolution of 1280 × 800 pixels. A long-pass filter was placed over the lens to transmit only the fluorescence signal and discard the reflected laser light (Sanvitale and Bowman, 2012). Materials The PLIF technique relies on the use of a laser sheet to excite the fluorescence of a dye diluted in the fluid and hence create an illuminated plane within the flow in which particles appear as dark shapes against a bright background. For the PLIF technique to work under optimum conditions, the refractive indices of the fluid and solid should match. The current experiments were performed with hydrocarbon oil (Cargille Laboratories) dyed with a fluorescent powder, Nile Red, and mixed with borosilicate glass beads (Sigmund Lindner GmbH). The fluid had a kinematic viscosity 16 times higher than that of water (16 cSt at 25°C) and a density 1.182 times lower than that of water (0.846 g/cm3 ), such that mixture consolidation behavior was equivalent to that using quartz particles that were one quarter the diameter in water. See Sanvitale and Bowman (2012) for further details on the experimental technique. Three granular materials consisting of spherical borosilicate glass beads were used in these experiments: two uniform particle size distributions, PSD1 and PSD2, with glass beads of 3 mm and 7.5 mm, respectively, and a more well-graded sample, PSD3 (coefficient of uniformity CU = d60 /d10 = 5, where dx denotes the percentage passing by mass), with mean particle size of 7.5 mm (Figure 2). These samples were intended to provide an insight into the effects of particle size and gradation on the impact dynamics. The influence on the flow dynamics and impact of the fluid content were also investigated by setting the initial fluid content fc , defined as massfluid /masssolid to 24 percent, 28 percent, and 32 percent for each

Figure 2. Particle size distributions (PSDs) for the solid materials used in the tests.

solid material investigated (equivalent to solid volume fractions of 0.61, 0.58, and 0.54, respectively). Hence, we report the results of nine experiments in all. Test Procedure Prior to each experiment, the flume was cleaned, avoiding the presence of dirt and oil films on the roughened bed and side walls. For each experiment, an exactly 10 kg aliquot of solid mass was used (Table 1). Oil to the desired fluid content was poured into the container and gently mixed with the glass beads to reduce the entrainment of air bubbles that would otherwise reduce optical transparency. Agitation of the mixture was maintained while the laser beam was turned on, the high-speed camera was activated, and the sluice gate was opened. At release, a triggering shutter connected to the gate activated the sensors that began recording at a sampling rate of 36 kHz, for a duration of 9 seconds. Low-pass filters were applied to the outcomes as described in the “Barrier Load” section later herein. RESULTS AND DISCUSSION Impact Kinematics Figure 3 shows images of the flow impact for each particle size distribution at different fluid content,

Table 1. Test program and testing materials. Particle Size Distribution PSD1 PSD1 PSD1 PSD2 PSD2 PSD2 PSD3 PSD3 PSD3

Fluid Content (%)

Solid Mass (kg)

d50 (mm)

Slope (°)

Low-Pass Filter (Hz)

24 28 32 24 28 32 24 28 32

10 10 10 10 10 10 10 10 10

3 3 3 7.5 7.5 7.5 7.5 7.5 7.5

1 1 1 1 1 1 5 5 5

20 20 20 20 20 20 20 20 20

270 450 630 350 370 380 70 90 105

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 127–138

129

Sanvitale, Bowman, and Cabrera

Figure 3. Sequences of the images recorded by the high-speed camera for (a) PSD1 at fc of 24 percent, (b) PSD2 at fc of 24 percent, (c) PSD3 at fc of 32 percent. The overlapped arrows describe the corresponding velocity ﬁeld estimated using PIV analysis.

130

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 127–138

Impact Dynamics of Saturated Granular Flows

fc : 24 percent for PSD1 and PSD2, and 32 percent for PSD3 (images for the other tests are not shown for brevity). The flow direction is from right to left. The images show different instants during the impact of the mixture against the barrier with respect to time t = 0, i.e., the time at which the sluice gate was opened. For all the tests, the impact process was characterized by a first stage during which individual saltating particles impacted the wall before the arrival of the flow front. For the test using PSD3, due to the segregation of the different particle sizes during downslope shearing (Sanvitale and Bowman, 2012, 2017), larger particles accumulated at the front. The images show the three different impact mechanisms (Gray et al., 2003; Armanini et al., 2011; Choi et al., 2015; Faug et al., 2015; and Albaba et al, 2018) observed during experiments. The first mechanism (Figure 3a) was displayed only by PSD1 (uniform grain size of 3 mm beads) at fc = 24 percent and consisted of a type of pile-up process. The surge front impacted the rigid barrier and deposited material at the base, and then the subsequent flow material impacted and piled up on top of the existing deposits. When the maximum pileup height was reached, the impact process rapidly attenuated. The second mechanism (Figure 3b) occurred after the impact by the formation of a reflected wave propagating upstream. The third mechanism (Figure 3c) consisted of the formation of a vertical jet travelling parallel to the vertical barrier that subsequently fell backward on the incoming surge, creating a secondary surge that propagated upstream. For PSD2 and PSD3 at fc = 24 percent, we observed the formation of a reflected wave, while at fc = 28 percent and 32 percent, we observed the jet-like behavior. The test for PSD1 at fc = 28 percent showed an intermediate behavior, with formation of an initially small jet that, once it fell back on itself, created a small surge propagating upstream that was immediately stopped by the incoming flow. Subsequently, the impact of the incoming flow transitioned more to be a type of reflected wave. Figure 3 also shows the results of the corresponding particle image velocimetry (PIV) analyses (Thielicke, 2014; Thielicke and Stamhius, 2014) conducted via image processing at specific stages during flow impact. The velocity field shown via the quiver plots (quiver lengths proportional to speed) displays the kinematics that are characteristic of the observed mechanisms. The velocity, vi , which is the average speed of the front of the incoming surge approaching the barrier before impact, is listed in Table 2. The front velocity was generally higher at larger fluid content. For the uniform gradings, at the same fluid content, the velocities were higher for the larger particle sizes, i.e., larger for PSD2 than PSD1. For the well-graded material,

Table 2. Experimental results. Particle Size Distribution PSD1 PSD1 PSD1 PSD2 PSD2 PSD2 PSD3 PSD3 PSD3

Fluid Content (%) 24 28 32 24 28 32 24 28 32

vi hi d50 (mm) (m/s) (mm) 3 3 3 7.5 7.5 7.5 7.5 7.5 7.5

0.54 1.12 1.68 2.21 2.34 2.70 1.07 1.46 1.76

40 26 16 38 39 34 26 22 21

Fr 0.90 2.29 4.38 3.75 3.91 4.86 2.19 3.28 4.05

PSD3, with the same mean particle size as PSD2, the front velocity was much lower, but it was higher than for PSD1 tests. The bed roughness may have played a role in the observed flow kinematics (Silbert et al., 2001; Goujon et al., 2003; Ahmadipur et al., 2019), with the scale of roughness of the 3 mm semi-spheres on the base inducing greater shearing within the uniform mixture of 3 mm beads in PSD1 compared to the 7.5 mm beads in PSD2. Tests with mixture PSD3 had 50 percent (by mass) particles larger than 7.5 mm and 50 percent smaller, producing lower flow velocities than for PSD2. In these tests, the presence of finer particles may have had a dampening effect on large particle collisions, and hence dissipated more energy within the body. The Froude number, vi , (1) Fr = √ gcosθhi where hi is the flow depth of the incident surge, and θ is the slope of the channel, quantifies the ratio between the inertial and gravitational forces. Table 2 reports the Froude numbers calculated for the experiments. The Fr results lie in the range between 0.90 and 4.86, with values lying closer together for the larger particle sizes. In all cases, Fr increased with fluid content, due both to increased velocity and reduced flow height. Armanini et al. (2011, 2019) found that the nature of the impact depends on the Froude number of the incoming front; when gravity dominates over inertia, there is a formation of a reflected wave, whereas when inertia dominates, jet-like behavior up the wall occurs upon impact. Specifically, Armanini et al. (2019) analyzed the dynamic impact of water and mixtures of water and sediments having a specific particle size distribution (d30 = 2.0 mm, d50 = 3.5 mm, d90 = 9.0 mm) on a barrier wall normal to flow. They found that for Fr < ∼3, a reflected wave formed; otherwise, a vertical jet was produced. This finding is not in complete agreement with our results (Table 2), especially for the mixture PSD2, for which we observed a reflected wave impact at fc = 24 percent, corresponding to Fr > 3.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 127–138

131

Sanvitale, Bowman, and Cabrera

Figure 4. Measured run-up height at the barrier (time t = 0 is the time of the ﬂow front arrival).

The transition from reflected wave to jet behavior lies in the 3 > Fr > 4 zone, but it is influenced also by particle distribution and fluid content. Figure 4 shows the evolution of the flow height at the barrier measured on the video footages (note the truncation of the plotted height for heights greater than the wall). PSD2 exhibited the greatest height at the barrier at the end of the test, which explains the greater final static load compared to the other mixtures. Both PSD2 experiments at fc = 28 percent and 32 percent were able to overtop the barrier. Overtopping also occurred for PSD3 at fc = 32 percent in the final part of the event due to the large fluid content, which was well above that needed to fully saturate (i.e., fill the voids between particles) the mixture, so that it could flow easily towards the barrier, increasing the height of the free surface of the fluid at values higher than the top of the wall. The PSD3 at fc = 28 percent test displayed an initial height peak at 0.28 seconds, which was due to the vertical jet travelling up the wall at the beginning of the impact. After that, a second spike followed at t = 1.1 seconds, due to the accumulation of fluid behind the barrier at the end of the impact, as occurred for the test PSD3 at fc = 32 percent. This excess of fluid after reaching the barrier was reflected and moved upslope. The height of PSD1 tests was always lower than barrier and reached values at the end of the tests comparable with PSD3. Barrier Load The recorded raw barrier load signals showed highfrequency spikes that were due to random effects depending on the resonance frequency of the load cell and on the single instantaneous impact of large particles. In order to filter the data, we followed the pro-

132

cedure proposed by Scheidl et al. (2012), applying a low-pass filter with a maximum high frequency. This high frequency was estimated by considering the average maximum front velocity, vi , from the PIV analysis, and the maximum particle diameter, as fi = vf /dmax . For PSD3, dmax as taken as d90 = 20 mm. The resulting low-pass filters for each test are listed in Table 1. Time histories for measured basal pressures and barrier loads for all tests are given in Figure 5 (fc = 24 percent), Figure 6 (fc = 28 percent), and Figure 7 (fc = 32 percent). In each case, the load on the barrier was characterized by an initial dynamic phase and a subsequent final static value. The dynamic phase was due to the impact loading exerted by the incoming flow against the barrier, while the static value was given by the pressure exerted by the deposited material behind the wall at the end of the event. The tests with fc = 24 percent for all particle size distributions exhibited a gradual increase of the barrier load due to the continuous accumulation of material behind the barrier. The PSD2 and PSD3 tests showed many spikes in the signal related to the instantaneous impacts of single large beads. The highest peak force was reached by the PSD2 tests because for these mixtures, more material was able to reach the barrier, increasing the height of the static deposit (Figure 4). The high mobility of these flows was confirmed by their front velocities, which were highest for all fluid contents (Table 1). Furthermore, it has to be noted that for the same mass of dry particles, the fluid necessary to saturate the sample was different for the three particle size distributions. Before testing, we measured that the saturation of the sample in the tank was reached with 22 percent, 20 percent, and 13 percent of fluid content for the 3 mm, 7.5 mm, and well-graded mixtures, respectively. Hence, the fluid content in excess of saturation that was available to fluidize the mixture was potentially higher for larger bead sizes. (Note that this picture becomes less clear for the well-graded material, PSD3, due to segregation as the flow developed.) The tests at fc = 28 percent exhibited larger Froude numbers (2.3, 3.9, and 3.3 for PSD1, 2, and 3, respectively), and the observed impact mechanism consisted of the formation of a vertical jet for PSD2 and PSD3. In these tests, during their dynamic interaction with the barrier, the flows developed a spike on the barrier load curve corresponding to the instant at which the jet fell back on the free surface of the incoming flow (Figure 6). This peak was followed by a transient decrease of the load due to the energy dissipation caused by the hydraulic jump subsequent to the falling jet breaking on the flow surface. For the PSD2 test, this spike represents the highest force, being larger than the final static load. For the PSD3 test, before the impulse due to the falling jet, the barrier experienced a series of

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 127–138

Impact Dynamics of Saturated Granular Flows

Figure 5. Fluid content 24 percent. (Left) Load on barrier. (Right) Basal pressures. (a) PSD1, (b) PSD2, (c) PSD3.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 127–138

133

Sanvitale, Bowman, and Cabrera

Figure 6. Fluid content 28 percent. (Left) Load on barrier. (Right) Basal pressures. (a) PSD1, (b) PSD2, (c) PSD3. The red oval points out the breaking of the jet on the free surface of the incoming ﬂow.

134

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 127–138

Impact Dynamics of Saturated Granular Flows

Figure 7. Fluid content 32 percent. (Left) Load on barrier. (Right) Basal pressures. (a) PSD1, (b) PSD2, (c) PSD3. The red oval points out the breaking of the jet on the free surface of the incoming ﬂow.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 127–138

135

Sanvitale, Bowman, and Cabrera

load peaks due to the largest particles accumulating at the front (due to particle size segregation) and colliding with the wall, although these transient peaks were lower than the final static load exerted on the wall by the total material accumulated during the event. Intermediate to this, for the PSD1 test at fc = 28 percent, the flow impact produced an initial small jet that gave rise to the formation of a surge propagating upstream. This was then stopped by the incoming flow to evolve into a type of reflected wave. The tests at fc = 32 percent for all the particle size distributions showed that the impact mechanism was always jet-like at larger Froude numbers. At this fluid content, the PSD1 mixture exhibited the largest front velocity, and after the sudden increase of the barrier load when the flow front arrived, it developed a vertical jet with a corresponding increase in the barrier force up to a maximum when the falling jet broke on the incoming flow. After this peak, the energy dissipation due to the formation of the hydraulic jump led to a decreasing load, which was followed by a gradual growth due to the accumulation of material from the tail of the debris-flow surge. For PSD2 and PSD3, which had larger particles, the fluid content had less influence on the flow behavior. PSD2 at fc = 32 percent showed the highest run-up of the vertical jet, which resulted both in some material overtopping the barrier and a higher barrier load when the falling jet broke. The well-graded material, PSD3, showed a similar impact behavior compared to that with fc = 28 percent, with the main difference being a greater final static load due to more material being mobilized to reach the barrier. The influence of the particle size on the response of the barrier is clear. PSD2, the uniform flows with the 7.5 mm particles, generated the greatest loads. For the well-graded PSD3 tests, similar to PSD2, several spikes in the barrier signal were recorded from the load cell, representing collisions of large particles against the barrier; however, the peak force during and after the impact reached final values similar to those of the PSD1 mixture. The peak load due to the initial dynamic impact of the flows appears to have been enhanced by the presence of a larger quantity of fluid for all the mixtures. In fact, the higher fluid content allowed the mixtures to be more fluidized and hence to reach higher velocities when they moved downslope. The larger fluid content also enhanced the mobility of the flows, increasing the final static load at the end of the impact due to the greater accumulation of material behind the wall (Figure 4). Only the PSD2 (uniform, 7.5 mm) tests exhibited a consistently higher peak barrier load from the initial dynamic impact than the static load exerted from the material deposited subsequently. The reason for this

136

is that once these flows hit the barrier at high speed (Table 2), they produced a higher jet wave than other tests, with large particles that were easily mobilized and pushed upward against the barrier. In contrast, it is clear from the high-speed images that for the fc = 28 percent tests with PSD3, most of the top part of the run-up wave was composed of fluid, because the large particles at the flow front were too heavy to be pushed any higher than approximately the middle height of the barrier (Figure 4). Furthermore, the presence of finer material can also have a damping effect on the large particle collisions. The combination of these factors can explain the similar value of the impact load between the 3 mm (PSD1) and the well-graded (PSD3) tests. Basal Pressure Development Figures 5 to 7 show the responses of the porepressure transducers (PPT) using a running average filtering window of 400 data points and the evolution of the basal total pressure, σtot . The PPT responses were dominated by the increase in the height of the fluidsaturated debris behind the barrier (which was effectively impermeable) after impact. Therefore, although flows initially passed over PPT4 and then PPT3 and then PPT2 in succession on their descending motion (resulting in relatively small recorded pressures of the order of 0.2 kPa), PPT2 (closest to the barrier) produced the largest and earliest response to this impact, with recorded pressures ranging between 1 and 2 kPa, except for the 3 mm test with fc = 24 percent, which was characterized by pore pressure below 0.5 kPa. For all the tests at fc = 24 percent, the impact was characterized by the arrival of an initially unsaturated flow front, for which the fluid pressure was absent due to the particles at the front running ahead of the fluid. This became more acute as the particle sizes segregated, with the largest particles trending to the front, as seen in field-scale debris flows (Iverson, 1997). Considering the measurements closest to the barrier (PPT2 and the basal load cell), differences were found in the pore-pressure behavior after impact between flows at different fluid contents, particularly for the 3 mm flows (PSD1), which, considering the ratio of pore pressure to total stress, for fc = 24 percent showed the pore pressure to be always below hydrostatic, while for fc = 28 percent, it was close to hydrostatic, and for fc = 32 percent, it was above hydrostatic immediately after the breaking of the falling jet. For the well-graded flows (PSD3) at both fluid content fc = 28 percent and 32 percent, the pore pressures were much greater than hydrostatic until the hydraulic jump, which developed after the breaking of the falling jet, ended. For the well-graded flows at fc = 24 percent,

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 127–138

Impact Dynamics of Saturated Granular Flows

after the arrival of the unsaturated front, the pore pressure was much larger than hydrostatic until the accumulation of material behind the barrier occurred, and then it dropped to typical hydrostatic values. For the 7.5 mm flows, the pore pressure was hydrostatic for fc = 24 percent, slightly above for 28 percent, and largest for 32 percent. These results demonstrate that excess (i.e., greater than hydrostatic) pore pressures are not necessarily generated within uniform flows of spheres, except where sufficient fluid is present and sufficient agitation is generated (e.g., during an impact event). For well-graded flows (at least for the chosen grading and fluid contents examined here), excess pore pressures are both generated and maintained at impact. This is likely to be due to both larger particles agitating the flow upon impact and fines reducing the mixture permeability and therefore maintaining the developed excess pore pressure for longer durations. CONCLUSIONS This paper presented an experimental investigation on the effects of fluid content and particle size on the impact force generated by a transparent debrisflow model on a rigid barrier. The debris-flow models were simulated by using refractive index-matched mixtures of borosilicate glass beads in a Newtonian fluid. Small-scale flume experiments were carried out using a channel equipped on the bottom with three porepressure transducers and a load cell for the measurement of the total normal stress and fluid pore pressure. A rigid barrier, instrumented with another load cell, was fixed normal to the flume bed at 2.25 m from the gate release. The evolution of the impact load, bed normal pressure, and fluid pore pressure for flows consisting of uniform and well-graded particle sizes at three different fluid contents, 24 percent, 28 percent, and 32 percent, were measured and analyzed. Results showed that excess pore pressures were not necessarily generated within uniform flows of spheres, except where sufficient fluid was present and sufficient agitation was generated (e.g., during an impact event). The particle size of the material had a strong influence on impact loading and overall response. The uniform flows with the largest particles generated the greatest load, while for the well-graded tests, the presence of fine particles within the flow could provide a dampening influence. Larger fluid content led to greater flow velocity and larger peak load in the initial dynamic phase of impact of the flows. Increasing the amount of fluid content also enhanced the overall mobility of the flows, increasing the final static load at the end of the impact due to the greater accumulation of material behind the wall.

ACKNOWLEDGMENTS This research was supported through the Engineering and Physical Sciences Research Council (EPSRC), UK project no. EP/M017427/1, “High speed granular debris flows: New paradigms and interactions in geomechanics.” The authors would like to acknowledge the assistance provided by technicians at the University of Sheffield in the construction of the apparatus. MAC was partly funded by the Early-stage Researcher fund (FAPA) from the Universidad de Los Andes, under grant agreement no. PR.3.2016.3667. REFERENCES Ahmadipur, A.; Qiu, T.; and Sheikh, B., 2019, Investigation of basal friction effects on impact force from a granular sliding mass to a rigid obstruction: Landslides, Vol. 16, pp. 1089–1105, https://doi.org/10.1007/s10346-019-01156-0. Albaba, A.; Lambert, S.; and Faug, T., 2018, Dry granular avalanche impact force on a rigid wall: Analytic shock solution versus discrete element simulations: Physical Review E, Vol. 97, No. 5, pp. 052903, https://doi.org/10.1103/PhysRevE.97.052903. Arattano, M. and Franzi, L., 2003, On the evaluation of debris flows dynamics by means of mathematical models: Natural Hazards and Earth System Sciences, Vol. 3, No. 6, pp. 539–544. doi:10.5194/nhess-3-539-2003. Armanini, A., 1997, On the dynamic impact of debris flows. In Armanini, A. and Michiue, M. (Editors), Recent Development on Debris Flows: Vol. 64, Springer, Berlin, pp. 208–226. Armanini, A.; Larcher, M.; and Odorizzi, M., 2011, Dynamic impact of a debris flow front against a vertical wall. In Genevois, R. and Douglas, L. (Editors), Proceedings of the 5th International Conference on Debris-flow Hazard Mitigation: Casa Editrice Università La Sapienza, Rome, Italy, pp. 1041– 1049. Armanini, A.; Rossi, G.; and Larcher, M., 2019, Dynamic impact of a water and sediments surge against a rigid wall: Journal of Hydraulic Research, Vol. 58, No. 2, pp. 314–325. doi:10.1080/00221686.2019.1579113. Bugnion, L.; McArdell, B.; Bartelt, P.; and Wendeler, C., 2011, Measurements of hillslope debris-flow impact pressure on obstacles: Landslides, Vol. 9, No. 2, pp. 179–187. doi:10.1007/s10346-011-0294-4. Choi, C. E.; Au-Yeung, S. C. H.; Ng, C. W. W.; and Song, D., 2015, Flume investigation of landslide granular debris and water runup mechanisms: Géotechnique Letters, Vol. 5, No. 1, pp. 28–32. Cui, P.; Zeng, C.; and Lei, Y., 2015, Experimental analysis on the impact force of viscous debris flow: Earth Surface Processes Landforms, Vol. 40, pp. 1644–1655. doi:10.1002/esp.3744. Faug, T.; Childs, P.; Wyburn E.; and Einav, I., 2015, Standing jumps in shallow granular flows down smooth inclines: Physics of Fluids, Vol. 27, No. 7, pp. 073304, https://doi.org/10.1063/1.4927447. Goujon, C.; Thomas, N.; and Dalloz-Dubrujeaud, B., 2003, Monodisperse dry granular flows on inclined planes: Role of roughness: European Physical Journal E Soft Matter, Vol. 11, No. 2, pp. 147–157. doi:10.1140/epje/i2003-10012-0. PMID:15011055. Gray, J. M. N. T.; Tai, Y. C.; and Noelle, S., 2003, Shock waves, dead zones and particle-free regions in rapid granu-

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 127–138

137

Sanvitale, Bowman, and Cabrera lar free-surface flows: Journal of Fluid Mechanics, Vol. 491, pp. 161–181. Hu, K.; Wei, F.; and Li, Y., 2011, Real-time measurement and preliminary analysis of debris-flow impact force at Jiangjia Ravine, China: Earth Surface Processes Landforms, Vol. 36, pp. 1268–1278. doi:10.1002/esp.2155. Hübl, J.; Suda, J.; Proske, D.; Kaitna, R.; and Scheidl, C., 2009, Debris flow impact estimation. In Eleventh International Symposium on Water Management and Hydraulic Engineering, Eds: Popovska C. and Jovanovski M., Ohrid, Macedonia, Vol 1, pp. 137–148. Hungr, O.; Leroueil, S.; and Picarelli, L., 2013, The Varnes classification of landslide types, an update: Landslides, Vol. 11, pp. 167–194. Iverson, R. M., 1997, The physics of debris flows: Reviews of Geophysics, Vol. 35, No. 3, pp. 245–296. Moriguchi, S.; Borja, R.; Yashima, A.; and Sawada, K., 2009, Estimating the impact force generated by granular flow on a rigid obstruction: Acta Geotechnica, Vol. 4, No. 1, pp. 57–71. Sanvitale, N. and Bowman, E. T., 2012, Internal imaging of saturated granular free-surface flows: International Journal Physical Modelling in Geotechnics, Vol. 12, No. 4, pp. 129–142. Sanvitale, N. and Bowman, E. T., 2017, Visualization of dominant stress-transfer mechanisms in experimental debris flows of different particle-size distribution: Canadian Geotechnical Journal, Vol. 54, No. 2, pp. 258–269. Scheidl, C.; Chiari, M.; Kaitna, R.; Mullegger, M.; Krawtschuk, A.; Zimmermann, T.; and Proske, D., 2012, Analysing debris-flow impact models, based on a small

138

scale modelling approach: Surveys in Geophysics, Vol. 34, No. 1, pp. 121–140. Silbert, L. E.; Ertaş, D.; Grest, G. S.; Halsey, T. C.; Levine, D.; and Plimpton, S. J., 2001, Granular flow down an inclined plane: Bagnold scaling and rheology: Physical Review E, Vol. 64, No. 5, pp. 051302. doi:10.1103/PhysRevE.64.051302. Stöhr, M.; Roth, K.; and Jahne, B., 2003, Measurement of 3D pore-scale flow in index-matched porous media: Experiments in Fluids, Vol. 35, No. 2, pp. 159–166. Takahashi, T., 2007, Debris Flow. London: Taylor & Francis, https://doi.org/10.1201/9780203946282. Thielicke, W., 2014, The Flapping Flight of Birds—Analysis and Application: Ph.D. thesis, Rijksuniversiteit Groningen, Groningen, Netherlands. http://irs.ub.rug.nl/ppn/ 382783069. Thielicke, W. and Stamhuis, E. J., 2014, PIVlab—Towards user-friendly, affordable and accurate digital particle image velocimetry in MATLAB: Journal Open Research Software, Vol. 2, No. 1, pp. e30. doi: http://dx.doi.org/10.5334/ jors.bl. Van Dine, D. F., 1996, Debris Flow Control Structures for Forest Engineering: Working paper, Ministry of Forest Research Program, Victoria, British Columbia, Canada. Zhang, S., 1993, A comprehensive approach to the observation and prevention of debris flows in China: Natural Hazards, Vol. 7, pp. 1–23. doi:10.1007/BF00595676. Zhou, G. G. D.; Song, D.; Choi, C. E.; Pasuto, A.; Sun, Q. C.; and Dai, D. F., 2018, Surge impact behaviour of granular flows: Effects of water content: Landslides, Vol. 15, No. 4, pp. 695–709.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 127–138

The Eﬀects of Particle Segregation on Debris Flow Fluidity Over a Rigid Bed NORIFUMI HOTTA* Graduate School of Agricultural and Life Sciences, The University of Tokyo, 1-1-1 Yayoi, Bunkyo-ku, Tokyo 113-8657, Japan

TOMOYUKI IWATA Chiba Prefectural Government Oﬃce, 1-1 Ichiba-cho, Chuo-ku, Chiba City, Chiba 260-8667, Japan

TAKURO SUZUKI Forestry and Forest Products Research Institute, 1 Matsunosato, Tsukuba, Ibaraki 305-8687, Japan

YUICHI SAKAI Graduate School of Science, Kyoto University, Kitashirakawaoiwake-cho, Sakyo-ku, Kyoto-shi, Kyoto, 606-8502, Japan

Key Terms: Debris Flow, Flume Test, Numerical Simulation, Particle-Size Segregation ABSTRACT It is essential to consider the fluidity of a debris flow front when calculating its impact. Here we flume-tested mono-granular and bi-granular debris flows and compared the results to those of numerical simulations. We used sand particles with diameters of 0.29 and 0.14 cm at two mixing ratios of 1:1 and 3:7. Particle segregation was recorded with a high-speed video camera. We evaluated the fronts of debris flows at 0.5-second intervals. Then we numerically simulated one-dimensional debris flows under the same conditions and used the mean particle diameter when simulating mixed-diameter flows. For the mono-granular debris flows, the experimental and simulated results showed good agreement in terms of flow depth, front velocity, and flux. However, for the bigranular debris flows, the simulated flow depth was less, and both the front velocity and flux were greater than those found experimentally. These differences may be attributable to the fact that the dominant shear stress was caused by the concentration of smaller sediment particles in the lower flow layers; such inverse gradations were detected in the debris flow bodies. Under these conditions, most shear stress is supported by smaller particles in the lower layers; the debris flow characteristics become similar to those of mono-granular flows, in con-

*Corresponding author email: hotta.norifumi@fr.a.u-tokyo.ac.jp

trast to the numerical simulation, which incorporated particle segregation with gradually decreasing mean diameter from the front to the flow body. Consequently, the calculated front velocities were underestimated; particle segregation at the front of the bi-granular debris flows did not affect fluidity either initially or over time. INTRODUCTION Stony debris flows, which contain mainly coarser sediment, have been modeled and described using constitutive equations based on Bagnold’s relation (Bagnold, 1954; Takahashi, 1991) or the principles of continuum mechanics (Hutter et al., 1996; Iverson, 1997; Berzi et al., 2010; and Bouchut et al., 2016), assuming dominance of the interparticle stress of frictional contacts and particle–particle collisions. Although the solid phase is often described using only Coulomb friction, which is independent of the shear rate, evidence for the importance of both interparticle friction and collisions for stony debris flows has been provided experimentally (Armanini et al., 2005, 2009; Lanzoni et al., 2017). To account for collisional stress in stony debris flows, researchers generally model the constitutive relations for coarse-grained debris flows under conditions of uniform particle size (Takahashi, 1991; Egashira et al., 1997; and Berzi and Jenkins, 2008). Hence, numerical simulations with the assumption of a representative particle size have been performed to reproduce and predict debris flow behaviors (Nakagawa and Takahashi, 1997; Osti and Egashira,

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 139–149

139

Hotta, Iwata, Suzuki, and Sakai

2008). However, natural debris flows have a wide range of grain sizes, from clay and silt to boulders (Major and Pierson, 1992; Berti et al., 1999; and Coe et al., 2008), which leads to particular flow characteristics, such as the suspension of finer sediment in a pore fluid (Iverson, 1997; Kaitna et al., 2016) or the segregation of coarser grains to the flow surface and the front (Takahashi et al., 1992; Stock and Dietrich, 2006; Suwa et al., 2009; and Yohannes et al., 2012). Particle size markedly affects debris flow fluidity. Fine sediment entrained into the pore water resulting in liquefaction leads to increasing density of the fluid phase and decreasing effective stresses of the solid fraction in a debris flow (Nishiguchi et al., 2012; Hotta et al., 2013; and Sakai et al., 2019). However, larger particles impart higher flow resistance, as collisional stresses are greatly affected by particle size (Takahashi, 1991; Egashira et al., 1997). Hence, the fluidity of the debris flow front, which is important in terms of impact forces, is affected by both particle size and particle admixing. The accumulation of boulders at the front causes the flow characteristics of the front to differ from those of the main body; these factors cannot be reflected in numerical simulations that use particles of uniform size. Particle segregation has been investigated extensively for dry granular matter. Segregation is considered and modeled as being induced by kinetic sieving and squeeze expulsion (Savage and Lun, 1988; Gray and Thornton, 2005), in which smaller particles fall into the pore spaces between coarse particles located in the lower layer in the flow and coarse particles are pushed upward to maintain the mass balance. Hill and Tan (2014) noted that both collisional stress and gravity contribute to segregation in dense flows. This inverse grading is observed in flume tests of the debris flow (Takahashi et al., 1992; Johnson et al., 2012). The uplifted coarse grains are delivered forward because of the vertical profile of velocity, and the coarse grains accumulate in the debris flow front. Based on this process, Takahashi et al. (1992) developed a numerical simulation to incorporate a gradually decreasing grain size distribution from the front to the main body of a debris flow. By contrast, Iwata et al. (2013) compared flume tests using particles of small and large sand of uniform size and mixtures and found that the mean diameter of the flow portion considering particle segregation did not account for the fluidity of debris flow. In terms of applicability of numerical simulation, the influence of particle segregation on the fluidity of a debris flow can be assessed effectively by comparing debris flows of uniform particle size to clarify the representative particle size. This study was performed to determine the influence of particle size segregation on the fluidity of multi-

140

Figure 1. Experimental setup.

granular debris flows. We flume-tested mono-granular and bi-granular debris flows and compared the results with those of numerical simulations to determine the effects of particle segregation on the debris flow front. MATERIALS AND METHODS Flume Test A channel of variable slope (10 m long and 10 cm wide) with a glass sidewall on the left side was used in all experiments (Figure 1). The slope angle was set to 15°. The upper 3.5 m of the channel was filled with sand particles to a depth of 10 cm and connected to a lower stream of 5 m in length; this was a rigid bed measuring 10 cm in height to the surface, to which sand particles of 0.29-cm diameter were glued to impart roughness. To prevent overflow, the sand was watered to near-saturation immediately before each test. A steady flow of water (2,000 mL/s) was supplied from the upper end of the channel to generate debris flow by eroding the deposited sand. Silica sands with particles measuring 0.29 and 0.14 cm in diameter were used at mixing ratios of 50 percent:50 percent (5:5) and 30 percent:70 percent (3:7; Table 1). Particle size distribution affects the flow characteristics of debris flows in several ways. For example, fine sediment and its liquefaction change the fluidity. In this study, we focused on particle segregation in stony debris flows. Sand particles of 0.29 and 0.14 cm were selected because they behave as representative stony debris flows under these experimental conditions (Hotta and Miyamoto, 2008; Hotta, 2012). Mono-granular debris flows using each particle size and bi-dispersed mixtures would simplify and clarify the particle segregation process. Table 1. Silica sands used in the flume tests. Mixing Ratio (0.29 cm:0.14 cm)

Mean Diameter (cm)

10:0 5:5 3:7 0:10

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 139–149

0.29 0.22 0.19 0.14

Particle Segregation and Debris Flow Fluidity

Eight ultrasonic displacement sensors (E4C; Omron, Kyoto, Japan) were placed above the channel at 0.5-m intervals from 0.5 to 4.5 m from the downstream end to monitor flow depth and timing. Temporal data were used to calculate front velocities. A highspeed video camera (EX-F1; Casio, Tokyo, Japan) was placed 0.5 m from the downstream end of the flume and recorded the debris flow from the side at 600 frames per second. We used the resulting images to evaluate the vertical velocity profiles and the locations of the larger (0.29 cm in diameter) particles by tracking the particles through the sequence of images. Five debris flow samples from the front edges were collected at intervals of approximately 0.5 seconds into a container with five separate compartments (Figure 1), and one sample was also obtained from the main body at the lower end of the channel. We measured sediment concentrations and particle segregation. Numerical Simulation Debris flows are generally described using depthaveraged model equations based on laws of conservation of mass and momentum (Takahashi, 1991; Iverson, 1997). Debris flow mixtures can be treated as a quasi-single phase when the relative velocity between the solid and fluid phases is neglected (Takahashi, 1991; Egashira et al., 1997; Armanini et al., 2009; and Xia et al., 2018). By contrast, a two-phase model that includes individual momentum balance equations for the solid and fluid phases should be used when the relative velocity between the two phases is significantly different, resulting in an interaction between them (Iverson and George, 2014; Li et al., 2018). We used a quasi–single-phase model to perform the numerical simulation. We performed a one-dimensional numerical simulation of the debris flows. When modeling debris flows that contained particles of two different diameters, we used the volumetric mean diameter (Table 1). The equations included a continuity equation for the debris flow, a continuity equation for the sediment, and a momentum equation, as follows: ∂h ∂M + = E; ∂t ∂x

(1)

∂ (c̄h) ∂ (ct M) + = Ec∗ ; ∂t ∂x

(2)

∂ (uM) ∂H τ0 ∂M +β = −gh − ; ∂t ∂x ∂x ρm

(3)

where h is the ﬂow depth, M is the discharge rate per unit width, E is the bed entrainment rate, c̄ is the mean

cross-sectional sediment concentration, ct is the transported sediment concentration, c* is the sediment concentration deposited in the channel, β is a compensation coefficient for momentum, u is the cross-sectional average velocity, g is the acceleration due to gravity, H is the elevation of the flow surface (H = h + zb , where zb is the bed elevation), τ0 is the shear stress at the bed, and ρm is the density of debris flow. c̄ and ct are identical if one assumes a uniform profile of sediment concentration. Our measurements indicated that c* was 0.60. For τ0 , Miyamoto and Itoh (2002) developed constitutive equations consisting of static yield stress and dynamic shear stress, as follows: τ0 = τ0y + ρ fb u2 ,

(4)

where ρ is the density of water. The yield stress τ0y and the friction coeﬃcient fb are given as follows, respectively: 15 c̄ (5) τ0y = (σ − ρ) c̄ghcos θ tan φs ; c∗ h −2 25 Kg + Kf fb = ; 4 d

(6)

where σ is the density of the sediment particles (2.64), θ is the bed slope angle, φs is the internal friction angle of the sediment particles (34.0°), and d is the mean diameter of the sediment particles. Kg and Kf represent shear stress terms by particle–particle collision and turbulence in the pore water and are expressed as follows: 1 σ (7) Kg = kg 1 − e2 c̄ 3 ; ρ 5

Kf = k f

(1 − c̄) 3

; (8) 2 c̄ 3 where kg is an experimental constant that was reported to be 0.0828 (Miyamoto, 1985), according to Itoh et al. (1999), and e is the coefficient of restitution of sediment particles with a value of 0.85. kf reflects the interstitial space and is often given as a constant, which Egashira et al. (1988) evaluated as 0.16. In this study, we used the following equation reported by Suzuki (2007) by back-analyzing flume tests of debris flows with relatively low sediment concentration, as the sediment concentration of our flume tests for 0.14-cm cases was low (∼0.2): 1 1 (9) k f = − c̄3 + c̄2 + 0.025. 2 2 According to Eq. 9, kf corresponds to 0.16 when c is 0.43. To include erosion and deposition in numerical simulations of debris flows, the entrainment rate equation

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 139–149

141

Hotta, Iwata, Suzuki, and Sakai

should be incorporated into the governing equations (Hotta et al., 2015; Iverson et al., 2015). The existing entrainment rate equations can be divided into two groups: one considers the difference between the current and equilibrium conditions, such that debris flows reach an equilibrium state (Takahashi, 1991; Egashira et al., 2001), and the other incorporates the dependence on boundary traction derived from momentum exchange at flow-bed boundaries (Fraccarollo and Capart, 2002; Medina et al., 2008). Although entrainment equations in the former group do not explicitly depend on stress, they implicitly concern physical processes via the equilibrium state determined by the balance between the driving force and friction at the surface of the bed. We used the entrainment rate equation of Egashira et al. (2001) for E to close Eqs. 1 through 3: tanθe =

c̄ (σ/ρ − 1) tan φs ; c̄ (σ/ρ − 1) + 1

E = utan (θ − θe ) ;

(10)

(11)

where θe is the equilibrium bed slope, which can be calculated based on a given sediment concentration (Takahashi, 1978). Analyses We first compared the simulations with the experiments in terms of flow depth, discharge rate, and velocity. The representative diameters in the simulations corresponded to 0.29 cm, 0.14 cm, and the mean diameter (Table 1), respectively. After confirming that particle segregation occurred under the experimental conditions in this study, we performed a numerical simulation considering particle segregation similar to that reported previously by Takahashi et al. (1992), incorporating a gradual decrease in the representative particle size at the front of a debris flow. This calculation used the same model as introduced in the previous section but was applied to a rigid bed in whole, with a continuous steady debris flow from the upper end of the simulation domain without evaluation of erosion so that we could clarify differences in performance after simulating the distances descended by mono- and bi-granular debris flows. RESULTS Flow Depth and Discharge The experimental and simulated results showed close agreement in terms of the depths of monogranular but not bi-granular debris flows (Figure 2). For the mono-granular flows, flow depths of 0.29- and

142

Figure 2. Debris flow depths over time of (a) mono-granular flows and (b, c) bi-granular flows at particle mixing ratios (larger:smaller) of 5:5 and 3:7, respectively, at a point 0.5 m from the downstream end. The experimental flow depths were smoothed with an 0.4-second moving average.

0.14-cm particles were clearly different, and the numerical simulation succeeded in reproducing these differences, including the wave profile: greater at the front and then gradually decreasing in the flow body (Figure 2a). The mean diameter did not account for the experimental depths of bi-granular flows. These values were very similar to those predicted for monogranular flows of smaller particles (0.14 cm in diameter), regardless of the mixing ratio (5:5 or 3:7 of

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 139–149

Particle Segregation and Debris Flow Fluidity

Figure 3. Debris flow discharge over time of (a) mono-granular flows and (b) bi-granular flows.

particles measuring 0.22 and 0.19 cm in diameter; Figure 2b and c). By contrast, the discharges did not differ markedly according to grain size (Figure 3). In the experimental data, the discharges of mono-granular debris flows of particles 0.29 and 0.14 cm in diameter differed slightly, but the calculations did not reveal any distinct difference (Figure 3a). For the 0.14-cm sand, the calculations failed to reproduce the discharge at the front but showed agreement in the following part (5 to 9 seconds). The calculated and experimental data for the particles 0.14 cm in diameter and mixed-particle debris flow fronts disagreed (Figure 3b). The discharges were similar at particle mixing ratios of 5:5 and 3:7, as were the flow depths. Velocity The experimental and calculated frontal velocities of the mono-granular debris flows (both initially and over time) showed relatively good agreement (Figure 4a). Both experiments and calculations showed increasing velocity over time, which suggests that the debris flows accelerated and did not reach terminal velocity

Figure 4. Debris flow frontal velocities over time of (a) monogranular flows and (b, c) bi-granular flows at particle mixing ratios (larger:smaller) of 5:5 and 3:7, respectively. Flow distance indicates the distance from the upper end of the rigid bed section (see Figure 1).

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 139–149

143

Hotta, Iwata, Suzuki, and Sakai

Figure 6. Experimental particle mixing ratios of the fronts and main bodies of debris ﬂows.

Figure 5. Comparison of vertical velocity distributions of the front (left) and main body (right) of debris ﬂows at particle mixing ratios (larger:smaller) of (a) 5:5 and (b) 3:7. Solid lines indicate average ﬂow depths, with error bars showing the standard deviation.

within the flume, although water was constantly supplied from the upper end in a straight channel with a constant slope angle. The experimental frontal velocities of the bigranular debris flows were initially the same as those of the 0.29-cm particle mono-granular flow, regardless of the mixing ratio, and matched those of debris flows containing particles of diameters equal to the means of 0.14 and 0.29 cm (Figure 4b and c). We cannot make a judgment about whether the acceleration of the bigranular debris flows terminates when it reaches the velocity of the mean-diameter debris flow or continues to approach that of the 0.14-cm particle monogranular flow similar to flow depth (Figure 2) because of the limited flume length. Figure 5 shows the experimental vertical distributions of particle velocities within the bi-granular debris flows. The velocities of the 0.14- and 0.29-cm particles did not differ at the same depth. The velocity profile indicated that the inclination was steeper in the flow body (7.9 and 6.9 seconds after the front had passed) than at the front (photos taken at 0.3 seconds; Figure 5a and b, respectively). Although the flow body showed the typical velocity profile of a debris flow on a rigid bed, the frontal velocity profile was similar to that

144

of a debris flow on an erodible bed, in which the total stress at the bed is supported by static share stress, resulting in a moderate velocity profile near the bottom (Egashira et al., 1997; Itoh et al., 1999). Both 0.14- and 0.29-cm particles were tracked to determine the velocity profile, and no differences in speed were detected according to particle size. Particle Segregation In the experimental data, the larger particles (0.29 cm) accumulated at the front. The ratio of the larger particles showed the maximum value at the front end and a smooth decrease toward the flow body (Figure 6). The extent of accumulation clearly differed according to the mixing ratio; the flow body retained the initial mixing ratio. Figure 7 compares large particle accumulation in the upper flow between the front and the main body presenting apparent inverse grading; small particles predominated in the most inclined section of the velocity profile (Figure 5b). The extent of inverse grading was more significant in the main body. These results confirmed that particle segregation was induced in the bi-granular debris flows in this study, as reported previously (Takahashi et al., 1992; Johnson et al., 2012). Numerical Simulation Incorporating Particle Segregation We compared the simulated profiles of surge propagation for the mono-granular and bi-granular debris flows. Figure 8 shows the variation over time in the representative particle sizes for the mono-granular and bi-granular flows with reference to the experiments (Figure 6); the frontal accumulation of large

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 139–149

Particle Segregation and Debris Flow Fluidity

Figure 7. Frequency distributions of particles 0.29 cm in diameter in the upper and lower layers of the experimental debris ﬂows with particle mixing ratios (larger:smaller) of (a) 5:5 and (b) 3:7. N indicates the number of 0.29-cm particles in each image taken using a high-speed video camera. The relative ﬂow depth was normalized relative to the surface level.

boulders was simply modeled by gradually decreasing the particle size. The larger particles accumulated at the front and the smaller particles accumulated in the main body. For the mono-granular debris flow, the particle size of smaller particles was given. The numerical simulations showed that the initial flow depth of the bi-granular flow was higher than that of the mono-granular flow (Figure 9), because the particles at the front of the flow were larger, and flow resistance theoretically increases with particle size (Eq. 6). This result conflicted with our experimental results (Figure 2), which showed that our bi-granular debris flows were similar to those consisting of smaller particles. Therefore, the results of our calculations based on a particle segregation model that simply incorporated the mean particle size transition differed markedly from our experimental results. In the numerical simulations, the mono-granular flow exhibited steady motion (Figure 9a) that differed from the experimental results (Figure 4a). The frontal velocity of the bi-granular flow started at a lower speed

than did the mono-granular ﬂow and then increased over time, reaching 129 cm in 0 to 1 second and 163 cm in 5 to 6 seconds (Figure 9b), as the main body caught up to the front; this occurred because the main body consisted of smaller particles and thus had greater velocity. This result can be explained by the constitutive

Figure 8. Comparison of time series of representative particle sizes for mono-granular and bi-granular debris ﬂows.

Figure 9. Simulated proﬁles of surges for (a) mono-granular and (b) bi-granular debris ﬂows.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 139–149

145

Hotta, Iwata, Suzuki, and Sakai

equation (Eq. 6), which corresponds to the experimental results (Figure 4b and c). DISCUSSION Reliability of Numerical Simulation for Mono-Granular Debris Flows The numerical simulation reproduced well the experimental results for mono-granular debris flows. The flow depth in particular showed good agreement, including differences according to particle size (Figure 2). This suggests that constitutive equations sufficiently describe debris flow characteristics, as the flow depth and the particle size directly affect the flow resistance, as shown in Eqs. 4 and 6, in which the relative flow depth (h/d) can be seen to directly affect the shear stress. The discharge was not perfectly duplicated for smaller particles in the flow front (Figure 3a). This could be an issue with the entrainment equations. The entrainment equations of Egashira et al. (2001) are simpler than those of Hotta et al. (2015), among the other entrainment equations without any fitting parameters (Eqs. 10 and 11). The water supplied was identical in each case, and consequently the difference in discharge should have been induced by the sediment concentration determined by erosion in the upper erodible bed section of the flume in which the debris flow was generated. In Eq. 10, the entrainment rate equation governing the sediment concentration implicitly incorporates particle size through velocity (Hotta et al., 2015). The difference in discharge by particle size could therefore be achieved by variations in the entrainment process that are insufficiently reproduced by the model. In addition, erosion in the upper section of the flume generating a debris flow may affect the initial velocity at the upper end of a rigid bed section. The numerical simulation in this study described experimental debris flows including generation in the upper section. The channel slope in the erodible bed declined in line with erosion, and the decreased angle reduced the velocity of the debris flows as well as the sediment concentration. The entrainment process may affect the underestimation of the initial velocity of mono-granular debris flows (Figure 4a). Use of alternative entrainment equations may lead to further agreement between calculations and experiments. Influence of Particle Segregation on the Fluidity of Debris Flows As shown in Figures 2 and 4, the behavior of bigranular debris flows could not be modeled with the

146

mean particle diameter, whereas the behavior of monogranular flows could. This may be attributable to particle segregation; the uneven distribution of particles renders it inappropriate to use the mean particle diameter when modeling fluidity. Particle segregation is initially caused by inverse grading (Figure 7), as shown in previous reports on bi-dispersed dry granular flows (Savage and Lun, 1988; Gray and Thornton, 2005; and Goujon et al., 2007) and debris flows (Yamano and Daido, 1985; Takahashi et al., 1992; and Johnson et al., 2012). Small particles concentrate in the lower layer, characterized by a steeper velocity profile (Figure 5b), which suggests that most shear stress is borne by small particles. This is consistent with the observation that the behavior of the bi-granular flows corresponded to that predicted for flows with small particles only, regardless of the mixing ratio, after the flows had developed sufficiently (Figure 2b and c). Similar behavior was noted by Linares-Guerrero et al. (2007) based on numerical simulation using the discrete element method to model bi-dispersed dry granular flow. However, at the start of the debris flows, when particles of different size had not yet been vertically segregated, the fluidity of the bi-granular flows was similar to that of a mono-granular flow containing particles measuring 0.29 cm in diameter (Figure 4b and c). Large particles dispersed within the flow body may have dominated the internal stress environment, but as the flow descended, the flow velocity (including the frontal velocity) changed to that of a flow of smaller particles (Figure 4b and c) as a result of inverse grading within the main body (Figure 7). Thus, debris flow motion cannot be adequately described by numerical simulations featuring particles of uniform size. Representative Grain Size for Multi-Granular Debris Flows Our preliminary attempt to simulate debris flow motion incorporating the effects of particle segregation failed, as shown in Figure 9. The experiments showed accelerated frontal velocity even for the monogranular debris flows (Figure 4a), whereas steady motion was depicted in the calculation (Figure 9a), resulting from different settings of the initial conditions (i.e., a constant flow of debris from the upper end). For the bi-granular debris flows, however, the initial conditions could not account for the differences between the experiments for the flow depth (Figure 2) and the calculations (Figure 9b), and it is interesting that the calculated results looked more plausible based on the constitutive equations (Eqs. 4–8). Our procedure to incorporate particle segregation was similar to that of Takahashi et al. (1992) in

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 139–149

Particle Segregation and Debris Flow Fluidity

terms of evaluating the representative grain size locally. With reference to the constitutive equations (Eqs. 4–8), larger particles are responsible for greater shear stress, which was actually validated for the mono-granular debris flows (Figure 2a). However, in reality for the bigranular debris flows, shear stresses are predominantly supported by smaller particles, as discussed in the previous section, which implies that the local mean diameter is insufficient as the representative particle size to describe the flow characteristics. These findings cannot necessarily be applied to any debris flow. As mentioned regarding the experimental vertical distributions of particle velocities (Figure 5), the vertical profile of velocity is quite different between debris flows on a rigid bed and those on an erodible bed. For a debris flow on an erodible bed, the velocity profile near the bottom is moderate, in contrast to that on a rigid bed, which results in predominant static shear stress due to particle contact, and kinetic characteristics should be governed by particles present in the upper layer, where the velocity profile is steeper. Above all, the representative particle size of multi-granular debris flows should not be determined merely based on actual grain size distributions but should be determined considering the internal stress structure. Note that the use of the actual mean diameter simply incorporating particle segregation may be detrimental, as it could lead to the prediction of slower debris flows than would occur in reality. CONCLUSIONS Experimental mono-granular debris flows consisting of sand particles 0.14 and 0.29 cm in diameter were reproduced by numerical simulation, whereas bigranular debris flows were not well described by mean grain size. Flow depth was lower in the experimental results than in the simulations, which corresponded to those for mono-granular debris flows containing sand particles of 0.14 cm in diameter, regardless of the mixing ratio. The frontal velocity and flux were also lower in the experimental results during the initial stage of flow but reached those of the simulations when using the mean particle diameter over time. These results could be explained by the predominant support of shear stress by smaller sediment particles as the debris flow descends; these smaller particles tended to be concentrated in the lower flow layers as a result of inverse grading, resulting in a velocity profile with a steeper vertical distribution. The behavior of the bi-granular debris flows was assessed by numerical simulations based on a simple model that incorporated the longitudinal transition in mean particle size observed in the experiments, with larger particles at the front of the debris flow and

smaller particles in the flow body. The model failed to account for the experimental debris flows in particular the flow depth. Neither mean diameter nor incorporating particle segregation could not reproduce the behavior of bi-granular debris flows. These results suggest that mean diameter does not necessarily dictate debris flow fluidity and that a representative grain size should be provided, considering the internal stress structure that is responsible for the flow characteristics. ACKNOWLEDGMENTS We sincerely thank Dr. Yuji Hasegawa (Hiroshima University) for his support in the flume tests. This research was partially supported by Grant-in-Aid for Scientific Research 18H03957 (2018) from the Ministry of Education, Culture, Sports, Science and Technology. REFERENCES Armanini, A.; Capart, H.; Fraccarollo, L.; and Larcher, M., 2005, Rheological stratification in experimental free-surface flows of granular-liquid mixtures: Journal Fluid Mechanics, Vol. 532, pp. 269–319, https://doi.org/10.1017/S0022112005004283. Armanini, A.; Larcher, M.; and Fraccarollo, L., 2009, Intermittency of rheological regimes in uniform liquidgranular flows: Physical Review E, Vol. 79, No. 5, 051306, https://doi.org/10.1103/PhysRevE.79.051306. Bagnold, R. A., 1954, Experiments on a gravity free dispersion of large solid spheres in a Newtonian fluid under shear: Proceedings Royal Society London A, Vol. 225, No. 1160, pp. 49–63. Berti, M.; Genevois, R.; Simoni, A.; and Tecca, P. R., 1999, Field observations of a debris flow event in the Dolomites: Geomorphology, Vol. 29, pp. 265–274. Berzi, D. and Jenkins, J. T., 2008, A theoretical analysis of free-surface flows of saturated granular-liquid mixtures: Journal Fluid Mechanics, Vol. 608, pp. 393–410, https://doi.org/10.1017/S0022112008002401. Berzi, D.; Jenkins, J. T.; and Larcher, M., 2010, Debris flows: Recent advances in experiments and modeling: Advances Geophysics, Vol. 52, pp. 103–138, https://doi.org/10.1016/S00652687(10)52002-8. Bouchut, F.; Fernández-Nieto, E. D.; Mangeney, A.; and Narbona-Reina, G., 2016, A two-phase two-layer model for fluidized granular flows with dilatancy effects: Journal Fluid Mechanics, Vol. 801, pp. 166–221, https://doi.org/10.1017/jfm.2016.417. Coe, J. A.; Kinner, D. A.; and Godt, J. W., 2008, Initiation conditions for debris flows generated by runoff at Chalk Cliffs, central Colorado: Geomorphology, Vol. 96, pp. 270–297, https://doi.org/10.1016/j.geomorph.2007.03.017. Egashira, S.; Ashida, K.; and Sasaki, H., 1988, Mechanics of debris flow in open channel: Proceedings Japanese Conference Hydraulics, Japan Society Civil Engineers, Vol. 32, pp. 485–490 (in Japanese with English summary). Egashira, S.; Honda, N.; and Itoh, T., 2001, Experimental study on the entrainment of bed material into debris flow: Physics Chemistry Earth C, Vol. 26, No. 9, pp. 645–650. Egashira, S.; Miyamoto, K.; and Itoh, T., 1997, Constitutive equations of debris flow and their applicability: Proceedings

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 139–149

147

Hotta, Iwata, Suzuki, and Sakai International Conference Debris-Flow Hazards Mitigation 1st, San Francisco, pp. 340–349. Fraccarollo, L. and Capart, H., 2002, Riemann wave description of erosional dam-break flows: Journal Fluid Mechanics, Vol. 461, pp. 183–228, https://doi.org/10.1017/ S0022112002008455. Goujon, C.; Dalloz-Dubrujeaud, B.; and Thomas, N., 2007, Bidisperse granular avalanches on inclined planes: A rich variety of behaviors: European Physical Journal E, Vol. 23, pp. 199–215, https://doi.org/10.1140/epje/i2006-10175-0. Gray, J. M. N. T. and Thornton, A. R., 2005, A theory for particle size segregation in shallow granular freesurface flows: Proceedings Royal Society London A, Vol. 461, pp. 1447–1473. Hill, K. M. and Tan, D. S., 2014, Segregation in dense sheared flows: Gravity, temperature gradients, and stress partitioning: Journal Fluid Mechanics, Vol. 756, pp. 54–88. Hotta, N., 2012, Basal interstitial water pressure in laboratory debris flows over a rigid bed in an open channel: Natural Hazards Earth System Sciences, Vol. 12, pp. 2499–2505, https://doi.org/10.5194/nhess-12-2499-2012. Hotta, N.; Kaneko, T.; Iwata, T.; and Nishimoto, H., 2013, Influence of fine sediment on the fluidity of debris flows: Journal Mountain Science, Vol. 10, pp. 233–238, https://doi.org/10.1007/s11629-013-2522-y. Hotta, N. and Miyamoto, K., 2008, Phase classification of laboratory debris flows over a rigid bed based on the relative flow depth and friction coefficients: International Journal Erosion Control Engineering, Vol. 1, pp. 54–61, https://doi.org/10.13101/ijece.1.54. Hotta, N.; Tsunetaka, H.; and Suzuki, T., 2015, Interaction between topographic conditions and entrainment rate in numerical simulations of debris flow: Journal Mountain Science, Vol. 12, pp. 1383–1394, https://doi.org/10.1007/s11629-0143352-2. Hutter, K.; Svendsen, B.; and Rickenmann, D., 1996, Debris flow modeling: A review: Continuum Mechanics Thermodynamics, Vol. 8, pp. 1–35. Itoh, T.; Egashira, S.; Miyamoto, K.; and Takeuchi, T., 1999, Transition of debris flows over rigid beds to over erodible beds: Proceedings Japanese Conference Hydraulics, Japan Society Civil Engineers, Vol. 43, pp. 635–640(in Japanese with English summary). Iverson, R. M., 1997, The physics of debris flows: Review Geophysics, Vol. 35, No. 3, pp. 245–296, https://doi.org/10.1029/ 97RG00426. Iverson, R. M. and George, D. L., 2014, A depth-averaged debrisflow model that includes the effects of evolving dilatancy. I. Physical basis: Proceedings Royal Society London A, Vol. 470, 20130819, https://doi.org/10.1098/rspa.2013.0819. Iverson, R. M.; Reid, M. E.; Logan, M.; LaHusen, R. G.; Godt, J. W.; and Griswold, J. P., 2015, Positive feedback and momentum growth during debris-flow entrainment of wet bed sediment: Nature Geoscience, Vol. 4, pp. 116–121, https://doi.org/10.1038/ngeo1040. Iwata, T.; Hotta, N.; and Suzuki, T., 2013, Influence of particle size segregation in multi granular debris flow on the fluidity: Journal Japan Society Erosion Control Engineering, Vol. 66, No. 3, pp. 13–23 (in Japanese with English summary). Johnson, C. G.; Kokelaar, B. P.; Iverson, R. M.; Logan, M.; LaHusen, R. G.; and Gray, J. M. N. T., 2012, Grain-size segregation and levee formation in geophysical mass flows: Journal Geophysical Research, Vol. 117, F01032, https://doi.org/10.1029/2011JF002185.

148

Kaitna, R.; Palucis, M. C.; Yohannes, B.; Hill, K. M.; and Dietrich, W. E., 2016, Effects of coarse grain size distribution and fine particle content on pore fluid pressure and shear behavior in experimental debris flows: Journal Geophysical Research: Earth Surface, Vol. 121, pp. 415–441, https://doi.org/10.1002/2015JF003725. Lanzoni, S.; Gregoretti, C.; and Stancanelli, L. M., 2017, Coarse-grained debris flow dynamics on erodible beds: Journal Geophysical Research: Earth Surface, Vol. 122, pp. 592–614, https://doi.org/10.1002/2016JF004046. Li, J.; Cao, Z.; Hu, K.; Pender, G.; and Liu, Q., 2018, A depthaveraged two-phase model for debris flows over erodible beds: Earth Surface Processes Landforms, Vol. 43, pp. 817–839, https://doi.org/10.1002/esp.4283. Linares-Guerrero, E.; Goujon, C.; and Zenit, R., 2007, Increased mobility of bidisperse granular avalanches: Journal Fluid Mechanics, Vol. 593, pp. 475–504, doi:10.1017/S0022112007008932. Major, J. and Pierson, T., 1992, Debris flow rheology: Experimental analysis of fine grained slurries: Water Resources Research, Vol. 28, No. 3, pp. 841–857. Medina, V.; Hürlimann, M.; and Bateman, A., 2008, Application of FLATModel, a 2D finite volume code, to debris flows in the northeastern part of the Iberian Peninsula: Landslides, Vol. 5, pp. 127–142, https://doi.org/10.1007/s10346-007-0102-3. Miyamoto, K., 1985, Study on the Grain Flows in Newtonian Fluid: Ph.D. Thesis, Ritsumeikan University, 155 p. (in Japanese). Miyamoto, K. and Itoh, T., 2002, Numerical simulation method of debris flow introducing the erosion rate equation: Journal Japan Society Erosion Control Engineering, Vol. 55, No. 2, pp. 24–35 (in Japanese with English summary). Nakagawa, H. and Takahashi, T., 1997, Estimation of a debris flow hydrograph and hazard area: Proceedings International Conference Debris-Flow Hazards Mitigation 1st, San Francisco, pp. 64–73. Nishiguchi, Y.; Uchida, T.; Takezawa, N.; Ishizuka, T.; and Mizuyama, T., 2012, Runout characteristics and grain size distribution of large-scale debris flows triggered by deep catastrophic landslides: International Journal Erosion Control Engineering, Vol. 5, pp. 16–26, https://doi.org/10.13101/ ijece.5.16. Osti, R. and Egashira, S., 2008, Method to improve the mitigative effectiveness of a series of check dams against debris flows: Hydrological Processes, Vol. 22, pp. 4986–4996, https://doi.org/10.1002/hyp.7118. Sakai, Y.; Hotta, N.; Kaneko, T., and Iwata, T., 2019, Effects of grain-size composition on flow resistance of debris flows: Behavior of fine sediment: Journal Hydraulic Engineering, Vol. 145, No. 5, https://doi.org/10.1061/(ASCE)HY.19437900.0001586. Savage, S. B. and Lun, C. K. K., 1988, Particle size segregation in inclined chute flow of dry cohesionless granular solids: Journal Fluid Mechanics, Vol. 189, pp. 311–335. Stock, J. D. and Dietrich, W. E., 2006, Erosion of steepland valleys by debris flows: Geological Society America Bulletin, Vol. 118, pp. 1125–1148, https://doi.org/10.1130/B25902.1. Suwa, H.; Okano, K.; and Kanno, T., 2009, Behavior of debris flows monitored on test slopes of Kamikamihorizawa Creek, Mount Yakedake, Japan: International Journal Erosion Control Engineering, Vol. 2, pp. 33–45, https://doi.org/10.13101/ijece.2.33. Suzuki, T., 2007, Study on Flow Mechanism of Debris Flows on a Rough Fixed Bed: Ph.D. Thesis, The University of Tokyo, 150 p. (in Japanese).

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 139–149

Particle Segregation and Debris Flow Fluidity Takahashi, T., 1978, Mechanical characteristics of debris flow: Journal Hydraulic Division, ASCE, Vol. 104, pp. 1153–1169. Takahashi, T., 1991, Debris Flow: Balkema, Rotterdam, The Netherlands. 165 p. Takahashi, T.; Nakagawa, H.; Harada, T.; and Yamashiki, Y., 1992, Routing debris flows with particle segregation: Journal Hydraulic Engineering, Vol. 118, No. 11, pp. 1490– 1507, https://doi.org/10.1061/(ASCE)0733-9429(1992)118: 11(1490). Xia, C.; Li, J.; Cao, Z.; Liu, Q.; and Hu, K., 2018, A quasi single-phase model for debris flows and its compar-

ison with a two-phase model: Journal Mountain Science, Vol. 15, pp. 1071–1089, https://doi.org/10.1007/s11629-0184886-5. Yamano, K. and Daido, A., 1985, The mechanism of granular ﬂow of mixed diameter composed two diameters: Journal Japan Society Civil Engineers, Vol. 357, pp. 25–34 (in Japanese with English summary). Yohannes, B.; Hsu, L.; Dietrich, W. E.; and Hill, K. M., 2012, Boundary stresses due to impacts from dry granular ﬂows: Journal Geophysical Research, Vol. 117, F02027, https://doi.org/10.1029/2011JF002150.

Environmental & Engineering Geoscience, Vol. XXVII, No. 1, February 2021, pp. 139–149

149

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

EDITORS

Eric Peterson Department of Geography, Geology, and the Environment Illinois State University Normal, IL 61790 309-438-5669 ewpeter@ilstu.edu

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

Environmental & Engineering Geoscience February 2021 VOLUME XXVII, NUMBER 1 Special Issue on Debris Flows, Part 1 Paul M. Santi and Lauren N. Schaefer, Guest Editors

Volume XXVII, Number 1, February 2021

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

ENVIRONMENTAL & ENGINEERING GEOSCIENCE