<<

Prehistoric and modern debris flows in semi-arid watersheds: Implications for hazard assessments in a changing

Item Type text; Electronic Dissertation

Authors Youberg, Ann M.

Publisher The University of .

Rights Copyright © is held by the author. Digital access to this material is made possible by the University Libraries, . Further transmission, reproduction or presentation (such as public display or performance) of protected items is prohibited except with permission of the author.

Download date 10/10/2021 22:25:06

Link to Item http://hdl.handle.net/10150/312570

PREHISTORIC AND MODERN DEBRIS FLOWS IN SEMI-ARID WATERSHEDS: IMPLICATIONS FOR HAZARD ASSESSMENTS IN A CHANGING CLIMATE

By

Ann M. Youberg

______

A dissertation Submitted to the Faculty of the

DEPARTMENT OF AND WATER RESOURCES

In Partial Fulfillment of the Requirements

For the Degree of

DOCTOR OF PHILOSOPHY

with a major in Hydrology

In the Graduate College

UNIVERSITY OF ARIZONA

2013

THE UNIVERSITY OF ARIZONA GRADUATE COLLEGE

As members of the Dissertation Committee, we certify that we have read the dissertation prepared by Ann Youberg,

Entitled Prehistoric and Modern Debris Flows in Semi-Arid Watersheds: Implications for Hazard Assessments in a Changing Climate

and recommend that it be accepted as fulfilling the dissertation requirement for the Degree of Doctor of Philosophy.

th ______Date: November 25 , 2013 Victor R. Baker

th ______Date: November 25 , 2013 Robert H. Webb

th ______Date: November 25 , 2013 Peter A. Troch

th ______Date: November 25 , 2013 Jon D. Pelletier

th ______Date: November 25 , 2013 Philip A. Pearthree

Final approval and acceptance of this dissertation is contingent upon the candidate’s submission of the final copies of the dissertation to the Graduate College.

I hereby certify that I have read this dissertation prepared under my direction and recommend that it be accepted as fulfilling the dissertation requirement.

______Date: November 25th, 2013 Dissertation Director: Victor R. Baker

______Date: November 25th, 2013 Dissertation Director: Robert H. Webb

2

STATEMENT BY AUTHOR

This dissertation has been submitted in partial fulfillment of the requirements for an advanced degree at the University of Arizona and is deposited in the University Library to be made available to borrowers under rules of the Library.

Brief quotations from this dissertation are allowable without special permission, provided that an accurate acknowledgement of the source is made. Requests for permission for extended quotation from or reproduction of this manuscript in whole or in part may be granted by the head of the major department or the Dean of the Graduate College when in his or her judgment the proposed use of the material is in the interests of scholarship. In all other instances, however, permission must be obtained from the author.

SIGNED: Ann M Youberg

3

ACKNOWLEDGMENTS As with all projects this large and complex, there are many people to thank. First, I would like to thank all of my committee and especially my co-advisors, Vic Baker and Bob Webb, for their guidance, insights and mentoring. Bob Webb and Phil Pearthree did their best to keep me on track. I particularly want to thank Phil, who has been my Section Chief and mentor for the past 14 years at the Arizona Geological Survey. Phil planted the idea of this dissertation when we were surveying the post-Aspen Fire flood damage in Romero Canyon. Everyone at the Arizona Geological Survey has supported or encouraged me in some way. I particularly want to thank Lee Allison, director of the AZGS, for his continued support throughout this process, and for his ongoing encouragement and support for exploring new research opportunities that haven’t traditionally been part of the Survey’s program. Joe Cook is a most excellent field partner and has taught me how to better tell the geomorphic story. Janel Day, Steve Richard, and Adrian Sonnenschein helped me with GIS and data problems, and Jeri Young and Brian Gootee provided encouragement and perspective. I have been very fortunate to have worked with, and be mentored by, many wonderful people not associated with the University or the AZGS. I particularly want thank Sue Cannon for her mentoring, and also for her financial support. Sue’s knowledge of post- debris flows is unsurpassed and I appreciate her sharing her knowledge with me. Deb Martin, John Moody, Jason Kean, and Dennis Staley have provided countless hours of guidance and feedback. I look forward to continued collaborations. Many people US Forest Service have contributed, supported or encouraged this work in some way. In particular I want to thank Bob Lefevre and Randal Smith, both now retired, for their encouragement and for providing early opportunities for me to learn about post-wildfire erosion. Jennifer Ruyle has provided much appreciated support for this work. Salak Shafiquillah, Marc Stamer, Michele Girard, Dan Neary, Karen Koestner, Kyle Wright, Kit MacDonald, Polly Haessig, Chris Stewart, Jalyn Cummings, Adam Springer, and Peter Koestner all helped me in some way to learn about assessing the impacts from and what the important questions are. Erica Bigio has been a good friend, an excellent field partner and a collaborator throughout this process. I value her perspective and appreciate her help in solving problems. Steve DeLong provided guidance, encouragement and, by example, taught me how to approach problems. Finally, I particularly want to thank my husband, Bob Czaja, who was not keen on the idea of me pursuing a PhD but nevertheless supported me throughout. Thank you.

4

DEDICATION I dedicate this dissertation to Dave Youberg, Bob Czaja, and Francis Sutherland.

In memory of my mother Kathryn Margaret Myers Youberg. March 3, 1931 – February 13, 2010

5

TABLE OF CONTENTS

ABSTRACT ...... 10

CHAPTER 1: INTRODUCTION ...... 14

1.1 Statement of problem ...... 14 1.2 Background ...... 17 1.2.1 Debris flows in Arizona ...... 21 1.2.2 Post-wildfire debris flows ...... 24

CHAPTER 2: THE PRESENT STUDY ...... 29

2.1 Appendix A: Increased debris-flow activity and magnitude during the Pleistocene- Holocene climatic transition, , Pima County, Arizona; implications for modern debris-flow hazards...... 29 2.1.2 Abstract ...... 30 2.2 Appendix B: Rainfall intensity-duration thresholds and the predictive strengths of three post-wildfire debris-flow models for use in Arizona...... 31 2.2.1 Abstract ...... 31 2.3 Appendix C: Basin physiographic and soil burn severity factors influencing the generation of post-wildfire debris flows...... 32 2.3.1 Abstract ...... 33

CHAPTER 3: CONCLUSIONS ...... 34

3.1 Summary ...... 34 3.2 Implications for future post-wildfire debris-flow hazard assessments ...... 36 3.1 Rainfall assessments ...... 37 3.2 Basin assessments ...... 39 3.3 Conclusions ...... 43

APPENDIX A: INCREASED DEBRIS-FLOW ACTIVITY AT THE PLEISTOCENE- HOLOCENE CLIMATIC TRANSITION, SANTA CATALINA MOUNTINS, PIMA COUNTY, ARIZONA; IMPLICATIONS FOR MODERN DEBRIS-FLOW HAZARDS ...... 45

A.1 Abstract ...... 46 A.2 Introduction ...... 48 A.3 Study Area ...... 52 A.4 Methods ...... 53 6

A.4.1 Geologic mapping of debris-flow deposits ...... 53 A.4.2 Cosmogenic age dating of debris-flow deposits ...... 55 A.4.3 Debris-flow modeling ...... 58 A.5 Results ...... 59 A.5.1 Geologic mapping ...... 59 A.5.1.1 Finger Rock Canyon ...... 59 A.5.1.2 Pima Canyon ...... 60 A.5.2 Cosmogenic 10Be ages on debris-flow surfaces ...... 61 A.5.2.1 Finger Rock Canyon ...... 62 A.5.2.2 Pima Canyon ...... 63 A.5.2.3 Comparison of exposure ages of units in Finger Rock and Pima Canyons...... 65 A.5.3 Debris-flow modeling ...... 65 A.5.3.1 LAHARZ model results ...... 66 A.5.3.2 Canyon in- storage analysis ...... 66 A.6 Discussion ...... 67 A.6.1 Ages of debris-flow deposition ...... 68 A.6.2 Implications from debris-flow modeling ...... 71 A.6.3. Climatic conditions at the Pleistocene-Holocene transition ...... 72 A.7 Conclusions ...... 76 A.9 References ...... 79 A.10 Figure and Table Captions ...... 85 A.11 Figures ...... 91

APPENDIX B: RAINFALL INTENSITY-DURATION THRESHOLDS AND THE PREDICTIVE STRENGTHS OF THREE POST-WILDFIRE DEBRIS FLOW MODELS FOR USE IN ARIZONA ...... 100

B.1 Abstract ...... 101 B.2. Introduction ...... 103 B.2.1 Development of post-wildfire debris-flow models ...... 106 B.2.2 Development of rainfall intensity-duration threshold models ...... 107 B.3. Settings ...... 108 B.3.1 Colorado Plateau ...... 108 B.3.1.1 2010 , , Flagstaff, AZ ...... 110

7

B.3.1.2 2011 Wallow Fire, east-central Arizona and western New Mexico ...... 111 B.3.2 Central Highlands Transition Zone ...... 112 B.3.2.1 Gladiator Fire, , central Arizona ...... 113 B.3.3 Basin and Range ...... 114 B.3.3.1 2011 Horseshoe 2 Fire, , southeastern Arizona .. 116 B.3.3.2 Monument Fire, in southeastern Arizona ...... 117 B.4. Methods ...... 117 B.4.1 Data collection ...... 118 B.4.2 Data analysis ...... 118 B.4.2.1 Rainfall analysis ...... 118 B.4.2.2 Post-wildfire debris-flow modeling ...... 119 B.4.2.3 Receiver operating characteristic (ROC) analysis ...... 120 B.5. Results ...... 123 B.5.1 Basin responses ...... 123 B.5.1.1 Colorado Plateau ...... 124 B.5.1.2 Central Highlands Transition Zone ...... 125 B.5.1.3 Basin and Range ...... 125 B.5.2 Rainfall characteristics and objectively defined intensity-duration thresholds.. 126 B.5.3 Post-fire debris-flow probability and volume models assessments ...... 127 B.5.3.1 Colorado Plateau ...... 128 B.5.3.2 Central Highlands Transition Zone ...... 128 B.5.3.3 Basin and Range ...... 128 B.6 Discussion ...... 129 B.6.1 Rainfall intensity-duration thresholds ...... 129 B.6.1.1 Practical guidelines for using the OBJ_ID threshold models ...... 133 B.6.2 Post-fire debris-flow probability models ...... 133 B.6.2.1 Colorado Plateau ...... 133 B.6.2.2 Central Highlands Transition Zone ...... 134 B.6.2.3 Basin and Range ...... 135 B.6.2.4 Practical guidelines for using the models ...... 136 Recommendations the Colorado Plateau: ...... 138 Recommendations for the Transition Zone: ...... 138 Recommendations for the Basin and Range: ...... 139

8

B.6.3 Debris-flow volume model ...... 139 B.7 Guidelines for future studies ...... 140 B.7.1 Rainfall analyses ...... 141 B.7.2 Debris-flow probability models ...... 142 B.7.2 Debris-flow volume models ...... 143 B.8 Conclusions ...... 144 B.9 References ...... 146 B.10 Figure and Table Captions ...... 151 B.11 Figures ...... 158

APPENIDX C: BASIN PHYSIOGRAPHIC AND SOIL BURN SEVERITY FACTORS INFLUENCING THE GENERATION OF POST-WILDFIRE DEBRIS FLOWS ...... 168

C.1 Abstract ...... 169 C.3 Introduction ...... 171 C.3.1 Burn severity measures ...... 174 C.3.2 Debris-flow processes ...... 176 C.4 Study Area ...... 179 C.5 Methods ...... 181 C.6 Results and Discussion ...... 186 C.6.1 Basin morphometric factors ...... 186 C.6.2 Slope-area plots ...... 186 C.6.3 Slope intervals ...... 188 C.6.3 Soil burn severity ...... 189 C.6.6 Slopes and erosional thresholds ...... 193 C.8 Conclusions ...... 195 C.9 References ...... 198 C.10 Figure and Table Captions ...... 204 C.11 Figures ...... 209

REFERENCES ...... 220

9

ABSTRACT

Until recently, debris flows were not considered a significant hazard in Arizona. Historic and modern debris flows were generally confined to remote, mountainous areas and to the Grand

Canyon. In 2006, an extremely rare (RI ≈ 1000 yrs), five-day resulted in approximately

1000 hillslope failures and numerous debris flows throughout southeastern Arizona, impacting

infrastructure and housing. The number and extent of debris flows from this event was

unexpected, and it dramatically illustrated the potential for sizable debris flows to exit canyons and

adversely impact properties near canyon mouths. Moreover, during the past few decades Arizona

has seen a dramatic increase in the size and severity of wildfires, a trend seen across the western

U.S., and a concurrent increase in development of the wildland-urban interface. Reflecting the increase of wildfire severity, hydrologic and erosional responses from burned areas are larger, which places people and their property at risk to potential post-wildfire flooding and debris flows. Climate projections for the Southwestern U.S. call for increased temperatures, decreased overall precipitation but with the potential for increased extreme winter precipitation events, and

increased wildfire activity. Recent trends, therefore, can be expected to continue, illustrating the

need to assess the magnitude of modern debris-flow hazards, and to develop methods appropriate for Arizona’s varied physiographic regions to identify and evaluate these hazards. Here, we present three distinct but related debris-flow studies to address these needs.

In the first study, we assess prehistoric to modern debris-flow deposits on two alluvial fans to place debris-flow hazards in the context of both the modern environment and the last major period of . First, we use a combination of surficial geologic mapping and

10Be exposure dating to produce a debris-flow history. The largest debris flows, of latest

10

Pleistocene to early Holocene age, cover much of the apices of alluvial fans formed at the

mouths of these canyons and extend up to 3 km downslope. The 10Be ages in this study generally

corroborate relative-ages from the geologic mapping, but reveal considerable scatter for surfaces

believed to be of uniform age indicating the dual possibilities of inheritance from previous

cosmic-ray exposure, as well as the potential for composite deposits derived from numerous

debris flows. We then use a combination of modeling to assess probable magnitudes of the older

debris flows, and GIS analysis to estimate the potential volumes of material available for

transport to evaluate possible initiation mechanisms. In-channel and floodplain storage within the canyons is not sufficient to generate volumes likely needed to produce the larger late Pleistocene to early Holocene debris-flow deposits. These older debris flows were likely initiated by larger landslide or other mass movement failures, although evidence of these features was not observed. Recent saturation-induced debris flows from extreme precipitation and runoff-induced, post-wildfire debris flows indicate that most modern debris flows are likely to mainly entrain channel material, thus limiting potential debris-flow magnitudes and hazard zones.

In the second and third studies we shift our focus to post-wildfire debris flows. Teams conducting rapid assessments of post-wildfire geologic hazards need models to quickly identify areas at risks for post-fire debris flows, and forecast centers and emergency response agencies need rainfall intensity-duration thresholds that identify rainfall conditions likely to generate post- fire debris flows. In the second study we address these needs specifically for Arizona. We assess the predictive strengths of existing probability and volume models in three different physiographic regions of Arizona, and define a new rainfall threshold valid for Arizona.

Specifically, we test the volume model and three probability models using data from 80 recently

11

burned basins in the Colorado Plateau, the Central Highlands Transition Zone, and the Basin and

Range of Arizona. We show that all of the models have adequate predictive strength throughout

most of the state, with the exception of lower gradient basins on the Colorado Plateau where all

models performed poorly. We also show that the debris-flow volume model over-predicts in all of our study areas. Our analysis shows that the choice of a model for a hazard assessment depends strongly on the area where it will be used. We also developed objectively defined rainfall intensity-duration thresholds to identify rainfall conditions likely to generate post- wildfire debris flows in Arizona. The objectively defined I10 and I15 thresholds of 52 and 42 mm

h-1, respectively, have the strongest predictive strengths of the five threshold models developed,

although all of the threshold models performed well. Results from this study provide guidelines

and considerations for selecting an appropriate debris-flow model, and rainfall threshold values

for issuing forecasts and warnings for Arizona.

The third study was motivated by the second study. Here, we use the large dataset of

burned basin flow responses developed during the second study to explore various basin

physiographic and soil burn severity factors to identify patterns and criteria that can be used to

discriminate between potential non-debris-flow (nD) and debris-flow (D) producing basins.

Findings from this study show that a metric of percent basins area with high soil burn severity on

slopes ≥30° provides a stronger discrimination between nD and D basins than do basin metrics,

such as mean basin gradient or relief, which weakly discriminate between basin responses. Mean

basin elevation was also found to discriminate nD from D basins and likely is a proxy for forest

type and density, which relates to soil thickness and root density. Mean basin elevation may also

be a proxy for the measure of efficient energy and mass transfer (EEMT), an erosional term that

12

combines temperature and precipitation and correlates with elevation, bedrock erosivity, soil

production, and above-ground biomass. All of these factors influence the magnitude of post-

disturbance erosion. Finally, we found that post-fire channel heads mapped for the 2011

Monument Fire formed at essentially the same slope range (~30-400) as 2006 saturation induced

debris flows, a result that is both useful for identifying potential post-fire debris flow hazards and may provide information to guide the future development of physically based models.

13

CHAPTER 1: INTRODUCTION

1.1 Statement of problem

Climate projections for the Southwestern US call for increased temperatures, decreased

overall precipitation but with the potential for increased extreme winter precipitation events, and

increased wildfire activity (Garfin et al., 2013). Indeed, wildfire activity has already increased

dramatically across the western U.S., resulting in an earlier start and a longer fire season with

larger, hotter, and longer-burning fires (Westerling et al., 2006). In the Southwestern U.S.,

extreme winter drought from warming temperatures and increased vapor pressure deficits have

resulted in an increase in total area burned (Williams et al., 2010; Williams et al., 2013), and an

increase in patch size of high burn severity areas (Falk, 2013). These factors, in turn, are

generating larger hydrologic and erosional responses from burned areas, which places people and

their property to potential post-wildfire flooding and debris flows.

Concurrent with these trends is an increase of urban development on alluvial fans near mountain fronts, as well as expansion of residential areas into the wildland-urban interface

(WUI). Alluvial fans in the Southwestern U.S. are one of the dominant landscape features that

sustain infrastructure and housing. Although debris flows are major depositional events that lead

to long-term aggradation of alluvial fans (Melton, 1965b; McDonald et al., 2003), flood control

measures in this region typically focus on channel-bank stabilization, ignoring the potential for

channel aggradation and subsequent overbank deposition during debris flows (Webb et al.,

2008b). Housing and infrastructure within the WUI are at even greater risks to the aftermaths of wildfires (Stein et al., 2013). A study by the Arizona Republic showed development in the WUI increased 91% between 1980 and 2000 in Arizona, with over 230,000 homes and structures built

14

in, or adjacent to, the WUI since 1990 (Pitzl et al., 2013). Development on alluvial fans and within the WUI exposes housing and infrastructure to numerous geologic and hydrologic hazards, including flooding and debris flows.

In the arid and semiarid southwestern United States, historical debris flows commonly occur following extreme precipitation and, recently, more commonly from watersheds burned by wildfires (Wohl and Pearthree, 1991; Webb et al., 2008b; Parise and Cannon, 2012). Historical occurrences of debris flows (<150 years) have been documented in all physiographic provinces in Arizona, including the Grand Canyon and isolated peaks on the Colorado Plateau (Melton,

1965b; Melis et al., 1994; Griffiths et al., 1996; Jenkins, 2007; Webb et al., 2008a), the

Mogollon Rim and the Transition Zone (Pearthree and Youberg, 2004; Youberg, 2008), and numerous mountain ranges within the Basin and Range Province of central and southern Arizona

(Wohl and Pearthree, 1991; Pearthree and Youberg, 2006; Webb et al., 2008b). Documentation of historical debris flows reveal that, while debris flows do occur across Arizona, they are limited in number and are typically confined to steep watersheds in mountainous terrain and sparsely populated areas. As an exception to this general rule in the modern environment, wildfire-related debris flows have impacted both infrastructure and private property (Pearthree and Youberg, 2004).

Numerous Pleistocene and Holocene debris-flow deposits, on the other hand, provide ample geologic evidence of prehistoric debris-flow activity in most mountain ranges across

Arizona (Jenkins, 2007; Johnson et al., 2011; Ferguson et al., 2012; Pearthree et al., 2012;

Youberg and Pearthree, 2012; Young et al., 2012). Extensive and large caliber debris-flow deposits found on alluvial fans throughout the state record periods of aggradation restricted to the

15

wetter climate of the Pleistocene and early Holocene (older than about 8,000 years), and attest to

the primary importance of debris flows in constructing fans during that time. Few Pleistocene-

Holocene debris-flow deposits have been studied, but detailed geologic mapping of debris-flow deposits on fans along the front range of the Santa Catalina Mountains show mid-Holocene to modern debris-flow deposits are smaller and more limited in extent than Pleistocene to early

Holocene deposits (Youberg et al., 2008).

Continued development on alluvial fans and encroachment into the WUI places more people at risks to debris-flow hazards. Understanding modern debris-flow processes, areas of past debris flow deposition, and past debris-flow magnitude and frequency are all needed when placing debris-flow hazards into a spatial and temporal framework (Youberg et al., 2008).

Moreover, assessment methods that are appropriate for Arizona’s varied physiographic regions are needed to quantify risks and identify areas for mitigation measures. Finally, changes in the modern climate may reduce the potential for debris flows from extreme precipitation while enhancing the possibility of wildfire-related debris flows. Research questions surrounding these issues include:

(1) What were debris-flow magnitudes and frequencies during the late Pleistocene to mid-

Holocene period of optimal environmental conditions for large-scale debris flow processes,

and how do these patterns inform assessments of modern debris-flow hazards under the

context of current climate changes?

(2) How well do existing predictive models of post-wildfire debris-flow hazards perform in

Arizona’s varied physiographic provinces, and under what rainfall conditions are post-

wildfire debris flows generated in Arizona?

16

(3) Are there specific basin morphometric and/or burn severity factors that can be used to

identify basins, either before or after wildfires, likely to generate post-wildfire debris flows?

1.2 Background

Debris flows are unsteady, non-uniform, very poorly sorted, sediment-rich slurries that are generated by saturated soil-induced hillslope failures (Costa and Williams, 1984; Iverson,

2003a), or by runoff-induced mechanisms on disturbed or sparsely vegetated slopes (Coe et al.,

2008; McCoy et al., 2010; Kean et al., 2011; Kean et al., 2013). Debris flows differ from flood flows and hyperconcentrated flows both in the amount of sediment they contain, more than 60% sediment by volume (Pierson and Costa, 1987), and in flow behavior. Flood flows typically contain less than 40% sediment by volume, are Newtonian (Pierson and Costa, 1987), and transport clay, silt and sometimes sand as suspended sediment and gravel, generally, as bedload.

Hyperconcentrated flows have around 40-60% sediment by volume and have sufficient interaction between grains to keep sediment in suspension as long as flow velocities are maintained (Pierson, 2005b). Deposits from both flood and hyperconcentrated flows exhibit some degree of sorting by grain size (Pierson, 2005b).

Debris flows are grain-fluid mixtures that have unsteady flow characteristics which, within a single flow, can range from a solid plug to a very fluid flow as composition, pore pressures and grain-to-grain interactions evolve (Iverson, 2003a). This interaction between the fluid and solid phases is what distinguishes debris flows from other flows (Iverson et al., 1997).

The fluid matrix phase of a debris flow is composed of clay, silt, and sand in suspension and is driven by high pore pressure. The solid-particle phase is composed of coarse clasts that interact by frictional and gravitational forces. The solid and fluid phases maintain flow by transferring

17

momentum both within and between each phase simultaneously (Iverson, 1997). These

interactions prevent particles from settling even at low velocities. Thus debris-flow deposits exhibit minimal sorting and typically have distinct morphological characteristics such as levees and snouts (Iverson and Vallance, 2001; Pierson, 2005b). The lack of sorting and other depositional characteristics are used to identify and distinguish debris-flow deposits during geologic mapping of debris-flow deposits. Accurately identify debris-flow deposits, however, can sometimes be quite difficult due to the varied and changing flow properties within a single debris flow, and because most debris flows are followed by a flood that often reworks previously deposited material (Melis et al., 1997).

Debris flows can be initiated by unusually prolonged or intense precipitation falling onto saturated hillslopes (Cannon and Ellen, 1985; Webb et al., 2008b; Montgomery et al., 2009), or by shorter-duration falling onto disturbed or sparsely vegetated hillslopes (Cannon, 2001;

Cannon et al., 2008; Coe et al., 2008). Under the first condition, hillslope soils become saturated and pore pressures increase until shear strengths decrease to the critical point of failure (Costa,

1984; Iverson et al., 1997). Initiation occurs for these debris flows by slumping or rotational failures, or by shallow translational failures of thin soil over impervious surfaces, such as bedrock, that liquefies and rapidly mobilizes into viscous slurries through liquefaction or dilatancy (Costa, 1984; Iverson et al., 1997; Iverson et al., 2010). The initiation points for these saturation-induced debris flows can clearly be identified by the landslide scarps remaining on the hillslopes.

Unlike saturation-induced debris flows, runoff-induced debris flows typically do not have a clear initiation point or failure scarp. This makes it more difficult to pinpoint where and how a

18

debris flow initiated. Runoff-induced debris flows can occur in sparsely vegetated, high-

sediment yield basins (Griffiths et al., 2004; Coe et al., 2008; Webb et al., 2008a; McCoy et al.,

2010), or after disturbances such as wildfires when high-frequency, low-magnitude storms generate very high runoff volumes due to the removal of vegetation and other fire-induced hydrologic changes (Wohl and Pearthree, 1991; Cannon, 2001; Santi et al., 2008; Parise and

Cannon, 2012). Initiation of runoff-induced debris flows may occur through a variety of processes, including the fire hose effect in which runoff entrains colluvial material below cliffs

(Johnson and Rodine, 1984; Griffiths et al., 2004; Godt and Coe, 2007), progressive sediment entrainment of channel material by surface flow (Wohl and Pearthree, 1991; Cannon et al., 2001;

Gabet and Bookter, 2008; Santi et al., 2008; McCoy et al., 2012), mass failure of channel bed or bank materials (Takahashi, 2007), or by centimeter-scale soil slips (Gabet, 2003).

Recent research has shown that runoff-induced debris flows probably most commonly initiate through a series of erosional and depositional processes. Pulses of sediment-laden flows originating as hillslope rills are transported to a channel or concentrated flow area, deposited at an in-channel step or subtle change in and de-watered, progressively building up until the deposit becomes destabilized, at which point a subsequent pulse of sediment breaches the deposit and initiates a debris flow (Kean et al., 2013). This process, modeled as a sediment capacitor, has been documented at the USGS Chalk Cliffs Observatory in Colorado and in basins burned by wildfires in southern California (Kean et al., 2013).

Both storm and site characteristics influence the occurrence of debris flows. Thresholds describing rainfall intensity-duration (ID) relationships provide a means to assess triggering rainfall and debris-flow characteristics across regions. Rainfall intensity-duration threshold

19

curves been developed developed worldwide (Caine, 1980; Guzzetti et al., 2008), regionally (e.g.

Fiorillo and Wilson, 2004; Frattini et al., 2009), and across the western U. S. for saturation-

induced debris flows and landslides (Cannon and Ellen, 1985; Godt et al., 2006; Webb et al.,

2008b; Montgomery et al., 2009; Baum and Godt, 2010), and for runoff-induced debris flows

(Godt and Coe, 2007; Coe et al., 2008), some of which were wildfire related (Cannon et al.,

2008). In general, these rainfall intensity-duration threshold curves show longer durations and lower intensities for saturation-induced debris flows, and shorter duration and high intensities for runoff-induced debris flows. In addition to the characteristics of triggering rainfall, site characteristics that may increase the likelihood of debris-flow initiation include steeper slopes and exposed bedrock, which increases runoff and flow velocity, (Giraud, 2005), high antecedent moisture conditions in seasonally wet/dry areas (Anderson and Sitar, 1995; Wieczorek and

Glade, 2005; Baum and Godt, 2010), subsurface flow through jointing, fractures and fautls

(Montgomery et al., 2009), and substrate clay mineralogy and major-cation chemistry (Melis et al., 1994; Webb et al., 2008a). In areas recently burned by wildfires, consumption of vegetation, soil organics and fine roots results in soils and hydrologic changes that enhance runoff volumes and velocities, reduce erosivity thresholds and increase the likelihood of erosion and debris-flow initiation (Moody and Martin, 2001; Moody et al., 2008; Parsons et al., 2010; Moody and Ebel,

2013; Moody et al., 2013; Nyman et al., 2013). In areas where trees are killed by logging, fires or pests, root strength decreases with time, which can also increase the likelihood of slope failure

(Gerber and Roering, 2003).

A debris-flow hazard is defined by a combination of the probability of the occurrence and the potential magnitude of a debris flow (Jakob, 2005a). Magnitude can be expressed as debris-

20

flow volume, peak discharge, runout (Hungr, 1995; Rickenmann, 1999, 2005), or area inundated

(O'Brien et al., 1993; Berti and Simoni, 2007; Griswold and Iverson, 2008) and is directly related to recurrence intervals and sediment availability (Bovis and Jakob, 1999; Glade, 2005; Hungr et

al., 2008). The prediction component of a debris-flow hazard assessment must be based on some

type of model, physical or empirical, to provide relevant and meaningful information (Wilcock

and Iverson, 2003). Physically based models use classical laws of , such as conservation

of mass, momentum, and energy, to succinctly describe processes and simplified concepts, and

to explore connections between processes, rates, thresholds and resulting landforms (Bras et al.,

2003; Iverson, 2003b; Wilcock and Iverson, 2003). Empirical models are developed to describe,

or predict, a process using historical datasets, observations and statistical analyses. While

physically based models can be applied across broad regions, empirical models are only valid for

areas in which they were developed. Both types of models have been applied during debris-flow

hazard assessments.

1.2.1 Debris flows in Arizona

Although there is ample evidence of past debris flow activity across Arizona, there are

only a few published studies looking specifically at debris flows. Certainly the most studied

debris flows in Arizona are found in the Grand Canyon. These studies have addressed ages of

debris flows (Bowers et al., 1997; Cerling et al., 1999), triggering rain events, initiation,

frequency, and magnitude (Webb et al., 1989; Melis et al., 1994; Griffiths et al., 1996, 1997,

2004), debris flow behavior (Melis et al., 1997) and effects of debris flows on the Colorado

River (Melis et al., 1993; Hereford et al., 1996; Hereford et al., 1998; Webb et al., 1999; Hanks

and Webb, 2006).

21

The oldest documented debris flow in the Grand Canyon has been dated to 5.4 ka

(thousand years ago) (Melis et al., 1994). Webb and others (1999) looked at multiple debris-flow

deposits from one particular canyon, Prospect Canyon, as it exits into the Colorado River and

forms Lava Falls Rapid. Results indicate the oldest set of debris flows were deposited

approximately 3 ka, based on cosmogenic 3He (Webb et al., 1999). Eight other inset debris-flow

deposits were found to range from approximately 2.2 ka to modern, with estimated return

intervals of 5 to 20 years for small debris flows, and 102-103 years for the largest debris flows

(Webb et al., 1999). In another study, debris flows tributaries throughout Grand Canyon were

found to have frequency intervals ranging from 10-15 years in some eastern tributaries to less

than one per century in other tributaries, with an average of 30-50 years throughout the canyon

(Melis et al., 1994).

Prior to 2006, modern debris-flow activity in Arizona, outside of the Grand Canyon, was

mainly wildfire related (Wohl and Pearthree, 1991; Pearthree and Youberg, 2004; Jenkins et al.,

2011). The trend of decreasing size and extent of debris-flow deposits from the late Pleistocene

to late Holocene, along with the dearth of historical non-wildfire related debris flows, suggested

that debris flows generated by extreme precipitation did not represent a significant geologic

hazard in the modern environment of Arizona (Webb et al., 2008b). That view was challenged when an unusual multiday pattern in late July, 2006, resulted in approximately 1000 hillslope failures in four mountain ranges across southeastern Arizona (Pearthree and Youberg,

2006; Magirl et al., 2007; Webb et al., 2008b). Debris flows damaged or destroyed infrastructure in Coronado National Memorial (Huachuca Mountains) and in Sabino Canyon (Santa Catalina

Mountains), while debris flows in the northern Chiricahua Mountains and in Aravaipa Canyon in

22

the were numerous but isolated (Webb et al., 2008b; J. P. Cook, unpublished

data). Most of the hillslope failures transformed into debris flows, with many coalescing into

larger flows. In the Santa Catalina Mountains larger, coalesced debris flows exited or nearly

exited the mouths of five canyons along the front range, and caused significant alluvial fan

flooding at the mouth of one canyon (Webb et al., 2008b).

Griffiths and others (2009) analyzed radar and rain gage data from the Santa Catalina

Mountains to estimate return intervals for this series of storms. Return intervals for individual

daily storms were less than two years, while return intervals for average 2-day storms were greater than 50 years, and greater than 200 -500 years for the 4-day storm with return intervals

up to, and greater than, 1000 years in some areas (Griffiths et al., 2009). High antecedent soil moisture conditions were important for initiation of these saturation-induced debris-flow

(Griffiths et al., 1996; Webb et al., 2008b).

Debris-flow frequency in the Santa Catalina Mountains are on the order of thousands of years (Youberg et al., 2008), which are typically longer than those found for the Grand Canyon.

Return intervals of debris-flow triggering storms, however, are significantly less than return intervals for debris flows in both the Santa Catalina Mountains and Grand Canyon (Webb et al.,

2005). The occurrence of so many debris flows triggered by extreme precipitation in 2006, in

addition to numerous post-wildfire debris flows during the last few decades, highlights the fact

that debris flows still pose a significant geologic hazard in Arizona. Under modern climate and

fire regimes, however, the greatest threat of debris flows in Arizona may be from wildfires.

23

1.2.2 Post-wildfire debris flows

With the exception of the debris flows in the Grand Canyon and other limited documentation of debris flows related to extreme precipitation across the state, most of the recent

debris flow activity in Arizona has been linked to wildfires. Over the past several decades, there

has been an increase in fire-related debris flows as the size and severity of wildfires have

increased (Pearthree and Youberg, 2004; Schaffner and Reed, 2005). Since the mid-1970s, the

annual area burned by wildfires in Arizona has increased dramatically (Grissino-Mayer and

Swetnam, 2000). Indeed, eight of Arizona’s ten largest historic wildfires have occurred since

2002, burning over 670,000 ha (Southwest Coordination Center,

http://gacc.nifc.gov/swcc/index.htm). While the occurrence of widespread wildfires in American

Southwest is not a new phenomenon, as tree-ring records show regionally synchronous fires have recurred for centuries (Swetnam and Betancourt, 1998), the way in which wildfires behave has

changed dramatically. Ponderosa Pine forests that previously burned by frequent, low intensity

surface fires, and mixed conifer forests that burned in small, patchy mosaics of severity, now often burn by high-intensity crown fires across extensive areas, frequently resulting in very large contiguous expanses of denuded forests and damaged soils (Schoennagel et al., 2004; Williams et al., 2010). Several factors have contributed to this change, including fire suppression efforts, livestock grazing, fuels buildup, and climate change (Westerling et al., 2003; Fule et al., 2012).

Key climatic factors vary by forest type, and include long-term drought, or wet year(s) followed by a dry year (Swetnam and Baisan, 1996; Grissino-Mayer and Swetnam, 2000; Swetnam and

Anderson, 2008; Littell et al., 2009). Recently, the long-term effects of fire suppression, in conjunction with changes in climate, have begun to have a greater influence on the size and severity of wildfires (Swetnam and Baisan, 1996; Littell et al., 2009; Williams et al., 2010). 24

Post-wildfire hydrologic responses are strongly related to soil burn severity (e.g. DeBano,

2000; Neary et al., 2005; Moody et al., 2008; Nyman et al., 2011; Shakesby, 2011). Watersheds burned by intense fires and classified as high soil burn severity indicate near complete consumption of surface vegetation, litter and duff (Parsons et al., 2010), which results in decreased surface roughness and increased hydrologic connectivity (Larsen et al., 2009; Cawson et al., 2013). Soils burned at high severity have disaggregated soil structure due to consumption of soil organics and fine roots (Cannon and Reneau, 2000; Moody and Nyman, 2012), and reduced infiltration capacity due to hyper-dry soil conditions, fire-induced soil-water repellency, decreased hydraulic conductivity, surface sealing and air entrapment (Shakesby et al., 2000;

Moody et al., 2005; Doerr et al., 2006; Moody et al., 2009; Ebel et al., 2012), These changes result in increased runoff volumes with higher boundary shear stresses, and thin layers of non- cohesive soils with lower erosional thresholds (Moody et al., 2005; Nyman et al., 2013). Thus, critical shear stress thresholds for sediment entrainment are crossed with less contributing area than in pre-fire conditions (Istanbulluoglu et al., 2004; Gabet and Bookter, 2008; Wohl, 2013), and rills develop very high on the upper portions of hillslopes (Moody and Martin, 2001; Moody and Nyman, 2012). Rills concentrate fluid flow and transport pulses of sediment downslope into coalescing flow areas, valley heads or channels, where runoff-induced debris flows can form as described by the sediment capacitor model (Kean et al., 2013), the fire hose effect or other mechanisms (Johnson and Rodine, 1984; Gabet and Bookter, 2008; Santi et al., 2008; Parise and

Cannon, 2012). Excess runoff and erosion results in flows that may be either sediment-laden flood flows or debris flows (Moody and Martin, 2001). Post-fire sediment-laden flood flows occur more frequently than debris flows, but debris flows can be significantly more destructive than floods (Gartner et al., 2004). Some factors identified as influencing the occurrence of post- 25

fire debris flows or floods include burn severity, , catchment size and gradient, and storm

characteristics and movement through the basin (Wohl and Pearthree, 1991; Cannon and Gartner,

2005; Cannon et al., 2008). Post-fire erosion, leading to floods and debris flows, are also a function of the rainfall intensity (Moody and Martin, 2009; Cannon et al., 2011; Kean et al.,

2011; Staley et al., 2012).

Unlike saturation-induced debris flows, post-fire debris flows usually occur due to runoff during the first few storms after the fire when antecedent soil moisture is low or absent (Kean and Staley, 2011). Post-fire debris-flows tend to be generated by relatively common, 2- to 10-

year frequency storms. Cannon and others (2008) analyzed rainfall intensities from convective storms that generated 25 post-fire debris flows in southwestern Colorado, and frontal storms that generated 68 post-fire debris flows in southern California. Results showed that return intervals for these storms were typically two years or less with average intensities ranging from 1.0 to 32.0 mm h-1 in Colorado and 1.3 to 20.4 mm h-1 in California (Cannon et al., 2008). Moody and

Martin (2009) analyzed sediment yields under different regional rainfall regimes, which were

based on rainfall seasonality (winter:summer) and the 2-year, 30-minute intensity domains (37 to

58 mm h-1). These researchers found that the majority of post-fire sediment is eroded from channels, and that the sediment yields are related to the timing of rainfall following a fire

(Moody and Martin, 2009). In Arizona, monsoon rains tend to immediately follow the wildfire season, often providing the means whereby fires ultimate get extinguished.

Fire-related debris flows have been generated by monsoonal storms all across Arizona

(Wohl and Pearthree, 1991; Pearthree and Youberg, 2004; Youberg and Pearthree, 2011). Wohl and Pearthree (1991) documented fire-related debris flows in the Huachuca Mountains generated

26

from a summer monsoon thunderstorm following the 1988 Peak Fire. They also found evidence

for numerous other debris flows in several basins in the Huachuca Mountains, including debris

flows following a 1977 fire (Wohl and Pearthree, 1991). In Dorothy Ryan Canyon, one of the

main study basins for the 1988 debris flows, two older debris flow deposits were identified in a

scoured channel bank. Based on soil development in the exposed deposits, the two oldest debris

flow deposits were interpreted to be late Pleistocene and early Holocene. Debris-flow recurrence intervals for basins in the Huachuca Mountains were estimated to be a few hundred to several thousand years (Wohl and Pearthree, 1991).

Jenkins and others (2007) looked at recent and paleodebris flows in several basins on

Kendrick Peak, north of Flagstaff. Debris flows generated after the 2000 Pumpkin Fire eroded channels and exposed older debris-flow deposits. Jenkins (2007) identified 56 sediment units in her study basins, of which 34 were debris flows and 45 were fire-related (debris flows, hyperconcentrated flows, and soils). Radiometric 14C indicated that the debris-flow deposits were clustered in intervals in the majority of basins around (Jenkins, 2007). More recently, debris flows were generated from monsoon storms following the 2004 Nuttal-Graham

Fire in the , the 2004 Willow Fire in the , the 2010

Schultz Fire near Flagstaff, the 2011 Monument and Horseshoe 2 Fires in the southeastern part of the state, the 2011 Wallow Fire in east-central Arizona, and the 2012 Gladiator Fire north of

Phoenix (Pearthree and Youberg, 2004; Youberg et al., 2011; Youberg and Pearthree, 2011).

Although post-fire debris flows have been documented in Arizona, methods to assess the

potential post-fire debris-flow hazards have not been established for Arizona. Following three

recent major fires, the USGS Water Science Center used a USGS post-fire debris-flow hazard

27

model (Model A, Cannon et al., 2010) to assess the hazards from the Wallow, Horseshoe 2 and

Monument Fires (Ruddy, 2011a, b, c). This model, and two related models (Models B and C,

Cannon et al., 2010), were developed using data from the intermountain western U.S., but not with data from Arizona. Moreover, none of these models had been tested in Arizona for effectiveness prior to their use in 2011.

28

CHAPTER 2: THE PRESENT STUDY

This dissertation represents a collection of research projects organized into three journal

article manuscripts dealing with debris flows in Arizona. Documentation of historical and modern debris flows reveal that debris flows do occur across Arizona, and still present a

potential geologic hazard. While saturation-induced debris flows from prolonged or intense

rainfall may not be as likely as in past climatic condition, the threat from post-wildfire debris

flows is increasing. The work reported on in this dissertation reflects studying debris flows from a hazards perspective. The research contained in this study represents the efforts of many collaborative researchers and colleagues, some who are co-authors on the following chapters.

2.1 Appendix A: Increased debris-flow activity and magnitude during the Pleistocene-

Holocene climatic transition, Santa Catalina Mountains, Pima County, Arizona; implications for modern debris-flow hazards.

Co-Authors: A. Youberg, R.H. Webb, C.R. Fenton, P.A. Pearthree.

Submitted to , reviewed, in revision.

The study in Appendix A builds upon a 2008 study (Youberg et al., 2008) using a

combination of geologic mapping, 10Be cosmogenic age dating techniques, modeling and GIS

analysis to assess the frequency and magnitude of debris-flow deposits at the mouths of two canyons, Pima and Finger Rock, emanating from the Santa Catalina Mountains, and also the potential initiation mechanisms of late Pleistocene to early Holocene debris flows. The manuscript presented here was submitted to Geomorphology and is now in revision.

29

2.1.2 Abstract

Hazard mitigation for extreme events such as debris flows requires geologic data,

particularly for alluvial fans near mountain fronts in the southwestern United States. In July

2006, five consecutive days of monsoonal storms caused hundreds of debris flows in

southeastern Arizona, particularly in the southern Santa Catalina Mountains north of Tucson.

Before 2006, no historical debris flows from the Santa Catalina Mountains reached the populated

mountain front, although abundant evidence of prehistoric debris flows is present on downslope

alluvial fans. We used a combination of surficial geologic mapping and 10Be exposure dating to

produce a debris-flow history for Pima and Finger Rock Canyons. The largest debris flows, of

latest Pleistocene to early Holocene age, covered much of the apices of alluvial fans formed at the

mouths of these canyons and extended up to 3 km downslope. These debris-flow deposits were

inset against higher and older alluvial surfaces with few, limited debris-flow deposits of late

Pleistocene age. The 10Be ages in this study have considerable scatter for surfaces believed to be

of uniform age, indicating the dual possibilities of inheritance from previous cosmic-ray exposure, as well as the potential for composite deposits derived from numerous debris flows.

We then used an empirical inundation model, LAHARZ, to assess probable magnitudes of the older debris flows to evaluate possible initiation mechanisms. In-channel and floodplain storage within the canyons is not sufficient to generate volumes likely needed to produce the larger later

Pleistocene to early Holocene debris-flow deposits. The abundance of latest Pleistocene and early

Holocene deposits suggests that large debris flows were either generated during Pleistocene or during the instability associated with climate change at the Pleistocene-Holocene transition. Without disturbances, such as watershed-scale wildfires, large fan-forming debris flows are unlikely to occur in Pima and Finger Rock Canyons in present-day climate regime. 30

2.2 Appendix B: Rainfall intensity-duration thresholds and the predictive strengths of

three post-wildfire debris-flow models for use in Arizona.

Ann Youberg, Susan H. Cannon, Jason W. Kean, Dennis M. Staley.

Manuscript intended for submission to Natural Hazards. In prep.

Appendices B and C are companion studies and focus on the aftermaths of wildfires. In

Appendix B, the predictive strengths of three multivariate statistical debris-flow probability models and one debris-flow volume model, developed with data outside of Arizona, were assessed for use in Arizona’s varied physiographic regions. Objectively defined rainfall intensity-duration thresholds were developed to provide guidelines for issuing warnings in

Arizona. Guidelines for use of the debris-flow probability models were also provided.

2.2.1 Abstract

Wildfires across the western U.S. are increasing in size and severity, while at the same time encroachment of development into the wildland-urban interface putting more people at risk to wildfires and their aftermaths. Teams conducting rapid assessments of post-wildfire geologic hazards need models to quickly identify areas at risks for post-fire debris flows, and forecast centers and emergency response agencies need rainfall intensity-duration thresholds that identify rainfall conditions likely to generate post-fire debris flows. While empirical methods to estimate the rainfall thresholds, probability, and volume of post-fire debris flows have been developed for specific regions of the Western U.S., it is not known how well these tools perform in other regions with different geologic and hydrologic characteristics. Here, in an effort to expand the domain where accurate post-fire debris-flow hazards assessments and forecasts can be made, we assess the predictive strengths of existing probability and volume models in three different 31

physiographic regions of Arizona, and we define a new rainfall threshold valid for Arizona.

Specifically, we test the volume model and three probability models using data from 80 recently

burned basins in the Colorado Plateau, the Central Highlands Transition Zone, and the Basin and

Range of Arizona. We show that all of the models have adequate predictive strength throughout

most of the state, with the exception of lower gradient basins on the Colorado Plateau (2011

Wallow Fire) where all models performed poorly. We also show that the debris-flow volume model over-predicts in all of our study areas. Our analysis shows that the choice of a model for a hazard assessment depends strongly on the area where it will be used, and the degree to which having true outcomes outweighs the risk of failed or false alarms. We also developed objectively defined rainfall intensity-duration thresholds to identify rainfall conditions likely to generate post-wildfire debris flows in Arizona. The objectively defined I10 and I15 thresholds of 52 and 42

mm h-1, respectively, have the strongest predictive strengths of the five threshold models

developed, although all of the threshold models performed well. Results from this study provide

guidelines and considerations for selecting an appropriate debris-flow model, and rainfall threshold values for issuing forecasts and warnings for Arizona.

2.3 Appendix C: Basin physiographic and soil burn severity factors influencing the generation of post-wildfire debris flows.

Ann Youberg, Victor R. Baker, Philip A. Pearthree

Manuscript intended for submission to Natural Hazards. In Prep.

Appendix C a multi-phased exploratory study, delving into an analysis of parameters

used by the models in the previous study, and other parameters to identify hillslope-scale

patterns useful in assessing potential post-fire debris-flow basins. Results from this study show

32

that a combination of two metrics, as described by percent of basin area with high soil burn

severity on slopes ≥30°, can be used to help discriminate between basins likely to produce post-

wildfire runoff-induced debris flows and those that are not.

2.3.1 Abstract

Post-wildfire debris-flows hazards impose increasing risks as wildfires increase in size

and severity. Here we analyze physiographic and soil burn severity data from 86 basins recently

burned by wildfires in Arizona, 56% of which produced debris flows. The data come from 3

distinctive physiographic provinces, and they correspond to multivariate statistical model

parameters that have been widely applied to post-wildfire debris-flow hazard assessments in the

intermountain western U.S. We extracted morphometric factors (basin ruggedness, slopes, etc.)

in a GIS framework from 10 m DEMs and soil burn severity parameters from 30 m dNBR

continuous and 10 m categorical data. Physiographic model factors did discriminate non-debris- flow (nD) from debris-flow (D) producing basins, though only weakly, for storms that followed

5 major 2010-2012 fires (Schultz, Wallow, Gladiator, Horseshoe 2, and Monument). However, slope-area plots for the basins revealed a consistent difference in slopes between nD and D basins. By studying 12 discrete intervals of slope values ranging between 14° and >42° we found that a very effective discriminating criterion is the percent basin area with slopes above a threshold of 30°. Moreover, the discrimination between nD and D basins becomes stronger with a combination of percent basin areas with high soil burn severity on slopes above this threshold value. Thus, a combination of steep slopes and high soil burn severity can be used to identify basins subject to post-wildfire debris flows according to specific criteria identified in this study.

Another criterion shown to discriminate nD from D basins is mean elevation, which probably is

33

important because of the zonal vegetation and related soil organic and root cohesion factors that

characterize Arizona. Finally, we found that post-fire channel heads mapped for the 2011

Monument Fire formed at essentially the same slope range (~30-400) as 2006 saturation induced

debris flows, a result that is both useful for identifying potential post-fire debris flow hazards and

may provide information to guide the future development of physically based models.

CHAPTER 3: CONCLUSIONS

3.1 Summary

The three studies presented here (Appendices A-C) assessed paleo and modern debris

flows to (1) understand debris-flow hazards in the modern environment within the context of past

and current climate changes, (2) assess the predictive strengths of three models to provide tools

for assessing post-wildfire debris-flow hazards in Arizona, (3) develop an Arizona post-wildfire

rainfall intensity-duration threshold curve for issuing warnings, and (4) explore measures of

basin physiographic and burn severity factors to identify patterns in specific, easily measured

criteria that may help discriminate between non-debris-flow and debris-flow producing basins.

The study in Appendix A shows that debris flow processes were most active during the late Pleistocene to early Holocene. Evidence of older debris-flow deposits was not observed but may be buried. Results from this study show that earlier debris flows were of higher magnitude and extended farther from the mountain front than more recent debris flows. This is likely due to decreasing sediment availability over time (Bardou and Jaboyedoff, 2008). The 2006 debris flows throughout southeastern Arizona, and numerous recent post-wildfire debris flows

throughout the state, were generated from thin (<~1.5 m) saturation-induced hillslope failures or

from runoff in concentrated flow areas that then scoured sediment from channels. Inundation 34

modeling results show, to a first order approximation, that sediment volumes needed to generate

late Pleistocene to early Holocene debris-flow deposits are not available in the modern channel networks. Although there is no observable evidence of large landslides in either Pima or Finger

Rock Canyons today, it is likely that larger, deep-seated landslides were one of the dominant mechanisms when debris-flow activity began in these drainages. Under current climate changes and hillslope conditions, debris flows generated in the modern environment will not be as large or travel as for onto alluvial fans as past flows did, and thus the hazard from saturation-induced debris flows is limited. With current and predicted climate changes, however, the greater potential hazards are from the aftermaths of wildfires.

The study in Appendix B presents the results from assessing the predictive strengths of three multivariate debris-flow probability models and one debris-flow volume model. These models were assessed with data from five different burned areas in diverse geologic and geomorphic settings. Results show that all three models perform adequately in four cases, but with variable results, and all three models failed in one case (Wallow Fire). Due to the variable results, complicated guidelines for choosing which model to use, and an appropriate threshold, in different areas of the state were also presented in Appendix B. Finally, objectively defined rainfall intensity-duration thresholds for peak intensities between I5-I60, and one statewide

rainfall intensity-duration threshold curve were developed for issuing post-wildfire debris-flow

warnings in Arizona (Appendix B, Figure B.9). More data are needed, however, to develop

objectively defined thresholds for the different physiographic regions in Arizona. Likely the

thresholds will be higher in the south where monsoon is stronger than in the north, similar to

findings by Moody and Martin (2009).

35

The third study (Appendix C) was motivated by the second study (Appendix B). Using

the large dataset of burned basin flow responses developed during the second study, various

basin physiographic and soil burn severity factors were explored to identify patterns and criteria

that can be used to discriminate between potential non-debris-flow (nD) and debris-flow (D) producing basins. Findings from this study show that a metric of percent basins area with high soil burn severity on slopes ≥30° provides a stronger discrimination between nD and D basins than basin metrics such as mean basin gradient or relief, which weakly discriminate between basin responses. Mean basin elevation was also found to discriminate nD from D basins and likely is a proxy for forest type and density, which influence soil production, root density and the magnitude of post-disturbance erosion (Istanbulluoglu et al., 2004; Pelletier et al., 2013). Finally, in the Huachuca Mountains, runoff-induced post-fire channel heads were found to form on slopes similar to saturation-induced hillslope failures (~30-40°). These findings are also similar to results by Kean et al. (2013) that show runoff-induced debris flows modeled by the sediment capacitor model were the generated on comparable slopes.

3.2 Implications for future post-wildfire debris-flow hazard assessments

Results from the first study (Appendix A) indicate that post-wildfire debris-flow pose greater risks in Arizona than do saturation-induced debris flows from extreme precipitation events, although these risks are not zero. Results from the studies presented in Appendices B and

C indicate that additional, or possibly different, methods are needed to adequately assess the risks of post-wildfire debris flows. Physical models that assess conditions for generating saturation-induced debris flows are not, in general, appropriate for post-wildfire debris flows.

Immediately after wildfires, hillslope soils, especially in high burn severity areas, are dry or

36

hyper dry with no antecedent moisture (Kean et al., 2011; Moody and Ebel, 2012). As precipitation falls throughout the post-wildfire season, hillslopes regain moisture, but debris flows from later in the summer often still do not have initiation scarps and are likely generated by runoff. Therefore models that consider pore pressures and transient soil moisture (e.g.,

Montgomery and Dietrich, 1994; Iverson, 2000) will not adequately capture the important processes leading to the generation of post-wildfire debris flows. A new model developed by

Kean and others (2013) couples several equations to describe the complicated processes of runoff-induced debris-flow initiation and surge magnitude and frequency. This sediment capacitor model is an important and significant step forward in understanding these runoff- induced debris flows and towards more consistent hazard assessments. While this work, and other studies (e.g. Moody and Nyman, 2012; Nyman et al., 2013) have helped clarify processes leading to post-wildfire debris-flow generation, many questions remain. For example, how can channels be evaluated to quantify the amount of sediment available for entrainment and transport? Or, how quickly do reload after a scouring event? Answers to these questions have implications for both emergency assessments and the development of physically based models.

3.1 Rainfall assessments

Rainfall has been used in various ways to characterize conditions under which debris flows are generated. Rainfall intensity-duration thresholds curves describe rainfall intensities for various storm durations that are likely to generate debris flows from either saturated soils or runoff. Rainfall intensity-duration thresholds developed for runoff-induced debris flows are more distinct because other hydrologic factors, such as antecedent moisture, do not influence initiation

(Kean et al., 2013). Objectively defined rainfall intensity threshold values are developed for

37

different time intervals (Appendix B, I5-I60) from which a curve can be developed. The

individual threshold values provide more flexibility for agencies that issue warnings, as threshold

values can be chosen based on available rainfall/radar data and storm movements (Appendix B).

Here, these values provide an opportunity to compare rainfall intensity during initiation of

saturation-induced debris flows to runoff-induced debris flows. Griffiths and others (2009) conducted detailed analyses of radar and rain gauge data from the 2006 extreme precipitation event in the Santa Catalina Mountains. They found that, on the day of the failures, that intensities

-1 -1 -1 were 28 mm h (I5), 22 mm h (I15), and 17 mm h (I60) (Griffiths et al., 2009). These rates were

very similar to the preceding four days. Comparing these values with the objectively defined

thresholds developed during the second study (Appendix B), post-wildfire rainfall intensities 78

-1 -1 -1 mm h (I5), 42 mm h (I15), and 22 mm h (I60). From this comparison, it is clear that peak intensities for shorter durations (≤I30) are much higher for runoff-induced post-wildfire debris flows than saturation-induced debris flows. Intensity values for longer time intervals (I60) are similar, however, echoing findings by Kean and others (2011). And, as expected and discussed in

Appendix B (Figure B.9), post-wildfire runoff-induced debris flows are generated from shorter duration, higher intensity rainfall than what generates saturation-induced debris flows.

Rainfall data are also used in the multivariate statistical debris-flow probability and volume models (Cannon et al., 2010). The probability models relate to average storm intensities,

while the volume model employs storm totals. Since these models were developed, however,

newer research has shown that debris-flow initiation is best correlated to peak 10-minute

intensities, but only volume is related to short-duration peak intensity (≤30 minutes) and poorly

correlated to total storm rainfall (Kean et al., 2011). Debris flow volumes are also related to

38

channel bed moisture, with wetter channel beds entraining sediment more easily and therefore

producing larger debris flows than dry beds (Iverson et al., 2011; McCoy et al., 2012). While hillslope soils dry out very quickly after a post-fire rain (Kean and Staley, 2011), channels may accumulate and retain moisture, making the bed more erodible.

In the five burned areas in this study (Appendices B and C), the first documented debris flows typically occurred within a week or two after containment of the fires with additional

flows occurring throughout the monsoon season. The earliest debris flow event occurred one day

before confinement, and the latest occurred on October 1st. Both these debris flows occurred in

tributaries within Miller Canyon on the Monument Fire in the Huachuca Mountains. Debris

flows continue to occur three years after the 2010 Schultz Fire and two years after the 2011

Horseshoe 2 Fire. These recent debris flows were likely saturation-induced from heavy 2011

monsoon rains and illustrate the ability of these areas to quickly reload channels with

transportable sediment.

3.2 Basin assessments

The pace of studies assessing post-wildfire erosion, debris-flow generation, and the mechanisms of these processes has dramatically increased in past decade or so. There are two main approaches to these studies, one is process based and the other focuses on emergency assessment. Appendix B addressed the later, Appendix C touched on the former. Below is a synthesis of the state of knowledge regarding the generation of runoff-induced, post-wildfire debris flows, including some considerations for future assessments of fire prone areas, from both process based and emergency assessments viewpoints. Process based questions relate more to landscape evolution, disturbance regimes, and recurrence intervals. Emergency assessment

39

questions deal with how to determine the important process based information in order to evaluate risks from wildfires. Both types of questions are critical and inform the development of each field.

The emerging conceptual model of post-wildfire debris-flow generation focuses on two components of a basin, upper basin hillslopes and channels. In general, upper basin hillslopes strongly control the generation of post-wildfire and need to be assessed in the context of slopes, soil burn severity and potential magnitude of post-wildfire erosional responses. Channel conditions determine if the flow will be propagated and how large the flow will be. Thus channels need to be assessed in the context of channel slope, confinement, bed moisture levels, and in particular, sediment available for transport. Channel erosion accounts for approximately

80% of eroded post-fire sediment (Moody and Martin, 2001), and may be higher when only considering debris flows.

Results from the study described in Appendix C show that a metric of percent basin area with high soil burn severity on slopes ≥30° strongly discriminates between nD and D basins

(Figure C.10). Mean basin area also helps to discriminate nD from D basins and may be a proxy for forest type and density. Forest density is directly related to surface roughness, root cohesion, slope stability and the magnitude of episodic erosion after disturbances such as wildfire (Roering et al., 2003; Istanbulluoglu et al., 2004). In an emergency assessment environment, forest type and density, as extracted from either LANDFIRE data (http://www.landfire.gov/) or local forests

GIS layers, can help focus field efforts for assessing debris-flow hazards. For process-based assessments, a more appropriate method is to consider the eco-pedo-geomorphic evolution of the study area (Pelletier and Rasmussen, 2009a, b; Pelletier et al., 2013).

40

Pelletier and others (2013) have shown that mean basin elevation is also a proxy for

efficient energy and mass transfer (EEMT), an erosional term that combines temperature and

precipitation into a single variable, correlates well to rates of bedrock in semi-arid

environments, and has been used to quantify soil production from eroding bedrock. EEMT has

been shown to increase linearly with elevation in southeastern Arizona mountain ranges, and

non-linearly with related variables of above-ground biomass and soil thickness (Pelletier et al.,

2013). Vegetation on upper basin hillslopes, considered as either forest type or AGB, increases

surface roughness, decreases runoff and increases shear stress needed to detach and transport soil

particles thus increasing soil depth. Therefore, the feedback between parameters is such that

higher EEMT leads to thicker soils, increased AGB, and increased slope stability (Roering et al.,

2003; Pelletier et al., 2013). When a fire does occur, and vegetation is completely removed, soils

and hydrologic changes occur that increase runoff volumes and velocities, and lower erosional

thresholds.

Immediately after a fire a highly erodible, thin (<~2-3 cm), non-cohesive surface soil

layer is present on burned hillslopes due to consumption of soil organics and fine roots (<2 mm

diameter) (Nyman et al., 2013). The variable shear strength of the burned soil is a function of burn severity and related to the density of fine roots at the surface (<~1 cm) (Moody and Nyman,

2012). At depths below the non-cohesive layer, soil shear strength is related to silt, clay and bulk density and increases dramatically but similarly across a broad range of sites (Nyman et al.,

2013). Consumption of surface vegetation, duff and litter results in decreased surface roughness,

increased hydrologic connectivity (Larsen et al., 2009; Cawson et al., 2013), and increased

runoff volumes and velocities, crossing critical shear boundary conditions with less contributing

41

area (Parsons et al., 1996; Moody and Nyman, 2012). Consequently, rills develop very high on upper hillslopes (Abrahams et al., 1995; Moody and Martin, 2001), concentrating fluid flow and transporting pulses of sediment downslope into coalescing flow areas, valley heads or channels.

Runoff-induced debris flows form in these areas as described by the sediment capacitor model

(Kean et al., 2013), the fire hose effect or other mechanisms (Johnson and Rodine, 1984; Gabet and Bookter, 2008; Santi et al., 2008; Parise and Cannon, 2012). The post-fire hillslope sediment availability is related to the non-cohesive soil layer, and also to background erosivity rates

(Nyman et al., 2013; Pelletier et al., 2013).

The thickness of the non-cohesive layer can be determined in the field by using a shear vane (Nyman et al., 2013), which provides an opportunity for both burned area emergency response (BAER) teams and researchers to calculate erosional thresholds and available hillslope sediment that can be transported by rills into valley heads were runoff-induced debris flows can form.. BAER teams can easily incorporate shear vane tests at sites where hydrophobicity is being tested, and on upper hillslopes in basins identified, via probability modeling and GIS analysis, as potential debris-flow producing basins. These data can also help constrain and calibrate landscape evolution models to better understand the influence of episodic erosion on the landscape.

Below the debris flow initiation zone, critical channel conditions that determine flow characteristics and debris-flow volumes include sediment available for entrainment and transport, the moisture level of bed material (McCoy et al., 2012), slopes and channel confinement. Once a debris flow is initiated, the viability and volume of the debris flow is related to the amount of channel bed material available to scour and entrain, and channel bed moisture

42

levels. A debris flow can be initiated and begin to travel down channel, but if no additional

sediments are available to scour and entrain, then the flow may transition to a hyperconcentrated

or flood flow as more runoff enters the channel (Kean et al., 2011). Channel bed wetness is

related to ease of scouring and entraining sediment during the passage of a debris-flow front.

Wet channel beds are more easily entrained by passing surge fronts and result in larger volumes than dry beds (Iverson et al., 2011; McCoy et al., 2012).

Assessing the amount of channel sediment available for transport is difficult, time- consuming, and so far, an intractable problem. Channel sediment available for entrainment and transport is a function of climate, bedrock and soil erosivity, soil depths, vegetation types and density, slopes, and time since the last channel scouring event. On a broad scale, and during a pre-fire assessment, these parameters can be assessed using the landscape modeling methods developed by Pelletier and others (2013). Such analyses would help constrain channel reloading and recurrence intervals. In an emergency assessment, however, these factors are much more difficult to evaluate, and currently no methods or models exist to extract this information, other than by field observations. This is an area critically in need of further research.

3.3 Conclusions

A clear picture of post-wildfire debris-flow generation has recently emerged. Changes to soils and hydrology on upper basin hillslopes result in a thin layer of non-cohesive hillslope soils that are easily transported by rills into coalescing flow areas. These pulses of sediment contribute to the initiation of runoff-induced post-wildfire debris flows. Questions remain concerning how to efficiently evaluate channels to quantify sediment available for entrainment and transport, and

43

bed material moisture levels, both of which contribute to debris-flow volumes. Modeling is needed to predict how quickly channels reload and hillslope soils recover.

44

APPENDIX A: INCREASED DEBRIS-FLOW ACTIVITY AT THE PLEISTOCENE- HOLOCENE CLIMATIC TRANSITION, SANTA CATALINA MOUNTINS, PIMA COUNTY, ARIZONA; IMPLICATIONS FOR MODERN DEBRIS-FLOW HAZARDS

Ann Youberg*,a, Robert H. Webbb,1, Cassandra R. Fentonc, Philip A. Pearthreea a Arizona Geological Survey, 416 W. Congress, Suite 100, Tucson, AZ 85701 USA b U.S. Geological Survey, 520 N. Park Avenue, Tucson, AZ 85719 USA c Helmholtz-Zentrum Potsdam – Deutsches GeoForschungsZentrum, Telegrafenberg, D- 14473 Potsdam, Germany * corresponding author tel: +1 520.770.3500; fax: +1 520.770.3505 E-mail: [email protected] 1Present address: School of Renewable Natural Resources, University of Arizona, Tucson, AZ 85719 USA

Prepared for Geomorphology Submitted, 8 June 2013, Reviewed, In Revision

45

A.1 Abstract

Hazard mitigation for extreme events such as debris flows requires geologic data,

particularly for alluvial fans near mountain fronts in the southwestern United States. In July

2006, five consecutive days of monsoonal storms caused hundreds of debris flows in

southeastern Arizona, particularly in the southern Santa Catalina Mountains north of Tucson.

Before 2006, no historical debris flows from the Santa Catalina Mountains reached the populated

mountain front, although abundant evidence of prehistoric debris flows is present on downslope

alluvial fans. We used a combination of surficial geologic mapping and 10Be exposure dating to

produce a debris-flow history for Pima and Finger Rock Canyons. The largest debris flows, of

latest Pleistocene to early Holocene age, covered much of the apices of alluvial fans formed at the

mouths of these canyons and extended up to 3 km downslope. These debris-flow deposits were

inset against higher and older alluvial surfaces with few, limited debris-flow deposits of late

Pleistocene age. The 10Be ages in this study have considerable scatter for surfaces believed to be

of uniform age, indicating the dual possibilities of inheritance from previous cosmic-ray exposure, as well as the potential for composite deposits derived from numerous debris flows.

We then used an empirical inundation model, LAHARZ, to assess probable magnitudes of the older debris flows to evaluate possible initiation mechanisms. In-channel and floodplain storage within the canyons is not sufficient to generate volumes likely needed to produce the larger later

Pleistocene to early Holocene debris-flow deposits. The abundance of latest Pleistocene and early

Holocene deposits suggests that large debris flows were either generated during Pleistocene climates or during the instability associated with climate change at the Pleistocene-Holocene transition. Without disturbances, such as watershed-scale wildfires, large fan-forming debris flows are unlikely to occur in Pima and Finger Rock Canyons in present-day climate regime. 46

Keywords: debris flows; surficial geology; cosmogenic 10Be dating; LAHARZ, arid; semi-arid; Arizona

47

A.2 Introduction

Urban development on alluvial fans near mountain fronts exposes housing and

infrastructure to numerous geologic and hydrologic hazards. In the arid and semiarid

southwestern United States, historical debris flows have commonly occurred following extreme

precipitation, or in recently burned watersheds (Wohl and Pearthree, 1991; Pearthree and

Youberg, 2004; Cannon et al., 2008; Kean et al., 2011). In this environment, debris flows are extreme events that deliver large amounts of large-caliber sediment to active alluvial fans (Wells and Harvey, 1987), dominating fan deposition patterns and profoundly influencing flow paths during smaller floods. Floodplain management strategies typically focus on 100-year floods or smaller, and flood control measures in this region generally utilizing channel-bank stabilization measures or requiring building above the calculated 100-year water surface, ignoring the potential for channel aggradation and subsequent overbank deposition during debris flows

(Webb et al., 2008b). An understanding of past debris-flow magnitude and frequency, as well as of prehistoric areas of deposition, is essential for placing the hazard of debris flows into a temporal framework in order to assess risk (Youberg et al., 2008). Moreover, because evidence left by debris flows tends to persist in the landscape for a long time, the significance of abundant undated debris-flow deposits is difficult to evaluate when assessing modern hazards.

During the last week of July 2006, southern Arizona experienced five consecutive days of early morning storms generated from monsoonal moisture mixing with a persistent low-pressure system centered over northwestern New Mexico (Magirl et al., 2007). This series of storms culminated on July 31, when the last pulse of rain fell on already saturated watersheds in the

Santa Catalina Mountains north of Tucson (Griffiths et al., 2009). Floods of record occurred in several washes throughout the region, and hundreds of hillslope failures and debris flows were 48

generated in at least four southeastern Arizona mountain ranges (Pearthree and Youberg, 2006;

Webb et al., 2008b). Debris flows occurred in nine canyons along the eastern half of the Santa

Catalina Mountains (Figure A.1). Debris flows exited or nearly exited the mouths of five of these canyons (Figure A.1), damaging roads, bridges, and one private house close to the mountain front (Webb et al., 2008b).

There are no reports of previous historical debris flows that exited the Santa Catalina mountain front, nor any that affected developed areas or infrastructure. Evidence of prehistoric debris flows, however, is abundant on the now-developed fans at the mouths of all canyons along the mountain front (Youberg et al., 2008). While debris flows were previously recognized as a hazard in Arizona’s mountains and canyons (Webb et al., 1988; Wohl and Pearthree, 1991;

Webb et al., 1999; Pearthree, 2004), the number and extent of debris flows from the 2006 event was unexpected, and it dramatically illustrated the potential for sizable debris flows to exit canyons and adversely impact properties near canyon mouths.

Following the 2006 debris flows, two studies were conducted to assess the magnitude and frequency of the 2006 event (Webb et al., 2008b), and of prehistoric debris-flow deposits found on the alluvial fans along the mountain front (Youberg et al., 2008). Webb et al. (2008b) documented floods of record on six watercourses in southeastern Arizona, and 435 shallow, translational hillslope failures in the Santa Catalina Mountains with an average failure depth of

0.75 m ± 0.4 m that, as a whole, produced a total of approximately 670,400 m3 of sediment.

Youberg et al. (2008) documented debris-flow deposits on all fans along the mountain front ranging in age from late Pleistocene to modern, with older deposits containing larger caliber clasts and extending farther from the mountain front than modern deposits (Youberg et al.,

49

2008). 10Be cosmogenic samples were collected from boulders in four canyons in an attempt to

provide absolute ages of deposits. These results generally corroborated the relative age dating of

the geologic mapping, but showed significant inheritance and reworking issues (Youberg et al.,

2008).

Two companion studies expanded on the work of Webb et al. (2008b), providing a

detailed assessment of the 2006 storm event (Griffiths et al., 2009) and an assessment of the efficacy of a statistically calibrated, empirical, inundation model, LAHARZ, for mapping hazard zones in arid and semi-arid environments (Magirl et al., 2010). Recurrence intervals for the individual daily storms were less than two years, while return intervals for average 2-day storms were greater than 50 years, and approximately 1000 years for the entire multi-day storm

(Griffiths et al., 2009). Results from this study re-enforce the rarity of the 2006 storm event.

LAHARZ is a GIS-based empirical model developed to map hazard zones from volcanic lahars (debris flows) (Iverson et al., 1998; Schilling, 1998). LAHARZ predicts the lateral and downstream inundation based on user-defined volumes, mobility coefficients derived from a worldwide dataset, and topography as described by a digital elevation models (DEM) (Iverson et

2/3 2/3 al., 1998). LAHARZ uses a two-thirds power-law of the form A = c1V and B = c2V to

describe the relationship between area and volume, where A is the cross-sectional area and B is

the planimetric area of inundation (Iverson et al., 1998). Lahars contain large amounts of clay

due to their volcanic origin and thus are very mobile. The coefficients to describe lahars are c1 =

0.05 and c2 = 200 (Iverson et al., 1998). LAHARZ was later extended to non-volcanic related

debris flows. Based on a worldwide dataset, the coefficients for debris flows in LAHARZ are c1

= 0.1 and c2 = 20 (Griswold and Iverson, 2008). Webb et al. (2008b) used data from six of the

50

2006 debris flows to develop southern Arizona coefficients of c1 = 0.1 and c2 = 40. Magirl et al.

(2010) used the southern Arizona coefficients with estimated sediment volumes (Webb et al.,

2008b) and field mapped deposits to model debris-flow deposition from a single failure and from aggregate failures. Results from this study found that the southern Arizona coefficients produced quite good results for single failures, but slightly over-predicted cross-sectional area and under- predicted planimetric area (downstream extent) for aggregate failures (Magirl et al., 2010).

Model runs with the worldwide debris-flow coefficients produced cross-sectional areas that were too wide and planimetric areas that were too short, and runs with lahar coefficients produce cross-sectional areas that were too narrow and planimetric areas that were far too long (Magirl et al., 2010). Overall, LAHARZ with the southern Arizona coefficients produced fairly accurate inundation extents with limited data, indicating that LAHARZ could be a useful tool for assessing debris-flow hazards (Magirl et al., 2010).

In this study, we build upon the studies of Youberg et al. (2008) and Magirl et al. (2010) in a continued attempt to assess debris-flow frequency and magnitude in order to place debris-flow hazards into the context of the modern environment. Here, we focus our efforts on two alluvial fans at the mouths of Pima and Finger Rock Canyons (Figure A.1). We conducted additional geologic mapping to assess areas of the fans not included in Youberg et al. (2008), re-assessed 10Be ages using a newer half-life calculation, and utilized LAHARZ to estimate a first-order approximation of debris-flow volumes needed to emplace the mapped late Pleistocene to early Holocene deposits on each fan. We also conducted field assessments within each canyon to assess potential sediment available for transport within the channel networks, and to look for evidence of mass movements.

This allowed us to assess potential initiation mechanisms and to further assess modern debris-flow hazards. 51

A.3 Study Area

Pima and Finger Rock Canyons drain the southwestern side of the Santa Catalina

Mountains north of Tucson, Arizona (Figure A.1). The Santa Catalina Mountains are a

metamorphic core complex uplifted from deep crustal levels during Oligo-Miocene lithospheric

extension, and are bounded by the moderately to gently south-dipping Catalina detachment fault

to the south and the post-mid-Miocene high-angle Pirate fault to the west (Spencer and

Reynolds, 1989; Dickenson, 1999; Spencer and Pearthree, 2004). Rock units exposed in Pima

and Finger Rock Canyons are middle Proterozoic Oracle Granite intruded by leucogranite sills of

the Eocene Wilderness-suite granites (Dickenson, 1991; Force, 1997; Spencer and Pearthree,

2004). The peraluminous, two-mica luecogranite sills of the Wilderness-suite granites are the main cliff-forming units in the study area (Keith et al., 1980). Drainages emanating from the southern flank of the Santa Catalina Mountains are structurally controlled bedrock channels with orientation and channel shape mainly influenced by joint density (Pelletier et al., 2009), but also by corrugations of the detachment fault (Spencer, 2000). The longitudinal profiles of the channels have large-scale steps (1-10 m) with nickpoints located in the thicker, more resistant sections of leucogranite sills of the Wilderness-suite granites with lower joint density, and lower gradient alluvial reaches located in areas of higher joint density and less resistant rocks of Oracle

Granite (Pelletier et al., 2009).

The alluvial fans at the mouths of the Pima and Finger Rock Canyons were chosen because the upper parts of each fan near the mountain front are covered by boulder deposits from past debris flows (Figure A.2), and both canyons and the apices of their alluvial fans are undeveloped and managed by the U.S. Forest Service (USFS). Neither canyon had debris flows

52

in 2006, yet both canyons have ample evidence of prehistoric debris flows and small historic events of much lower magnitude than occurred in eastern drainages in 2006.

Pima and Finger Rock Canyons are oriented south-southwest with similar elevation headwaters (Figure A.1). Finger Rock Canyon is the shorter and steeper of the two basins (Table

A.1). Downslope of the undeveloped USFS lands, both alluvial fans have undergone low- to moderate-density development with infrastructure designed to withstand streamflow floods, not debris flows. The lower alluvial fan of Finger Rock Canyon was developed in the 1970s, and development on the Pima Canyon alluvial fan began in the early 2000s.

A.4 Methods

A.4.1 Geologic mapping of debris-flow deposits

During the first study (Youberg et al., 2008) we mapped debris-flow deposits from just upstream of the USFS boundary to the downstream extent of mappable deposits. In that study we were most concerned with how far beyond the mountain front into developed areas past debris flows had traveled. In this study we conducted additional geologic mapping on the undeveloped sections of the alluvial fans emanating from Pima and Finger Rock Canyons (Figs 3-6 modified from Youberg et al., 2008). Debris-flow deposits were mapped using a combination of extensive field observations aided with 1960 (scale 1:9,000) and 2007 (scale 1:12,000) black-and-white aerial photographs; 2002 color digital orthophotographs; and 2007 LiDAR-derived topography

(1 m resolution). The undeveloped sections of each fan were mapped from the apex downstream to the USFS boundary, which was ~1.5 km downstream for Pima Canyon and ~1.2 km downstream for Finger Rock Canyon (Figures A.3 and A.4). Unlike the lower fan areas where

53

development has masked or substantially modified debris-flow deposits, the upper fans have well-preserved debris-flow deposits (Figure A.2).

Relative ages of debris-flow deposits were estimated using characteristics of clast weathering (Melton, 1965a; Bull, 1991), soil development of the underlying alluvial surfaces

(Gile et al., 1981; Machette, 1985; Birkland et al., 1991; Bull, 1991; Birkland, 1999), and

position in the landscape. Specific characteristics of each map unit are given in Table A.2.

Debris-flow deposits may or may not have a matrix of fine-grained particles because these may

have been removed during recessional flow or by post-event reworking (Webb et al., 1988). As a result, the near-surface parts of debris-flow deposits, particularly older deposits, typically have an open framework. Clasts in older debris-flow deposits also generally exhibit darkened weathering rinds; some degree of oxidation (orange surface patina); thin, discontinuous

carbonate coatings on buried clasts; fracturing and splitting; and, in some deposits, burial up to

1-m depth. In contrast, clasts in younger deposits have no weathering rinds and weak to non-

existent surface patinas, and thus are brighter and fresher looking, with little to no soil

accumulation or clast fracturing (Youberg et al., 2008).

During this mapping phase, we also made field observations to document conditions

within each canyon. Pima Canyon is larger and has a wider valley with a floodplain composed of

terraces with abundant debris-flow deposits. Finger Rock is smaller, and the canyon bottom is

narrower, as is the floodplain. The extent, height and material in the floodplains were noted

along the main channel. We also documented tributaries with evidence of past debris flows,

typically levees, and looked for evidence of landslides and other mass failures.

54

A.4.2 Cosmogenic age dating of debris-flow deposits

Cosmogenic dating, in its simplest usage, reflects the amount of time a rock or mineral has been exposed at the ’s surface to cosmic ray bombardment (Gosse and Phillips, 2001).

Cosmogenic 10Be is the most extensively used cosmogenic nuclide for establishing the exposure

ages of surficial landforms containing quartz (Bierman et al., 1995; Gosse and Phillips, 2001;

Matmon et al., 2005; Matmon et al., 2006; Dühnforth et al., 2007), which is abundant in the

boulders in this debris-flow study.

Production rates of cosmogenic 10Be increase with elevation and latitude. Production

rates can vary depending on longitude, which can account for significant variations in

atmospheric pressure and temporal fluctuations in the earth’s magnetic field (Stone, 2000; Lifton

et al., 2005). Production of cosmogenic nuclides is highest at a rock’s surface (e.g., the top 4 cm)

and then decreases exponentially with depth (Gosse and Phillips, 2001). Production at 1 m is

approximately 50% of that at the rock surface, and typically rocks below 2 m depth are

essentially shielded from cosmic rays and have negligible production of cosmogenic isotopes. In

the simplest scenario, large boulders found on debris-flow surfaces in this study were either shielded deeply in bedrock or in hillslope deposits prior to debris-flow transport, deposition and finally exposure at the debris-flow surface. In this case, all of the accumulation of cosmogenic nuclides occurred after boulders reached their current locations. Many clasts found in debris- flow deposits, however, likely would have been exposed to cosmic rays on hillslopes or along channels before they were deposited in their current locations, thus accumulating cosmogenic nuclides before debris-flow deposition (Bierman et al., 1995; Fenton and Pelletier, 2013).

55

The total concentration of cosmogenic 10Be nuclides in a rock/mineral is a combination

of 10Be accumulated since emplacement of the rock in a landform of interest (i.e., a debris-flow deposit) plus any 10Be that may have accumulated in the rock before deposition in the new

landform. Corrections can be made using depth profiles and/or amalgamation samples in

deposits to correct for an inherited nuclide from a prior exposure history (e.g., Anderson et al.,

1996; Wolkowinsky and Granger, 2004). Cerling et al. (1999) and Webb et al. (1999)

demonstrated that Holocene debris flows could be accurately dated with cosmogenic 3He using

only boulder surfaces and without any depth-profile corrections. An on-line calculator developed

for the -Earth project was used for calculating 10Be exposure ages of boulders in debris

fans emanating from Pima and Finger Rock Canyons (Table 4, Results Section; Balco et al.,

2008; http://hess.ess.washington.edu/math/al_be_v22/al_be_multiple_v22.php).

Multiple surfaces composed of debris-flow deposits were identified along the modern

channels of Pima and Finger Rock washes, just downstream from the canyon mouths. Rock

samples for cosmogenic dating were collected from boulder tops on surfaces of clast-supported

boulder trains that were arranged stratigraphically from youngest (lowest and closest to the channels) to oldest (highest and farthest from channels). The degree of desert varnish decreased with decreasing geomorphic/stratigraphic age. The surfaces associated with different levels of debris-flow deposition had sufficient horizontal extent to be mappable (for example, see Figure

A.2). For each surface, a minimum of three boulders were selected with the following characteristics: (1) the boulders had similar amounts of weathering and/or desert varnish and appeared to be representative of primary deposition on the surface; (2) the boulders were stable and did not appear to have rolled since deposition; (3) the boulders had a relatively high amount of quartz with low feldspar and mica; (4) the top of the boulders either were at, or protruded 56

above, the surrounding surface with minimal shielding from cosmic ray bombardment; (5) the

top of the boulders either were relatively level or had a measurable slope that would allow

estimation of self-shielding, and (6) grussification or other evidence of high weathering rates

were absent (Table A.3). Using a hammer and chisel, samples representing 3-6 cm depth were collected from the tops of suitable boulders.

Boulder samples yielded abundant quartz, from which 10Be contents were extracted and

used to calculate exposure ages (see Table A.4, Results Section). The standard laboratory

procedure of Bookhagen (2007) was used to produce 10Be targets for accelerator mass

spectrometer analysis, with some modifications as noted in Youberg et al. (2008). Samples were

prepared at the University of Arizona’s cosmogenic nuclide facility in the Department of

Geosciences. Replicates of samples 070108-02 and 070109-03 were also prepared at Lawrence

Livermore National Laboratories for quality control (Table A.4). For a detailed description of

sample preparation refer to Youberg et al. (2008).

The 10Be targets were analyzed on the accelerator mass spectrometer (AMS) at the

Lawrence Livermore National Laboratory in Livermore, California (https://cams.llnl.gov/). This

AMS is the most appropriate one to use in the United States for samples with Holocene exposure

ages because it has a 10Be/9Be ratio resolution of 1.046 . 10-16, which is required to obtain

10Be/9Be measurements for very young samples (i.e., 1,000-year-old debris-flow deposits). To gain higher beam strengths from targets, this laboratory switched from silver (Ag) to niobium

(Nb) (R.C. Finkel, personal communication, 2007), which increases the yield in the spectral widths for beryllium and increases the accuracy of measurements. Measured 10Be/9Be results

were converted to 10Be concentrations (at/g) and used to estimate ages (Table A.4).

57

A.4.3 Debris-flow modeling

Debris-flow modeling was conducted in two stages. First, LAHARZ was used to

calculate a first-order approximation of debris-flow volumes necessary to emplace the mapped

late Pleistocene-early Holocene debris-flow deposits. Inundation modeling was conducted using

LAHARZ with three sets of coefficients describing lahars, worldwide debris flows, and southern

Arizona debris flows, and four volumes based on orders of magnitude of 104, 105, 106, and 107

m3. Landslide scarps or other evidence of mass movement failures were not observed in either

canyon, therefore we modeled by selecting the beginning points of the depositional zones

(Griswold and Iverson, 2008; Magirl et al., 2010), which were identified on the basis of field mapping of deposits at each canyon mouth. Topography used in LAHARZ was 1 m resolution

DEMs derived from airborne LiDAR collected by the Pima County Flood Control District in

2007.

The second stage was conducted in GIS. The stream channel network within each canyon was extracted using the TauDEM toolbox (Tarboton, 2005) and spatial analyst tools in Esri’s

ArcMap to calculate channel lengths and Strahler stream order. The stream network was used to

estimate the amount of material that could have been available by channel scour alone. Estimates

of available channel material were made by compiling cumulative channel lengths for each

Strahler stream order and using a combination of field observations to estimate channel widths

and depths for different stream orders, and measured width-depth ratios from four channels

scoured by debris flows in 2006 in the Huachuca Mountains (Youberg et al., 2006; Youberg, unpublished data). Tributary channels were included in the network and in calculations only if evidence of past debris-flow activity was observed. Floodplains along the main channels were

58

mapped as one polygon; the main channel area was subtracted from the floodplain area. A range

of potential channel network storage volumes were then calculated and compared with the

modeled LAHARZ volumes.

A.5 Results

A.5.1 Geologic mapping

Mapping of debris-flow deposits on the fans of Pima and Finger Rock Canyons revealed

similar patterns of debris-flow occurrence. Older Pleistocene deposits with moderately strongly

developed soils (Table A.2; Figures A.3 and A.4, unit I) are located highest in the landscape and

are generally confined to the apex of the alluvial fan complex. The largest and most extensive

deposits on both fans are estimated to be latest Pleistocene to early Holocene in age based on

deposit characteristics (unit IY). Near the fan apexes, unit IY was divided into two units, IYa

(higher and presumably older) and IYb, based on inset relationships. Downstream, where this

relationship is not evident, deposits of this age group are undivided (IY). Two Holocene units,

Y1 and Y2, are inset below unit IY (Figures A.3 and A.4). The lateral extents and topographic

positions of IYa, IYb, Y1 and Y2 provide information about channel migration and incision

since the late Pleistocene.

A.5.1.1 Finger Rock Canyon

The Finger Rock Canyon alluvial fan complex contains all of the debris flow units

described above (Figure A.3). Finger Rock Wash has cut a narrow, bedrock-confined channel

that comprise now entrenched below older debris-flow deposits near the canyon mouth. The lower part of the canyon and the upper part of the alluvial fan is an excellent study area, as debris-flow deposits of various ages are preserved and have not been obscured by development. 59

Late Pleistocene to early Holocene debris-flow deposits (units I and IY) are extensive

near the apex of the Finger Rock alluvial fan (Figure A.3). These deposits are the highest

elevation debris-flow deposits in Finger Rock Canyon, and are up to 10 m above the entrenched

active channel (Figure A.5). Two different levels of IY deposits can be identified on the upper

fan where deposits of unit IYb are typically inset 1-2 m below unit IYa.

Intermediate debris-flow deposits (unit Y1, early to mid-Holocene) are generally inset

less than 1 m below IYb deposits and typically are 2-5 m above the active channel on the upper fan (Figure A.5). Y1 deposits are found surrounding several IYb deposits, attesting to changing channel flow paths during avulsing debris flows. Young debris-flow deposits (unit Y2, mid to late Holocene) are generally 1-3 m above the active channel on the upper fan, and inset approximately 2 m below Y1 deposits.

A.5.1.2 Pima Canyon

Pima Canyon has a well-defined channel entrenched below a broad floodplain within the canyon. At the canyon mouth, bedrock is exposed in the channel bottom. Pima Canyon Wash exits the mountain front and flows westward between a large latest Pleistocene to early Holocene alluvial fan to the south and bedrock to the north (Fig. 4; Klawon et al., 1999). Approximately 1

km from the fan apex, the present-day channel turns southward at the interface with a mid-

Pleistocene alluvial fan. Topography and sedimentary deposits on the latest Pleistocene to early

Holocene fan surface document channel migration from a more southerly flow to the modern

westerly flow. Deposition and build-up of the fan by older debris-flow deposits (units IY, IYa,

and IYb) diverted channel flow westward (Figure A.4). Younger debris-flow deposits (units Y1

and Y2) are inset and north of IY debris-flow deposits. Much of the evidence for debris-flow

60

deposits on the Pima Canyon fan outside of USFS lands is now obscured by residential

development.

Pleistocene deposits (unit I) define the southeastern extent of late Pleistocene and

younger debris flows in Pima Canyon (Figures A.4 and A.6). Pleistocene debris-flow deposits are perched on an older fan surface, are 1-2 m higher than adjacent IY deposits, and are approximately 5 m above the active channel (Figure A.6). Unit I deposits are similar in appearance to IY deposits, but soils are moderately reddened and clasts are well-weathered. IY

deposits are approximately 3-4 m above the active modern channel on the upper fan (Figure

A.6). Near the fan apex IY deposits are divided into higher IYa and slightly lower IYb deposits.

The inset relationship of IYa and IYb, as well as an abandoned channel on the south side of the

latest Pleistocene to early Holocene fan, indicate channel avulsion. Downslope and away from

the mountain front, inset relationships between IYa and IYb cannot be discerned and are mapped

as IY undifferentiated (Figure A.4).

Intermediate debris-flow deposits (unit Y1, early to mid-Holocene) are generally 2-3 m

above the active channel and 1-2 m below unit IY (Figure A.6). Intermediate debris-flow

deposits are highest near the mountain front. Young debris-flow deposits (unit Y2) are typically

0.5-1.5 m above the active channel. In general Y2 deposits are inset 1-2 m below Y1 deposits,

but some Y2 deposits on-lap and cover Y1 deposits.

A.5.2 Cosmogenic 10Be ages on debris-flow surfaces

Cosmogenic 10Be provide age estimates of boulders from four terrace levels on the

alluvial fans of Pima and Finger Rock Canyons (Figures A.5 and A.6). Exposure ages of 21

dated boulders from eight mapped debris-flow deposits ranged from 63.9 to 5 ka (thousand years

61

before present) (Table A.4, Figure A.7). While the exposure ages of the sampled boulders show

considerable scatter, both for individual debris-flow deposits and with respect to the stratigraphic

position (surfaces in stratigraphically younger positions have older 10Be ages), the measurement

uncertainties associated with these results are consistently low (Table A.4, 9-15%). Measurement

uncertainties are a combination of the ability of the lab to make the measurement (internal

uncertainty) and variations in production rate and other factors such as shielding (external

uncertainties). Uncertainty with the exposure ages are related to inheritance problems and the

potential for multiple debris flows to be represented by one mapped debris-flow unit.

A.5.2.1 Finger Rock Canyon

In Finger Rock Canyon, the relatively old, stratigraphically high debris-flow unit IYa was sampled approximately 8 m above the active channel (Figure A.5b). Two boulders (FR-01, -02)

yielded exposure ages of 12.3±1.1and 13.4±1.2 ka that overlap within uncertainty (Table A.4,

Figures A.3, A.5 and A.7). Replicate samples for FR-02 yielded indistinguishable ages (13.4±1.2 ka and 13.3±1.2 ka). The range of exposure ages including uncertainty (11.2-14.6 ka) at this sample site is consistent with the estimated age of latest Pleistocene to early Holocene assigned to map unit IYa (Table A.2).

Exposure ages of boulders sampled from stratigraphically younger deposits mapped as unit IYb show considerably more scatter (Table A.4). These deposits are 6-7 m above the active channels. Samples FR-04, -05, -06, and -07 from four boulders in a western deposit have exposure ages ranging from 37.2 to 7.6 ka, while samples FR-09, -10, and -11 from three boulders in an eastern deposit have exposure ages ranging from 20.5 to 10.1 ka (Figures A.3, A.5 and A.7). Since stratigraphic relationships indicate unit IYb is younger than unit IYa, exposure

62

ages from samples in unit IYb older than 14.6 ka likely reflect exposure prior to emplacement in

their current positions, but may also indicate that age differences between IYa and IYb are not

significant. Sample FR-06 yields an exposure age of 7.6±0.7 ka, which does not overlap the

remaining ages even within uncertainty. Based on the stratigraphic position of boulder FR-06 (6 m above the active channel) and relative-age dating characteristics of unit IYb, this exposure age is likely to be too young and may represent erosion or exhumation. Age-data from these boulders highlights the uncertainty and difficulty in accurately dating these debris-flow deposits using cosmogenic nuclides. Still, 10.1 to 14.8 ka is a reasonable age estimate for map unit IYb from the Finger Rock Canyon catchment.

The stratigraphically lowest debris-flow deposit (unit Y1) sampled in Finger Rock

Canyon sits approximately 3 m above the active wash (Figure A.5). Three boulders (FR-18, -17, and -16) sampled from this deposit have exposure ages of 16.2±1.6, 9.2±0.9 and 5±0.5 ka, respectively (Figures A.3, A.5 and A.7). The two younger ages are consistent with unit Y1’s stratigraphic position and surface and deposit characteristics, which suggest an age of early to middle Holocene (Table A.2). Sample FR-18 yields an exposure age more consistent with exposure ages from higher IYa and IYb debris-flow deposits. It is likely that this boulder (FR-

18) was reworked from the older deposit into the Y1 debris flow during deposition.

A.5.2.2 Pima Canyon

In Pima Canyon, samples were collected from boulders in four mapped debris-flow deposits adjacent to the active channel (Figure A.6b). The oldest mapped debris-flow deposits

(unit I and IYa, Figure A.4) were not sampled due to access restrictions. The oldest and stratigraphically highest debris-flow deposit (IYb) that was sampled is about 3-4 m above the

63

active channel. Three boulders samples (P-01, -03, -04) yield exposure ages ranging from 63.9 to

17 ka (Table A.4, Figures A.4, A.6 and A.7). Sample P-01 and its replicate yielded exposure ages of 17±1.6 and 18±1.6 ka. These are indistinguishable and consistent with the previously estimated latest Pleistocene to early Holocene age for this unit. Samples P-03 and -04 yielded exposure ages much older (>31.8 ka) than mapping of the debris-flow deposit suggests (Table

A.2). These samples are considered as outliers, affected by previous exposure.

Two boulders were sampled from the stratigraphically lower Y1 deposit about 3 m above the active channel on the south side of the Pima Canyon drainage. These samples (P-06 and -07) yielded exposure ages of 13.3±1.2 and 13.6±1.2 ka (Table A.4; Figures A.6 and A.7). These overlapping ages are younger than exposure ages from than the Pima Canyon IYb unit

(exposure-age estimate: 15.4-19.6 ka), but they are older than relative age estimates of early to middle Holocene for unit Y1. This suggests that samples P-06 and -07 may have a previous exposure history. Boulder sample FR-14 was collected from another Y1 deposit on the north side of the Pima Canyon channel, approximately 2 m above the active wash. Sample P-14 yielded an exposure age of 9.4±0.9 ka (Figures A.4 and A.7). While this date is more consistent with the relative age of the Y1 deposit based on geologic mapping, it is difficult to assess the exposure age of the deposit with only one boulder sampled.

Boulders sampled from the stratigraphically lowest deposit (Y2) in Pima Canyon were collected about 1 m above the active wash. These samples (P-10, -11, and -12) yielded exposure ages of 8.6, 29.8, and 30.2 ka, respectively (Figures A.4 and A.7). Ages yielded by samples -11 and -12 are much older than the middle to late Holocene age estimated for Y2 units based on topographic and surface characteristics. The third sample (P-10) yields an exposure age that is

64

younger than those of samples -11 and -12; however, the age of sample P-10 (8.8±0.8 ka)

overlaps with the youngest exposure age in Pima Canyon’s stratigraphically older Y1 unit. The

Y2 deposit is not very high above the active channel, and boulders from older debris-flow deposits were likely remobilized and deposited in the younger unit.

A.5.2.3 Comparison of exposure ages of map units in Finger Rock and Pima Canyons

Cosmogenic ages obtained from various map units provide different estimates for

surfaces considered to be of similar age in these canyons. On the Finger Rock fan, cosmogenic

ages are consistent with stratigraphic relationships and are reasonably consistent with age

estimates based on geomorphic criteria if old outliers are excluded. Cosmogenic age estimates

obtained from surfaces on the Pima Canyon fan are less internally consistent and are generally

older than geomorphic age estimates. Boulders in both of the Pima and Finger Rock deposits

yielded very consistent exposure ages, yet are mapped as three different geologic units based on

relative age-dating characteristics. This illustrates the limitations of geologic mapping and

exposure-age dating for assigning numerical ages to specific map units from separate drainages.

Within each drainage, the relative mapping and exposure-age estimates are consistent, but they are not necessarily consistent when comparing debris-flow history in geomorphically disconnected catchments.

A.5.3 Debris-flow modeling

Debris-flow modeling was conducted in two steps. First, we used LAHARZ to find a first-order approximation of volume needed to emplace mapped IY deposits that were mapped

(Figure A.8). Then, for each canyon, we extracted a channel network with associated Strahler

65

stream order and calculated potential channel material volumes using field-based width to depth

ratios.

A.5.3.1 LAHARZ model results

LAHARZ modeling was conducted using three sets of coefficients describing lahar,

worldwide debris flows, and southern Arizona debris flows. We selected the top of the

deposition zone based on the mapped locations of debris-flow deposits at the mouths of each

canyon, and tested four volumes magnitudes of 104, 105, 106, and 107 m3.

Results from our model runs were similar to those found by Magirl et al (2010). Model

runs using the lahar coefficients were too narrow (cross-sectional area) and traveled much too far downstream (planimetric area), beyond the edge of the DEM. The worldwide debris flow coefficients were just the opposite, expanding too far laterally (cross-sectional) and upslope, but not far enough downstream (planimetric area). Model runs with the southern Arizona coefficients show inundation more similar to the mapped deposits (Figure A.8), but still expanding slightly too far laterally nor far enough downstream, although results with this model are better than with the worldwide model. Results from the LAHARZ modeling indicate volumes on the order of 105 to 106 m3 were necessary for emplacing mapped IY deposits.

A.5.3.2 Canyon in-channel storage analysis

Field assessments were conducted within each canyon to assess the amount of material stored along the main channel, and to document tributary channels that have evidence of past debris flows. The mapping in Pima Canyon is more extensive because the canyon bottom is wider and the trail is located along the main channel, providing views of stored channel and floodplain material, and conditions at confluences with tributaries. Evidence of past debris flows 66

on tributaries is well preserved at or just above the confluences. Mapping in Finger Rock Canyon is limited because the trail leaves the canyon floor almost immediately making accurate observations more difficult.

Figure A.9 shows the Pima Canyon tributaries with evidence of past debris-flow activity mapped in blue. The mapped floodplain along the main channel is outlined in yellow, and two

LAHARZ volumes, 105 and 106 m3, are shown along with the mapped debris-flow deposits on the alluvial fan (Figure A.9). The Strahler orders of tributary channels ranged from 1-4, and the main channel had Strahler orders of 4 and 5. Width to depth ratios used to calculate the potential magnitude of in-channel storage ranged from 2 to 5. Volume estimates of potential in-channel storage in Pima Canyon ranged from 3 x 104 m3 to 1.4 x 105 m3 for channel storage, and 1.4-3 x

105 m3 for floodplain storage, for a total potential in-channel storage ranging from 2-4.5 x 105 m3.

A.6 Discussion

We used a combination of new geologic mapping and a re-assessment of cosmogenic

10Be exposure ages to assess the timing and frequency of debris flows emanating from two canyons in the southern Santa Catalina Mountains near Tucson, Arizona. Geologic mapping differentiated debris-flow deposits by relative ages using topographic and landscape position, clast weathering, and soil development (Table A.2, Figures A.3 and A.4). Relative age estimates for mapped deposits range from late Pleistocene to late Holocene. After careful evaluation taking into account the geologic context of our samples, cosmogenic 10Be exposure ages support the relative age dating within each drainage (e.g., in Finger Rock Canyon), but do not provide simple age constraints for individual map units present in both Finger Rock and Pima Canyons. Wide

67

age discrepancies occurred due to inheritance, re-working, and composite deposits, illustrating

some of the difficulty in dating alluvial fans comprised of debris-flow deposits.

We then used a statistically calibrated, empirical model to assess potential debris flow volumes needed to emplace mapped late Pleistocene-early Holocene debris-flow deposits (Figure

A.8), and calculated the magnitude of potential in-channel and floodplain storage within the canyons to evaluate probable initiation mechanisms.

A.6.1 Ages of debris-flow deposition

Geologic mapping can only provide broad age ranges of debris-flow deposits given the considerable uncertainty in weathering, rates of varnish deposition, and soil formation.

Numerous relatively young (latest Pleistocene to Holocene) debris-flow deposits were mapped, but the number of separate mapped deposits does not represent the number of debris-flow occurrences. A single flow may leave several deposits, typically as levees. Numerous separate debris flows may successively contribute material to a single composite deposit. This was observed in 2006 when new boulders were deposited on top of older levees (Youberg et al.,

2008). Many deposits on the Pima and Finger Rock alluvial fans have characteristics indicating they were emplaced through multiple events or pulses within events. Conversely, evidence of individual debris flows may not have been preserved due to reworking, erosion and removal, or deposition and burial from either active channel processes or subsequent debris flows. This uncertainty precludes the ability of mapping to provide information on the absolute number of debris flows represented within a debris-flow surface.

We encountered similar limitations using 10Be cosmogenic dating to estimate absolute

ages of surfaces. Some boulders sampled from a single mapped deposit (FR-01 and FR-02; P-06

68

and P-07; and P-11 and P-12) show consistency in 10Be ages, but most units yielded ages of

considerable variability (Table A.4; Figures A.3, A.4, and A.7). The ranges of 10Be exposure

ages (Figure A.7) show a wide range of ages from boulders within five of the eight sampled

deposits. The most obvious explanation for apparent discrepancies in cosmogenic ages is that

inheritance or re-working resulted in cosmogenic nuclide accumulation prior to emplacement of

boulders in their current locations (Bierman, 1994; Nichols et al., 2006; Fenton and Pelletier,

2013).

Inheritance or re-working results in exposure-age estimates that are older than the actual

emplacement age of a deposit. If we make this assumption, as has been done in other studies (i.e.

Bierman et al., 1995), then the youngest exposure age dates are likely to most closely

approximate deposit age. This may be a reasonable assumption for debris flows in the study area

as there are numerous, large, old-appearing debris-flow deposits within Finger Rock and Pima

Canyons that could have contributed boulders with extensive prior exposure histories to younger debris flows. Evidence of composite deposits in our study, however, complicates this assumption. There is evidence for both composite deposits and re-working in Pima and Finger

Rock Canyons. In addition, uncertainties related to fan- and/or boulder-surface erosion and complex fan histories can add to the difficulty of correlating different deposits or geomorphic surfaces (Bierman et al., 1995; Matmon et al., 2005; Matmon et al., 2006).

While we acknowledge the limitations of both geologic mapping and cosmogenic 10Be

dating in this study, the combination of field relationships and exposure ages does provide useful

information about the late Pleistocene and Holocene history of debris-flow activity in the Santa

Catalina Mountains (Figure 7). Cosmogenic 10Be dating is broadly consistent with the relative

69

ages of the deposits assigned by geologic mapping, and exposure ages decrease with

stratigraphic/topographic position. The extensive debris-flow deposits on the upper alluvial fans

of Pima and Finger Rock Canyons, along with similar deposits mapped elsewhere along the

mountain front (Youberg et al., 2008), record periods of aggradation during the latest Pleistocene

to early Holocene and attest to the primary importance of debris flows (e.g., unit IY) in

constructing alluvial fans during that time. Although the depth of channel entrenchment at the

time of their deposition may well have been less than it is today, multiple lines of evidence

suggest that at least some of these debris flows were substantially larger than any that have

occurred in the middle and late Holocene. Previous geologic mapping documented large debris-

flow deposits, including snout-like deposits, with similar varnish and weathering characteristics

up to 3 km down channels below the fan apex (Youberg et al., 2008). The extent and caliber of

these deposits suggests that some debris flows during this time were fairly large. Mapping also

shows that areas of middle to late Holocene debris-flow deposits (e.g., unit Y2) are smaller in size and extent than the latest Pleistocene-early Holocene debris flow deposits. Moreover, there is no clear evidence that these debris flows continued beyond the mountain front fans. Our data suggest that debris flows were either uncommon before the Late Glacial Maximum, or evidence for older debris flows was buried by late Pleistocene deposition. We infer that the increase in debris flow activity in the late Pleistocene was related to some combination of vegetation change in the watershed and increased intensity of precipitation related to the glacial/interglacial climate transition (Van Devender et al., 1990; Wagner, 2006; Pigati et al., 2009).

70

A.6.2 Implications from debris-flow modeling

LAHARZ modeling was conducted to help constrain potential debris-flow volumes

needed to emplace IY deposits. There are, however, limitations to LAHARZ. This model does

not, nor is it intended to, accurately map debris-flow depositional patterns. Rather, the model

identifies probable extents of lateral (cross-sectional) and downstream (planimetric) inundation

using empirical data and modern topography. While this approach is appropriate for hazard

mapping in the current environment, it does limit the effectiveness of the modeling here. We

cannot know the actual topography of the depositional zone during the late Pleistocene and early

Holocene, although the mapped deposits do provide a general picture.

In both Pima and Finger Rock Canyons, bedrock is exposed in the channels at the fan

apices. Therefore, the channels cannot have been any deeper than they are today, but they could

have been higher. Indeed, mapping in Finger Rock indicates that the modern channel is incised

up to 10 m deeper than during the late Pleistocene (Figures A.3 and A.5). Deposits on the Pima

Canyon fan show channel migration through time with the late Pleistocene channel located

farther south than the modern channel (Figure A.4). Understanding the limitations of modeling,

and the information provided by mapping, helps for interpreting model results. And while other

models provide a better assessment of depositional patterns (e.g. FLO-2D, O'Brien et al., 1993), the issue of modern versus prehistoric topography remains. LAHARZ model results provide adequate order-of-magnitude informative for assessing past debris-flow volumes.

Results from the LAHARZ modeling indicate volumes on the order of 105 to 106 m3 are

needed to emplace IY deposits. In-channel and floodplain storage calculations from Pima

Canyon indicate potential volumes of 105 - 105.5 m3, if all channels were completely evacuated in

71

one event. This scenario would look similar to the 2006 event where relatively shallow (<~1 m),

translational failures occur in hollows at the tops of channels or hillslope swales, and the

resulting debris flows scour channels below the failures to bedrock, a “bulking up” phenomenon

that has been observed elsewhere in Arizona (Webb et al., 1999). Many of the 2006 debris flows in tributary channels did not exit the canyon but were deposited at the confluence with the main channel where they remain today (Webb et al., 2008b). Indeed, evidence of debris-flow deposits within the Pima Canyon floodplain deposits show that debris flows in the canyon do not always travel to the fan. In addition, the 2006 sediment volumes from all nine canyons total 6.7 x 105

m3. This is a near-approximate volume required from one canyon for the IY deposits. Thus, is it

unlikely that a 2006 event could produce the volumes necessary for the IY deposits, and

mechanisms that produce larger volumes are necessary.

During our fieldwork, we searched for evidence of larger landslides or mass failures.

Remnants of landslides are present in Sabino Canyon to the east of Finger Rock Canyon (Force,

1997), and Banks (1976) mapped two small landslide deposits in Pima Canyon, although we

couldn’t find these features during our field work. Large slump-block or rotational landslides would produce debris flows with volumes sufficient to emplace the extensive Pleistocene-

Holocene debris-flow deposits, and a combination of larger landslides and smaller channel scouring debris flows could account for the in-canyon storage, the extent of the Pleistocene-

Holocene debris-flow deposits and the scatter in the 10Be dates.

A.6.3. Climatic conditions at the Pleistocene-Holocene transition

The numerous, large and extensive Pleistocene-Holocene debris-flow deposits document

the occurrence of relatively abundant and large debris flows in the late Pleistocene, while the

72

fewer, smaller and less extensive middle to late Holocene debris-flow deposits indicate continued but declining activity through the Holocene. Proposed reasons for these geomorphic responses during and after the Pleistocene-Holocene transition include (1) increased sedimentation rates due to intense rain over hillslopes undergoing desertification and subsequently reduced sedimentation rates and fan-head entrenchment due to increased aridity that slowed sediment production (Harvey et al., 1999), (2) an increase in late- winter/spring Pacific frontal systems (April-June) or dissipating tropical storms (August-

October) (McDonald et al., 2003), or (3) cyclical wet periods followed by droughts and wildfires that resulted in initially large pulses of sediment that tapered off through the Holocene as the amount of stored material available for transport was depleted (Swanson, 1981; Bull, 1991;

Meyer et al., 1995; Kirchner et al., 2001; Pierce et al., 2004; Eppes and McFadden, 2008;

Frechette and Meyer, 2009). Arundel (2002) and others (Weng and Jackson, 1999; Betancourt,

2003) suggest an increase in wildfires as a response to late Pleistocene climate changes. In view of the very large and high severity wildfires of today it is tempting to invoke this mechanism. In the modern environment, high frequency, low magnitude convective storms produce significant erosion and debris flows in areas disturbed by wildfires (Wohl and Pearthree, 1991; Meyer et al.,

1995; Cannon, 2001; Pearthree and Youberg, 2004; Pierce et al., 2004; Frechette and Meyer,

2009; Jenkins et al., 2011).Post-wildfire debris flows, however, are typically runoff-induced and mainly scour channels. In addition, although the presence of wildfires in Arizona forests during the late Pleistocene have been documented (Weng and Jackson, 1999), no field evidence (i.e. charcoal or ash) was observed to associate wildfires with these debris-flow deposits. Rather prolonged or intense rainfall (i.e. winter frontal systems or dissipating tropical storms), or extreme storms such as the 2006 event (Griffiths et al., 2009), likely provided the rainfall 73

intensity necessary to create high runoff that effectively mobilizes and transports coarse

sediments (Webb and Betancourt, 1992; McDonald et al., 2003). Regional climate

reconstructions from proxy data provide a context for assessing the timing and potential

mechanisms by which deposits from the Pleistocene-Holocene transition were generated.

Speleothem data from Cave of the Bells (COB) in the to the south

of the Santa Catalina Mountains (Wagner, 2006; Wagner et al., 2010), groundwater data from

the San Pedro River valley to the east (Pigati et al., 2009), radiocarbon ages of Lake Cochise

highstands (Waters, 1989; Kowler et al., 2012; Lyle et al., 2012), and data from regional and

local packrat midden analyses (Van Devender, 1990; Arundel, 2002; Holmgren et al., 2006)

present a consistent but imprecise story of climate change in southeastern Arizona from the Late

Wisconsin through the Holocene. Data from speleothem growth rings (Wagner, 2006) and high

groundwater levels in the San Pedro River valley (Pigati et al., 2009) indicate a cool, wet climate

from 50 ka until ~15 ka when a rapid (~100-300 year) shift to winter warming and/or drying

began, coinciding with the Bølling-Allerød warming period (Wagner, 2006). Radiocarbon dating

of highstands from Lake Cochise shows lake levels peaked at several intervals, including ~19 ka,

16.9 - 16.3 ka, 15.2 - 15.0 ka, and 14.4 - 13.8 ka (Kowler et al., 2012). These highstand intervals

are consistent with other studies of Lake Cochise (Waters, 1989) and Lake Estancia (Lyle et al.,

2012). During the Bølling-Allerød warming period interval, groundwater and lake levels dropped

(Pigati et al., 2009; Kowler et al., 2012) and forests began transitioning upslope (Van Devender,

1990; Betancourt, 2003; Holmgren et al., 2007). A cooler, wetter period, roughly coinciding with the Younger Dryas, is indicated by COB speleothem data (Wagner, 2006) and a brief recovery to higher water tables in the San Pedro River valley (Pigati et al., 2009). The speleothem data suggests a potential drought from ~11.5-10 ka, as indicated by a growth-ring hiatus, and then 74

increasing multi-decadal variability with warmer, drier winters through the Holocene (Wagner,

2006). Groundwater levels in the San Pedro River valley dropped after ~10.6 ka and have remained lower during the Holocene (Pigati et al., 2009).

While different proxy data (vegetation, speleothems, and groundwater levels) record diverse and potentially differential responses to climate change, the similarity of the timing and of changes between the proxy data show strong, regionally consistent responses to the climate change during the Pleistocene-Holocene transition that could help to explain the occurrence of large debris flows in the southern Santa Catalina Mountains. COB speleothem and packrat-midden data are in accord with an increase in extreme storm activity in the late

Pleistocene due to periods of cool North Atlantic sea-surface temperatures, which would have shifted the jet stream southward but maintained moisture flow from the Pacific (Arundel,

2002; Wagner et al., 2010; Ballenger et al., 2011). The regional climate reconstruction supports our contention that the highest magnitude debris-flow activity began sometime during the latest

Pleistocene, possibly after the abrupt warming at the beginning of the Bølling-Allerød. This is consistent with other regional studies that found aggradation of coarse-grained alluvial fans from

~14-9 ka (Harvey et al., 1999; McDonald et al., 2003). This pattern of larger, more extensive debris flows at the beginning of activity with waning size and travel distance over time supports the premise of continuous hillslope erosion since the last climate change and dwindling sediment supplies (Bardou and Jaboyedoff, 2008).

Implications of these findings, in conjunction with the analysis of the 2006 debris flows

(Webb et al., 2008b), indicate that saturation-induced debris-flow hazards in southeastern

Arizona are limited to areas within the mountains and near the mountain fronts due to limited

75

sediment supplies. Infrastructure and flood-control measures would likely be most affected by debris flows due to channel aggradation, reduction of flood conveyance and subsequent overbank deposition. One possible scenario in which areas farther from the mountain front may be at risk is if a canyon is burned by a wildfire, and then followed by a large (RI ~50-100 yrs) storm, which would significantly increase the potential size and hazard of an ensuing debris flow

(Jakob, 2005b).

A.7 Conclusions

Geologic mapping and cosmogenic 10Be dating of debris-flow deposits were used to

assess the timing and magnitude of debris flows near the mountain front of the southern Santa

Catalina Mountains north of Tucson, Arizona. Geologic mapping placed the debris-flow deposits in a relative-age frame based on topographic and landscape position, clast weathering, and soil development, whereas the 10Be cosmogenic dating, despite problems with inheritance, re-

working, and composite surfaces, provides some absolute age control. Based on 10Be

concentrations in debris-flow boulders, ages are estimated for the following map units: IYa:

14.6-11.2 ka (Finger Rock Canyon); IYb: 19.6-15.4.ka (Pima Canyon) and 15.1-9 ka (Finger

Rock Canyon); Y1: 10.1-4.5 ka (Finger Rock Canyon); and Y2: <8.6 ka (Pima Canyon).

We then used an empirical inundation model to assess potential debris-flow volumes

needed to emplace mapped IY deposits and calculated in-channel and floodplain storage within

the canyons to evaluate possible initiation mechanisms. Volumes needed to emplace the

Pleistocene-Holocene debris-flow deposits are on the order of 105 to 106 m3. Volumes available

in upstream channels and floodplains are on the order of 105 - 105.5 m3 if all channels and

floodplains were completely scoured and carried to the fan. Therefore, late Pleistocene to early

76

Holocene debris flows were likely initiated by larger landslide or other mass movement failures, although evidence of these features was not observed.

Our data indicates that debris-flow activity began in the Santa Catalina Mountains during the latest Pleistocene, possibly sometime during the Bølling-Allerød warming period, had a maximum at the Pleistocene-Holocene transition, and continued through the early Holocene. The oldest debris-flow deposits are the largest and most extensive, reflecting the magnitude of sediment available for transport as the climate changed from the Late Glacial Maximum to a more arid regime.

Acknowledgments

The Pima County Regional Flood Control District, the Arizona Geological Survey, and the U.S. Geological Survey funded this work in response to the 2006 debris flows in southern

Arizona. We thank the many home owners, home owner associations, and the Coronado National

Forest for providing access to alluvial fans. Colby Bowser, from the Office of County Supervisor

Ann Day, was especially helpful in obtaining access permissions. Numerous discussions with Joe

Cook and Mike Cline of the Arizona Geological Survey, both in the office and the field, provided the framework for mapping these deposits and understanding the past debris-flow events. Chris Magirl and Peter Griffiths of the U.S. Geological Survey, and Evan Canfield, Mark

Krieski, and Akitsu Kimoto of Pima County Regional Flood Control District, engaged in many discussions about this work and provided thoughtful reviews that substantially improved this manuscript. Pima County Flood Control provided the 1 m LiDAR data. Bodo Bookhagen provided guidance for cosmogenic sample pretreatment, Helen Raichle assisted with sample

77

preparation, and Dylan H. Rood and Robert C. Finkel facilitated the cosmogenic dating at the

Lawrence Livermore National Laboratory.

78

A.9 References

Anderson, R.S., Repka, J.L., Dick, G.S., 1996. Explicit treatment of inheritance in dating depositional surfaces using in situ 10Be and 26Al. Geology, 24(1), 47-51. Arundel, S.T., 2002. Modeling Climate Limits of Plants Found in Sonoran Desert Packrat Middens. Quaternary Research, 58(2), 112-121. Balco, G., Stone, J.O., Lifton, N.A., Dunai, T.J., 2008. A complete and easily accessible means of calculating surface exposure ages or erosion rates from 10Be and 26Al measurements. Quaternary Geochronology, 3(3), 174-195, doi: doi:10.1016/j.quageo.2007.12.001. Ballenger, J.A.M., Holliday, V.T., Kowler, A.L., Reitze, W.T., Prasciunas, M.M., Miller, D.S., Windingstad, J.D., 2011. Evidence for Younger Dryas global climate oscillation and human response in the American Southwest. Quaternary International, 242(2), 502-519, doi: 10.1016/j.quaint.2011.06.040. Banks, N.G., 1976. Reconnaissance of the quadrangle, Arizona. U.S. Geological Survey Miscellanious Field Studies Map MF-747, scale1:62,500. Bardou, E., Jaboyedoff, M., 2008. Debris flows as a factor of hillslope evolution controlled by a continuous or a pulse process? Geological Society, London, Special Publications, 296(1), 63-78, doi: 10.1144/sp296.5. Betancourt, J.L., 2003. The Pleistocene-Holocene transition in subtropical North and South America, XVI INQUA congress; Shaping the Earth; a Quaternary perspective. Congress of the International Union for Quaternary Research. Desert Research Institute, Reno, Nevada, pp. 190. Bierman, P.R., 1994. Using in situ produced cosmogenic isotopes to estimate rates of landscape evolution: A review from the geomorphic perspective. Journal of Geophysical Research, 99(B7), 13885-13896, doi: 10.1029/94JB00459. Bierman, P.R., Gillespie, A.R., Caffee, M.W., 1995. Cosmogenic Ages for Earthquake Recurrence Intervals and Debris Flow Fan Deposition, Owens Valley, California. Science, 270(5235), 447-450. Birkland, P.R., 1999. Soils and Geomorphology. Oxford University Press, New York. Birkland, P.R., Machette, M.N., Haller, K.M., 1991. Soils as a tool for applied Quaternary geology. Utah Geological and Mineral Survey Miscellaneous Publication 91-3, pp. 63. Bookhagen, B., 2007. Chemical separation of Al and Be from quartz-bearing rocks. Stanford University, Department of Geosciences, Stanford, California, pp. 48. Bull, W.B., 1991. Geomorphic Response to Climate Change. Oxford University Press, New York. Cannon, S.H., 2001. Debris-flow generation from recently burned watersheds. Environmental and Engineering Geoscience, 7(4), 321-341.

79

Cannon, S.H., Gartner, J.E., Wilson, R.C., Bowers, J.C., Laber, J.L., 2008. Storm rainfall conditions for floods and debris flows from recently burned areas in southwestern Colorado and southern California. Geomorphology, 96(3-4), 250-269. Cerling, T.E., Webb, R.H., Poreda, R.J., Rigby, A.D., Melis, T.S., 1999. Cosmogonic 3He ages and frequency of late Holocene debris flows from Prospect Canyon, Grand Canyon, USA. Geomorphology, 27(1-2), 93-111. Dickenson, W.R., 1991. Tectonic setting of faulted Tertiary strata associated with the Catalina core complex in southern Arizona. Geological Society of America Special Paper 264, Boulder, Colorado, pp. 106. Dickenson, W.R., 1999. Geologic framework of the Catalina foothills, outskirts of Tucson (Pima County, Arizona). Arizona Geological Survey Contributed Map CM-99-B, Tucson, pp. 31. Dühnforth, M., Densmore, A.L., Ivy-Ochs, S., Allen, P.A., Kubik, P.W., 2007. Timing and patterns of debris flow deposition on Shepherd and Symmes creek fans, Owens Valley, California, deduced from cosmogenic 10Be. J. Geophys. Res., 112(F3), F03S15, doi: 10.1029/2006jf000562. Eppes, M.C., McFadden, L., 2008. The influence of bedrock weathering on the response of drainage basins and associated alluvial fans to Holocene climates, San Bernardino Mountains, California, USA. The Holocene, 18(6), 895-905. Fenton, C.R., Pelletier, J.D., 2013. Cosmogenic 3He age estimates of Plio-Pleistocene alluvial- fan surfaces in the Lower Colorado River Corridor, Arizona, USA. Quaternary Research, 79(1), 86-99, doi: 10.1016/j.yqres.2012.10.006. Force, E.R., 1997. Geology and mineral resources of the Santa Catalina Mountains, southeastern Arizona. Monographs in mineral resource science; no 1. Center for Mineral Resources, Tucson. Frechette, J.D., Meyer, G.A., 2009. Holocene fire-related alluvial-fan deposition and climate in ponderosa pine and mixed-conifer forests, Sacramento Mountains, New Mexico, USA. The Holocene, 19(4), 639-651, doi: 10.1177/0959683609104031. Gile, L.H., Hawley, J.W., Grossman, R.B., 1981. Soils and geomorphology in the basin and range area of southern New Mexico - guidebook to the Desert Project. Memior 39. New Mexico Bureau of Mines and Mineral Resources, Soccoro. Gosse, J.C., Phillips, F.M., 2001. Terrestrial in situ cosmogenic nuclides: theory and application. Quaternary Science Reviews, 20(14), 1475-1560, doi: 10.1016/s0277-3791(00)00171-2. Griffiths, P.G., Magirl, C.S., Webb, R.H., Pytlak, E., Troch, P.A., Lyon, S.W., 2009. Spatial distribution and frequency of precipitation during an extreme event: July 2006 mesoscale convective complexes and floods in southeastern Arizona. Water Resourses Research, 45, W07419, doi: 10.1029/2008wr007380. Griswold, J.P., Iverson, R.M., 2008. Mobility statistics and automated hazard mapping for debris flows and rock avalanches [electronic resource], U.S Geological Survey Scientific Investigations Report 2007-5276. U.S. Geological Survey, Reston, Virginia, pp. 59. 80

Harvey, A.M., Wigand, P.E., Wells, S.G., 1999. Response of alluvial fan systems to the late Pleistocene to Holocene climatic transition: contrasts between the margins of pluvial Lakes Lahontan and Mojave, Nevada and California, USA. koCatena, 36(4), 255-281, doi: 10.1016/s0341-8162(99)00049-1. Holmgren, C.A., Betancourt, J.L., Rylander, K.A., 2006. A 36,000-yr vegetation history from the Peloncillo Mountains, southeastern Arizona, USA. Palaeogeography, Palaeoclimatology, Palaeoecology. , 240(3-4), 405-422. Holmgren, C.A., Norris, J., Betancourt, J.L., 2007. Inferences about winter temperatures and summer rains from the late Quaternary record of C-4 perennial grasses and C-3 desert shrubs in the northern Chihuahuan Desert. Journal of Quaternary Science, 22(2), 141- 161, doi: 10.1002/jqs.1023. Iverson, R.M., Schilling, S.P., Vallance, J.W., 1998. Objective delineation of lahar-inundation hazard zones. Geological Society of America Bulletin, 110(8), 972-984, doi: 10.1130/0016-7606. Jakob, M., 2005. A size classification for debris flows. Engineering Geology, 79(3-4), 151-161. Jenkins, S.E., Sieg, C.H., Anderson, D.E., Kaufman, D.S., Pearthree, P.A., 2011. Late Holocene geomorphic record of fire in ponderosa pine and mixed-conifer forests, Kendrick Mountain, northern Arizona, USA. International Journal of Wildland Fire, 20(1), 125- 141. Kean, J.W., Staley, D.M., 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. J. Geophys. Res., 116(F4), F04019, doi: 10.1029/2011jf002005. Keith, S.B., Reynolds, S.J., Damon, P.E., Shafiqullah, M., Livingston, D.E., Pushkar, P.D., 1980. Evidence for multiple intrusion and deformation within the Santa Catalina-Rincon- Tortolita crystalline complex, southeastern Arizona. In: M.D. Crittenden, Jr., P.J. Coney, G.H. Davis (Eds.), Cordilleran Metamorphic Core Complexes. Memoir 153. Geological Society of America, Memoir 153, Boulder, CO, pp. 217-267. Kirchner, J.W., Finkel, R.C., Riebe, C.S., Granger, D.E., Clayton, J.L., King, J.G., Megahan, W.F., 2001. Mountain erosion over 10 yr, 10 k.y., and 10 m.y. time scales. Geology, 29(7), 591-594, doi: 10.1130/0091-7613(2001)029<0591:meoyky>2.0.co;2. Klawon, J.E., Dickinson, W.R., Pearthree, P.A., 1999. Surficial Geology and Geologic Hazards of the Northern Tucson Basin, Pima County, Arizona, Tucson North and Sabino Canyon Quadrangles, Open-File Report 99-21. Arizona Geological Survey, Tucson, pp. 28 p. Kowler, A.L., Bright, J., Quade, J., Holliday, V.T., 2012. Latest Pleistocene paleohydrology of the Willcox Basin, southeastern Arizona, and implications for Southwest paleoclimate, Geological Society of America Abstracts with Programs, Charlotte, NC, pp. 187. Lifton, N.A., Bieber, J.W., Clem, J.M., Duldig, M.L., Evenson, P., Humble, J.E., Pyle, R., 2005. Addressing solar modulation and long-term uncertainties in scaling secondary cosmic

81

rays for in situ cosmogenic nuclide applications. Earth and Planetary Science Letters, 239, 140-161. Lyle, M., Heusser, L., Ravelo, C., Yamamoto, M., Barron, J., Diffenbaugh, N.S., Herbert, T., Andreasen, D., 2012. Out of the Tropics: The Pacific, Great Basin Lakes, and Late Pleistocene Water Cycle in the Western United States. Science, 337(6102), 1629-1633, doi: 10.1126/science.1218390. Machette, M.N., 1985. Calcic soils of the southwestern United States. In: L. Weide David, L. Faber Marianne (Eds.), Soils and Quaternary geology of the Southwestern United States. Special Paper - Geological Society of America. Geological Society of America (GSA), Boulder, CO, United States, pp. 1-21. Magirl, C.S., Griffiths, P.G., Webb, R.H., 2010. Analyzing debris flows with the statistically calibrated empirical model LAHARZ in southeastern Arizona, USA. Geomorphology, 119(1-2), 111-124, doi: DOI: 10.1016/j.geomorph.2010.02.022. Magirl, C.S., Webb, R.H., Griffiths, P.G., Schaffner, M., Shoemaker, C., Pytlak, E., Yatheendradas, S., Lyon, S.W., Troch, P.A., Dislets, S.L.E., Goodrich, D.C., Unkrich, C.L., Youberg, A., Pearthree, P.A., 2007. Impacts of recent extreme Arizona Storms. Eos, Transactions, American Geophysical Union, 88(17). Matmon, A., Nichols, K., Finkel, R., 2006. Isotopic insights into smoothening of abandoned fan surfaces, Southern California. Quaternary Research, 66(1), 109-118, doi: DOI: 10.1016/j.yqres.2006.02.010. Matmon, A., Schwartz, D.P., Finkel, R., Clemmens, S., Hanks, T., 2005. Dating offset fans along the Mojave section of the San Andreas fault using cosmogenic 26Al and 10Be. Geological Society of America Bulletin, 117(5-6), 795-807, doi: 10.1130/b25590.1. McDonald, E.V., McFadden, L.D., Wells, S.G., 2003. Regional response of alluvial fans to the Pleistocene-Holocene climatic transition, Mojave Desert, California. Special Paper 368: Paleoenvironments and paleohydrology of the Mojave and southern Great Basin deserts, 368(0), 189-205. Melton, M.A., 1965. Debris-covered hillslopes of the southern Arizona desert: consideration of their stability and sediment contribution. The Journal of Geology, 73(5), 715-729. Meyer, G.A., Wells, S.G., Jull, A.J.T., 1995. Fire and alluvial chronology in Yellowstone National Park; climatic and intrinsic controls on Holocene geomorphic processes. Geological Society of America Bulletin, 107(10), 1211-1230. Nichols, K.K., Bierman, P.R., Foniri, W.R., Gillespie, A.R., Caffee, M.W., Finkel, R., 2006. Dates and rates of arid region geomorphic processes. GSA Today, 16(8), 4-11, doi: 10.1130/GSAT01608.1. O'Brien, J.S., Julien, P.Y., Fullerton, W.T., 1993. Two-Dimensional Water Flood and Mudflow Simulation. Journal of Hydraulic Engineering, 119(2), 244-261. Pearthree, P.A., 2004. Geologic Map of the Huachuca City 7½' Quadrangle, Cochise County, Arizona, Digital Geologic Map 36. Arizona Geological Survey, Tucson.

82

Pearthree, P.A., Youberg, A., 2004. Fire and Sediment Deposition. Arizona Geology, 34(3), 1-2. Pearthree, P.A., Youberg, A., 2006. Recent debris flows and floods in southern Arizona. Arizona Geology, 36(3), 1-5. Pelletier, J.D., Engelder, T., Comeau, D., Hudson, A., Leclerc, M., Youberg, A., Diniega, S., 2009. Tectonic and structural control of fluvial channel morphology in metamorphic core complexes: The example of the Catalina-Rincon core complex, Arizona. Geosphere, 5(4), 1-22, doi: doi:10.1130/GES00221.1. Pierce, J.L., Meyer, G.A., Jull, A.J.T., 2004. Fire-induced erosion and millennial-scale climate change in northern ponderosa pine forests. Nature (London), 432(7013), 87-90. Pigati, J.S., Bright, J.E., Shanahan, T.M., Mahan, S.A., 2009. Late Pleistocene paleohydrology near the boundary of the Sonoran and Chihuahuan Deserts, southeastern Arizona, USA. Quaternary Science Reviews, 28(3-4), 286-300. Schilling, S.P., 1998. LAHARZ: GIS Programs for Automated Mapping of Lahar-Inundation Hazard Zones, U.S. Geological Survey, Reston, VA, Open-Fire Report 98-638. Spencer, J.E., 2000. Possible origin and significance of extension-parallel drainages in Arizona's metamorphic core complexes. Geological Society of America Bulletin, 112(5), 727-735, doi: 10.1130/0016-7606(2000)112<727:poasoe>2.0.co;2. 1 Spencer, J.E., Pearthree, P.A., 2004. Geologic map of the Oro Valley 7 /2' Quadrangle and the Pusch Peak area, northeastern Pima County, Arizona, Arizona Geological Survey Digital Geologic Map 21, version 2.0, Tucson. Spencer, J.E., Reynolds, S.J., 1989. Middle Tertiary of Arizona and adjacent areas. In: J.P. Jenney, S.J. Reynolds (Eds.), Geologic Evolution of Arizona. Arizona Geological Society Digest 17, Tucson, pp. 539-574. Stone, J.O., 2000. Air pressure and cosmogenic isotope production. J. Geophys. Res., 105, 23753– 23759. Swanson, F.J., 1981. Fire and geomorphic processes. In: J. Hagen, D. Johnston, T.M. Bonnicksen, N.L. Christensen, J.E. Lotan, W.A. Reiners (Eds.), Fire regimes and properties. USDA Forest Service, pp. 401-420. Tarboton, D.G., 2005. Terrain Analysis Using Digital Elevation Models (TauDEM). Utah State University, Logan. Van Devender, T.R., 1990. Late Quaternary vegetation and climate of the Sonoran Desert, United States and Mexico. In: J.L. Betancourt, T.R. Van Devender, P.S. Martin (Eds.), Packrat middens; the last 40,000 years of biotic change. Univ. Ariz. Press, Tucson, AZ, pp. 134-165. Van Devender, T.R., Toolin, L.J., Burgess, T.L., 1990. The and paleoecology of grasses in selected Sonoran Desert plan communities. In: J.L. Betancourt, T.R. Van Devender, P.S. Martin (Eds.), Packrat middens; the last 40,000 years of biotic change. Univiversity of Arizona Press, Tucson, AZ, pp. 326-349.

83

Wagner, J.D.M., 2006. Speleothem record of southern Arizona paleoclimate, 54 to 3.5 ka. Ph.D. Unpublished dissertation, University of Arizona, Tucson, 129 pp. Wagner, J.D.M., Cole, J.E., Beck, J.W., Patchett, P.J., Henderson, G.M., Barnett, H.R., 2010. Moisture variability in the southwestern United States linked to abrupt glacial climate change. Nature Geosci, 3(2), 110-113, doi: http://www.nature.com/ngeo/journal/v3/n2/suppinfo/ngeo707_S1.html. Waters, M.R., 1989. Late Quaternary lacustrine history and paleoclimatic significance of pluvial Lake Cochise, southeastern Arizona. Quaternary Research, 32(1), 1-11. Webb, R.H., Betancourt, J.L., 1992. Climatic variability and flood frequency of the Santa Cruz River, Pima County, Arizona, U. S. Geological Survey Water-Supply Paper 2379. U. S. Geological Survey, Reston, VA, United States, pp. 40. Webb, R.H., Magirl, C.S., Griffiths, P.G., Boyer, D.E., 2008. Debris flows and floods in southeastern Arizona from extreme precipitation in late July 2006: Magnitude, frequency, and sediment delivery, U.S. Geological Survey Open-File Report 2008-1274. United States Geological Survey, Tucson, Arizona, pp. 95. Webb, R.H., Melis, T.S., Griffiths, P.G., Elliott, J.G., Cerling, T.E., Poreda, R.J., Wise, T.W., Pizzuto, J.E., 1999. Lava Falls Rapid in Grand Canyon: effect of late Holocene debris flows in the Colorado River, U. S. Geological Survey Professional Paper 1591. U. S. Geological Survey, Reston, VA, United States, pp. 90. Webb, R.H., Pringle, P.T., Reneau, S.L., Rink, G.R., 1988. The 1984 Monument Creek debris flow: Implications for the formation of rapids on the Colorado River in Grand Canyon National Park. Geology, 16, 50-54. Wells, S.G., Harvey, A.M., 1987. Sedimentologic and geomorphic variations in storm-generated alluvial fans, Howgill Fells, Northwest England Geological Society of America Bulletin, 98(2), 182-198. Weng, C., Jackson, S.T., 1999. Late Glacial and Holocene vegetation history and paleoclimate of the , Arizona. Palaeogeography, Palaeoclimatology, Palaeoecology, 153, 179-201. Wohl, E.E., Pearthree, P.A., 1991. Debris flows as geomorphic agents in the Huachuca Mountains of southeastern Arizona. Geomorphology, 4(3-4), 273-292. Wolkowinsky, A.J., Granger, D.E., 2004. Early Pleistocene incision of the San Juan River, Utah, dated with 26Al and 10Be. Geology, 32(9), 749-752, doi: 10.1130/g20541.1. Youberg, A., Cline, M.L., Cook, J.P., Pearthree, P.A., Webb, R.H., 2008. Geologic mapping of debris-flow deposits in the Santa Catalina Mountains, Pima County, Arizona, Open-File Report 08-06. Arizona Geological Survey, Tucson, pp. 41 p., 11 map sheets on CD, scale 41:6000. Youberg, A., Pearthree, P.A., Baker, V.R., 2006. Comparison of Debris Flows Generated in Adjacent Unburned and Recently-Burned Areas, Coronado National Memorial, Arizona, Eos Transactions AGU, San Francisco, pp. Abstact 12661.

84

A.10 Figure and Table Captions

Table A.1. Morphometric characteristics of study basins.

Table A.2. Generalized map unit descriptions for debris-flow surfaces on alluvial fans emanating from the Santa Catalina Mountains, southern Arizona.

Table A.3. Characteristics of debris-flow boulders sampled for cosmogenic-nuclide dating.

Table A4. 10Be cosmogenic exposure ages of boulders protruding from debris-flow surfaces on alluvial fans south of the Santa Catalina Mountains, southern Arizona.

Figure A.1. Location of the study site on the western end of the Santa Catalina Mountains north of Tucson, Arizona, in eastern Pima County. The study canyons, Pima and Finger Rock, are outlined in black, the eastern canyons that had debris flows in 2006 are lightly shaded and outlined in gray. Crosses show locations of hillslope failures. Arrows indicate canyons where debris flows exited, or nearly exited, canyon mouths in 2006.

Figure A.2. Preserved prehistoric debris-flow deposits near the apex of the fan emanating from Finger Rock Canyon.

Figure A.3. Map of debris-flow deposits on the fan of Finger Rock Canyon. Samples locations for cosmogenic 10Be dating are shown with yellow dots. Descriptions of geologic units are in Table A.2.

Figure A.4. Map of debris-flow deposits on the fan of Pima Canyon. Samples locations for cosmogenic 10Be dating are shown with yellow dots. Descriptions of geologic units are in Table A.2.

Figure A.5. Cross-section of Finger Rock Wash. A. Location of cross-section in relation to cosmogenic 10Be sample locations. B. Cross section extracted from 1 m LiDAR data showing the elevational relations among mapped deposits and the active wash.

Figure A.6. Cross-section of Pima Wash. A. The location of cross-section in relation to cosmogenic 10Be sample locations. B. Cross section extracted from 1 m LiDAR data showing the elevational relations among mapped deposits and the active wash.

Figure A.7. Diagram showing cosmogenic 10Be dates for deposits assigned relative ages during surficial geologic mapping. Points grouped by boxes show samples collected from boulders within the same mapped deposit. Open squares and circles denote outliers.

Figure A.8. Volumes estimates from LAHARZ modeling, showing results using the southern Arizona coefficients. Panel A: Pima Canyon. Panel B: Finger Rock Canyon. Volume estimates are 104 m3 (purple), 105 m3 (blue) and 106 m3 (yellow). (Volume 107 m3 not shown). 85

Figure A.9. Pima Canyon shown with a 1m resolution DEM derived from airborne LiDAR with mapped debris-flow deposits on the fan, and two LAHARZ volumes, 104 and 105 m3. The yellow polygon within the canyon represents the extent of the floodplain along the main channel, and composed mostly of debris-flow deposits. Hillslope channels shown in blue are tributaries that had evidence of past debris flows. These tributary channels, the main channel and the floodplain were used to assess the potential magnitude of in-channel storage.

86

Table A.1. Morphometric characteristics of study basins. Canyon Drainage Basin Elevation Average Average Relief Ratio Area (km2) Length Range Basin Channel (km) (m) Slope (o) Gradient (%)

Finger Rock 4.2 4.5 1269 34 20 0.28

Pima 10.9 6.0 1280 33 13 0.21

87

Table A.2. Generalized map unit descriptions.

Unit Relative Age Generalized Unit Description

I Pleistocene debris- Debris-flow deposits that are the highest, most weathered deposits in the flow deposits, landscape and have clearly reddened soil (5YR) undifferentiated

IY Older debris-flow Debris-flow deposits that are spatially removed from active fluvial systems, either deposits (latest as high-banks or laterally separate from younger deposits. Clasts are slightly to Pleistocene to early moderately weathered, lightly to moderately stained by oxidation, and commonly Holocene) exhibit fracturing and splitting. Surface soil between boulders are slightly reddened (7.5YR), and in some areas clasts are partially to almost completely buried by finer deposits. Clasts exposed in disturbed deposits often have thin, discontinuous carbonate coatings. In some areas, IY deposits can be further classified as IYa- highest debris-flow deposits above the channel, and IYb - debris-flow deposits of similar age but inset 1-2 m below IY deposits.

Y1 Intermediate debris- Debris-flow deposits found above active washes but entrenched below fan flow deposits (early surfaces near the mountain front, and confined to valley bottoms along incised to middle Holocene) drainages farther out from the mountain front. Clasts generally are slightly weathered with light surface oxidation and little rock fracturing. Clasts in Y1 deposits may be slightly buried from initial deposition and subsequent abandonment, soil accumulation, or overbank deposition. Soil color varies from gray (10YR) to brown (7.5YR).

Y2 Young debris-flow Debris-flow deposits found along banks and terraces of active washes, typically 1- deposits (middle to 2 m above channel floors. Fine-grained matrix sediments are generally absent late Holocene) from Y2 deposits, leaving only clasts that appear fresh and unweathered.

88

Table A.3. Characteristics of debris-flow boulders sampled for cosmogenic-nuclide dating. Boulder- surface Dip Sample thickness height (above and dip direction a-axis b-axis c-axis (depth from surrounding ground (° down from Sample Sample Boulder length length length boulder surface) /material) horizontal in Name Number Lithology (cm) (cm) (cm) cm cm azimuth direction) Catalina Flat to FR-01 070108-01 90 60 60 4 45 Gneiss subhorizontal Catalina FR-02 070108-02 80 69 30 4 20 5° to 90° Gneiss Catalina Flat to FR-04 070108-04 150 120 50 4 60 Gneiss subhorizontal Catalina FR-05 070108-05 100 90 50 4 50 5° to 0° Gneiss Catalina Flat to FR-06 070108-06 120 60 40 4 60 Gneiss subhorizontal Catalina Flat to FR-07 070108-07 80 50 40 4 30 Gneiss subhorizontal Catalina Flat to FR-09 070108-09 120 90 80 4 100 Gneiss subhorizontal Catalina Flat to FR-10 070108-10 ------3 -- Gneiss subhorizontal Catalina FR-11 070108-11 150 80 80 6 90 5° to 180° Gneiss Catalina Flat to FR-16 070108-16 70 60 40 4 40 Gneiss subhorizontal Catalina FR-17 070108-17 110 80 60 4 90 50° to 0° Gneiss Catalina Flat to FR-18 070108-18 80 50 30 4 30 Gneiss subhorizontal Flat to P-01 070109-01 Granite 90 90 50 5 50 subhorizontal Flat to P-03 070109-03 Gneiss 150 120 80 2 80 subhorizontal Quartz vein Flat to P-04 070109-04 380 190 150 4 150 in Gneiss subhorizontal Pegmatitic Flat to P-06 070109-06 90 90 70 5 90 Granite subhorizontal Flat to P-07 070109-07 Granite 140 100 70 4 70 subhorizontal Granitic P-10 070109-10 110 80 60 5 70 40° to 80° Gneiss Granitic Flat to P-11 070109-11 150 110 50 2 60 Gneiss subhorizontal Granitic Flat to P-12 070109-12 120 80 40 6 50 Gneiss subhorizontal Pegmatitic Flat to P-14 070109-14 80 80 40 4 -- Granite subhorizontal Note: ‘-- ‘ indicates unavailable data.

89

1 Table A4. 10Be cosmogenic exposure ages. Production rate Blank Sample Name Topo- Production (Scaling scheme for Internal External Sample Elevation Corrected ± Uncertainty Exposure (Geologic Latitude Longitude graphic rate (muons) : Lal(1991) Uncertaint Uncertaint Number (MSL) 10Bee (105 at/g) 2 age (yr) 3 Unit) Shielding 1 (atoms/g/yr) / Stone(2000)) y (kyr)4, 6 y (kyr) 5, 6 (105 at/g) (atoms/g/yr) FR-01 (YI1) 070108-01 -110.9100 32.3393 966 0.995 0.247 7.74 9.744 0.041 12.3 0.1 1.1 FR-02 (YI1) 070108-027 -110.9100 32.3393 966 0.995 0.247 7.74 10.620 0.211 13.4 0.3 1.2 FR-02 (YI1) 070108-027 -110.9100 32.3393 966 0.995 0.247 7.74 10.590 0.050 13.3 0.1 1.2 FR-04 (YI2) 070108-04 -110.9099 32.3387 961 0.995 0.246 7.72 9.636 0.047 12.2 0.1 1.1 FR-05 (YI2) 070108-05 -110.9099 32.3388 959 0.995 0.246 7.71 23.040 0.081 29.3 0.1 2.6 FR-06 (YI2) 070108-06 -110.9099 32.3386 957 0.995 0.246 7.69 5.977 0.022 7.6 0.0 0.7 FR-07 (YI2) 070108-07 -110.9099 32.3386 959 0.995 0.246 7.7 29.200 0.079 37.2 0.1 3.3 FR-09 (YI2) 070108-09 -110.9093 32.3382 954 0.995 0.246 7.68 11.670 0.043 14.8 0.1 1.3 FR-10 (YI2) 070108-10 -110.9093 32.3382 954 0.995 0.246 7.68 8.004 0.299 10.1 0.4 1.0 FR-11 (YI2) 070108-11 -110.9093 32.3382 956 0.995 0.245 7.56 15.890 0.052 20.5 0.1 1.8 FR-16 (Y1) 070108-16 -110.9095 32.3381 950 0.995 0.246 7.66 3.910 0.216 5.0 0.3 0.5 FR-17 (Y1) 070108-17 -110.9095 32.3381 951 0.888 0.246 7.66 6.492 0.336 9.2 0.5 0.9 FR-18 (Y1) 070108-18 -110.9095 32.3381 950 0.996 0.246 7.66 12.720 0.513 16.2 0.7 1.6 P-01 (YI2) 070109-01 -110.9432 32.3525 912 0.998 0.243 7.46 12.910 0.432 17.0 0.6 1.6 P-01 (YI2) 070109-01 -110.9432 32.3525 912 0.998 0.242 7.4 13.650 0.044 18.0 0.1 1.6 P-03 (YI2) 070109-037 -110.9432 32.3525 912 0.998 0.243 7.46 43.700 0.829 56.7 1.1 5.1 P-03 (YI2) 070109-037 -110.9432 32.3525 912 0.998 0.243 7.59 49.150 0.121 63.9 0.2 5.6 P-04 (YI2) 070109-04 -110.9433 32.3525 913 0.998 0.243 7.47 24.280 0.075 31.8 0.1 2.8 P-06 (Y1) 070109-06 -110.9430 32.3529 910 0.998 0.242 7.39 10.090 0.048 13.3 0.1 1.2 P-07 (Y1) 070109-07 -110.9430 32.3529 911 0.998 0.243 7.52 10.440 0.030 13.6 0.0 1.2 P-10 (Y2) 070109-10 -110.9417 32.3527 917 0.943 0.243 7.48 6.285 0.282 8.6 0.4 0.8 P-11 (Y2) 070109-11 -110.9418 32.3527 915 0.998 0.243 7.47 22.830 0.962 29.8 1.3 2.9 P-12 (Y2) 070109-12 -110.9417 32.3527 916 0.998 0.242 7.36 22.800 0.312 30.2 0.4 2.7 P-14 (Y1) 070109-14 -110.9419 32.3528 917 0.998 0.243 7.48 7.224 0.354 9.4 0.5 0.9 2 3 1 Total topographic shielding includes sloping of the sample surface and shielding from topographic highs. 4 2 Blank-corrected 10Be concentrations that are standardized to KNSTD standard. Values are corrected for total topographic shielding and sample thickness. 5 Uncertainty includes the uncertainty of: (1) the sample measurement on the Lawrence Livermore National Laboratories AMS (LLNL-CAMS); (2) the process- 6 blank correction; and (3) the chemical sample preparation, which includes the uncertainty of the Be concentration in the carrier solution. Fully processed blanks 7 had an average 10Be/9Be ratio of 1.046×10-16, which was subtracted from the 10Be/9Be ratio of the samples which ranged from 6.1.10-15 - 6.70.10-12. 8 3 All exposure ages were calculated using the Lal (1991)/Stone (2000) scaling scheme in the Balco et al. (2008) CRONUS-Earth online calculator (version 2.2, 9 main calculator version 2.1, constants version 2.2.1, and muons version 1.1; http://hess.ess.washington.edu/). A density of 2.65 g/cm3 and the standard 10 atmospheric pressure-elevation relation were used in the calculations. 11 4Internal uncertainty is the ability of the lab to make the measurement. All reported uncertainties are 2σ. 12 5External uncertainty is the variation in production rate and other factors needed to complete the calculation, such as variation in latitude-longitude shielding. All 13 reported uncertainties are 2σ . 14 6 Uncertainties of exposure ages and 26Al /10Be ratios do not include the uncertainty in production rates. 15 7Replicate samples to check for consistency between samples prepared at LLNL-CAMS and at the University of Arizona. 90

A.11 Figures

Figure 1. Study location map

91

Figure A.2. Preserved prehistoric debris-flow deposits.

92

Figure A.3. Map of debris-flow deposits on the fan of Finger Rock Canyon.

93

Figure A.4. Map of debris-flow deposits on the fan of Pima Canyon.

94

Figure A.5. Cross-section of Finger Rock Wash.

95

Figure A.6. Cross-section of Pima Wash.

96

Figure A.7. Diagram showing cosmogenic 10Be dates for deposits assigned relative ages during surficial geologic mapping.

97

Figure A.8. Volumes estimates from LAHARZ modeling

98

Figure A.9. Pima Canyon with channels and floodplains used to calculate storage volumes.

99

APPENDIX B: RAINFALL INTENSITY-DURATION THRESHOLDS AND THE PREDICTIVE STRENGTHS OF THREE POST-WILDFIRE DEBRIS FLOW MODELS FOR USE IN ARIZONA

Ann Youberg*,1, Susan H. Cannon2, Jason W. Kean2, Dennis M. Staley2

1 Arizona Geological Survey, 416 W. Congress, Suite 100, Tucson, AZ 85701 USA

2 U.S. Geological Survey, Denver, CO, USA

* corresponding author tel: +1 520.770.3500; fax: +1 520-770-3505 e-mail: [email protected]

Manuscript intended for submission to Natural Hazards

2013

100

B.1 Abstract

Wildfires across the western U.S. are increasing in size and severity, while at the same

time encroachment of development into the wildland-urban interface putting more people at risk

to wildfires and their aftermaths. Teams conducting rapid assessments of post-wildfire geologic hazards need models to quickly identify areas at risks for post-fire debris flows, and forecast centers and emergency response agencies need rainfall intensity-duration thresholds that identify rainfall conditions likely to generate post-fire debris flows. While empirical methods to estimate the rainfall thresholds, probability, and volume of post-fire debris flows have been developed for specific regions of the Western U.S., it is not known how well these tools perform in other regions with different geologic and hydrologic characteristics. Here, in an effort to expand the domain where accurate post-fire debris-flow hazards assessments and forecasts can be made, we assess the predictive strengths of existing probability and volume models in three different physiographic regions of Arizona, and we define a new rainfall threshold valid for Arizona.

Specifically, we test the volume model and three probability models using data from 80 recently burned basins in the Colorado Plateau, the Central Highlands Transition Zone, and the Basin and

Range of Arizona. We show that all of the models have adequate predictive strength throughout most of the state, with the exception of lower gradient basins on the Colorado Plateau (2011

Wallow Fire) where all models performed poorly. We also show that the debris-flow volume model over-predicts in all of our study areas. Our analysis shows that the choice of a model for a hazard assessment depends strongly on the area where it will be used, and the degree to which having true outcomes outweighs the risk of failed or false alarms. We also developed objectively defined rainfall intensity-duration thresholds to identify rainfall conditions likely to generate post-wildfire debris flows in Arizona. The objectively defined I10 and I15 thresholds of 52 and 42 101

mm h-1, respectively, have the strongest predictive strengths of the five threshold models

developed, although all of the threshold models performed well. Results from this study provide

guidelines and considerations for selecting an appropriate debris-flow model, and rainfall threshold values for issuing forecasts and warnings for Arizona.

102

B.2. Introduction

Wildfire activity has increased dramatically across the western U.S., resulting in an earlier start and a longer fire season with larger, hotter, and longer-burning fires (Westerling et

al., 2006). Examples of this upward trend, which began in the late 20th century, can be found all

across the west (Grissino-Mayer and Swetnam, 2000; Littell et al., 2009). Indeed, eight of

Arizona’s ten largest historic wildfires have occurred since 2002, burning over 670,000 ha

(Southwest Coordination Center, http://gacc.nifc.gov/swcc/index.htm). While the occurrence of

widespread wildfires in American Southwest is not a new phenomenon, as tree-ring records

show regionally synchronous fires have recurred for centuries (Swetnam and Betancourt, 1998), the way in which wildfire behaves has changed dramatically. Ponderosa Pine forests that

previously burned by frequent, low intensity surface fires, and mixed conifer forests that burned

in small, patchy mosaics of severity, now often burn by high-intensity crown fires across

extensive areas frequently resulting in very large contiguous expanses of denuded forests and

damaged soils (Schoennagel et al., 2004; Williams et al., 2010). Several factors have contributed

to this change including fire suppression efforts, livestock grazing, fuels buildup, and climate

change (Westerling et al., 2003; Fule et al., 2012). Key climatic factors vary by forest type and included long-term drought, or wet year(s) followed by a dry year (Swetnam and Baisan, 1996;

Grissino-Mayer and Swetnam, 2000; Swetnam and Anderson, 2008; Littell et al., 2009).

Recently, the long-term effects of fire suppression, in conjunction with changes in climate, have begun to have a greater influence on the size and severity of wildfires (Swetnam and Baisan,

1996; Littell et al., 2009; Williams et al., 2010).

103

Concurrent with the recent increase in wildfire activity has been an increase in the

encroachment and development of the wildland-urban interface (WUI), placing people at a

greater risk not only from wildfires, but also from the aftermath of fires (Stein et al., 2013).

Following wildfires, hydrologic changes in burned watersheds occur due to loss of vegetation,

duff and litter, and changes to the soil that may include fire-induced soil-water repellency or hyper-dry soils, disintegration of soil structure, surface sealing from rainsplash impacts, new flow paths through burned out roots, and the presence of ash over soil (Coelho et al., 2004; Doerr

et al., 2006; Moody et al., 2009; Ebel et al., 2012; Stoof et al., 2013). These changes result in decreased interception and infiltration, significantly increasing runoff and erosion by orders of magnitude (DeBano, 2000; Moody and Martin, 2001; Neary et al., 2005; Moody and Ebel,

2012). The excess runoff and erosion leads to sediment-laden flood flows and debris flows. Post- wildfire sediment-laden flood flows occur more frequently than debris flows, but debris flows can be significantly more destructive than floods (Gartner et al., 2004). To further complicate matters, fire season in the Southwestern U.S. ends with the onset of monsoonal rainfall which often ultimately extinguishing wildfires. Thus there is very little time to assess post-wildfire damages to resources, evaluate potential hazards to values-at-risk, implement mitigation measures and execute emergency response plans.

Methods that quickly identify potentially hazardous areas provide more time for implementation of mitigation and emergency measures. Burned Area Emergency Response

(BAER) team hydrologists have well established methods and models for assessing potential increases in the magnitude of post-fire floods. Traditionally BAER team geologists have used evidence of past debris flows, burn severity and topographic information to qualitatively assess

104

the likelihood of post-fire debris flows (USDA Forest Service, 2010). While this is an effective and useful method for small areas, time constraints due to impending monsoon rains can make it quite difficult to adequately assess potential debris-flow hazards over large burn areas. Recently developed multivariate statistical models for predicting the probability of post-wildfire debris flows and associated debris-flow volumes are new tools available for BAER team geologists in the intermountain western U.S. (Cannon et al., 2010). One of these models, model A, was recently used in Arizona (Ruddy, 2011a, b, c) and New Mexico (Tillery et al., 2011a; Tillery et al., 2011b) to assess the probability of post-fire debris flows following several wildfires in 2011.

Although results from these assessments provided quick information to land managers and local emergency response agencies, the reliability of model results in the Southwestern US are unknown since the models were developed with data outside this region, and have not been evaluated for use in the varied physiographic provinces of the Southwestern U.S.

Modeling can provide an assessment of the likelihood and potential magnitude of post- fire debris flows, however identifying potential hazard zones is only one component of the emergency response assessment. Rainfall is the primary driver of the debris flows, thus identifying triggering rainfall intensities is critical. Debris flows are typically generated in small, steep basins from high-intensity, short-duration rainfall. In southern California, for example, rainfall characteristics were found to be the most significant factor for debris-flow generation, and the timing and the magnitude of post-wildfire debris flows were most strongly controlled by short-duration (less than 30 minutes), high-intensity rainfall (Kean et al., 2011). Consequently, real-time warning systems based on flow monitoring stations in or near the burn area do not provide sufficient time to evacuate an area. Instead, warnings with longer lead times need to be

105

based on rainfall characteristics that typically generate debris flows. These warnings can be made

by comparing expected rainfall (or measured rainfall upwind of the burn area) to empirical

rainfall intensity-duration thresholds known to cause debris flows (Caine, 1980; Coe et al., 2008;

Godt et al., 2008; Guzzetti et al., 2008; Cannon et al., 2011; Staley et al., 2012).

This paper addresses both issues by assessing the predictive strengths of three post-

wildfire debris-flow probability models and a post-fire debris-flow volume model, and by developing objectively defined rainfall intensity-duration thresholds for predicting debris-flow

triggering rainfall. To complete these tasks, we developed a database of storm rainfall

characteristics and associated post-wildfire basin flow responses in 80 recently burned basins

across Arizona. Basin flow response data is binary. For each basin and storm, the flow response

is classified as either a debris flow (1) or not a debris flow (0), which includes floods,

hyperconcentrated flows and no responses. We then assessed the predictive strengths of the post-

wildfire debris-flow and volume models by comparing model results with basin flow responses

using receiver operating characteristic (ROC) analysis methods. We also used ROC to develop

and objectively define rainfall intensity-duration thresholds above which post-wildfire debris

flows in Arizona.

B.2.1 Development of post-wildfire debris-flow models

Cannon et al. (2010) developed a set of multivariate statistical models to predict the

likelihood of post-wildfire debris flows and their potential magnitudes at selected basin outlets.

The debris-flow probability models were developed using data from 388 basins in 15 recently burned areas in Colorado, Utah, Idaho and Montana, and the debris-flow volume model was developed with data from burned basins in California, Utah and Colorado (Table B.1) (Cannon et

106

al., 2010). Basin flow responses data were collected within the first 2 years after fire from short-

duration (<1 h), low-recurrence interval (<2-10 yrs) storms (Cannon et al., 2010). The models

use a geographic information system (GIS) to extract basin morphometric parameters, soil burn

severity characteristics and soils data. The models also use average storm rainfall intensity. The

probability of debris-flow occurrence is determined by P = ex/(1+ex) with the variable ‘x’

calculated by the multivariate statistical models (Cannon et al., 2010). In this study we use

models A, B and C (Table B.1) to determine the probability of post-wildfire debris flows.

B.2.2 Development of rainfall intensity-duration threshold models

Rainfall intensity is the main driver of post-wildfire debris flows. Analyzing rainfall from storms that generate debris flows provides information about conditions that generates them. One of the most common ways to assess triggering rainfall is through the development of rainfall intensity-duration (ID) thresholds. Models describing threshold curves have the form I = αDβ

where I is rainfall intensity (mm h-1), D is storm duration (hours), and α and β are empirically derived parameters

Rainfall ID threshold curves for post-wildfire debris flows are developed using a combination of rainfall data and corresponding basin flow responses (debris flows, floods/hyperconcentrated floods, no responses) (Cannon et al., 2008; Staley et al., 2012). The goal is to produce a model that predicts rainfall conditions likely to trigger debris flows within the appropriate timeframe and with minimal false or failed alarms. False alarms occur when the threshold is set too low and rain falling at intensities above the threshold does not produce debris flows. Conversely, failed alarms occur when the threshold is set too high, and rain falling at intensities below the threshold produce debris flows. Objectively defined ID thresholds are

107

developed for various time intervals to maximize accurate warnings while minimizing false and

failed alarms (Staley et al., 2012). ID thresholds are developed by characterizing rainfall

intensities from individual storms for various time intervals (5-, 10-, 15-, 30- and 60-minutes)

and grouped by binary basin response (Staley et al., 2012). ROC methodology provides a means to objectively identify the best ID threshold for each duration interval and to assess threshold performances and effectiveness for issuing warnings (Fawcett, 2006; Staley et al., 2012).

B.3. Settings

In this study we collected post-wildfire rainfall data and basin responses during the first summer following five recent wildfires that burned areas in Arizona’s three physiographic provinces; the Colorado Plateau, the Basin and Range, and the Central Highlands Transition

Zone (Figure B.1). Each zone has distinct geology and geomorphic characteristics, and vegetation communities that reflect the topography and the climate. We document flow responses following rainfall in 80 individual basins burned between 2010 and 2012 to assess the predictive strengths of the post-wildfire debris-flow models. Ten rain gauges recorded rainfall information from 82 storms, 12 of which produced debris flows.

B.3.1 Colorado Plateau

The Colorado Plateau covers the northern one-third Arizona. It is bounded on the south by the Mogollon Rim (Figure B.1) and on the west by the in western Grand

Canyon (Nations and Stump, 1981). Elevations range from approximately 900 m in lower Grand

Canyon to 3,840 m at the top of the San Francisco Peaks near Flagstaff, with generally lower elevations towards the north. The Plateau is characterized by canyons and gently sloping mesas and tables composed of primarily horizontal sedimentary rocks (Nations and Stump, 1981).

108

Towards the southern boundary, the Colorado Plateau has isolated mountain ranges created

during Cenozoic volcanic activity (Nations and Stump, 1981).

The climate of the Colorado Plateau is arid to semi-arid, varying with elevation, with

warm to hot summers and cold winters. Mean annual temperatures range from 10-16°C at the

lower elevations to -2-6°C on the highest mountains (Prism Climate Group,

http://prism.nacse.org/normals/). Diurnal fluctuations may be as high as 16°C (Western Regional

Climate Center, http://www.wrcc.dri.edu/).The desert mesas of northeastern Arizona receive an

annual average precipitation of 100-300 mm while the higher plateaus and mountain ranges may

receive up to 1,020-1,270 mm (Prism Climate Group, http://prism.nacse.org/normals/).

Precipitation is bimodal with approximately half of the annual precipitation falling as rain during

the summer monsoon (July - September) and a second peak of precipitation during the winter

(November - March), with snowfall contributing proportional amounts in upper elevations

(Sheppard et al., 2002). Wildfire season on the Colorado Plateau is in the warm, dry foresummer,

generally May and June, and is immediately followed by summer monsoon from July to mid-

September.

Rainfall intensity is the strongest driver of post-wildfire runoff and erosion. Moody and

Martin (2009) developed rainfall regimes for the western United States tying the two-year, 30- minute rainfall intensity and the seasonal rainfall ratio with post-wildfire sediment yields. 2푦푟 30 The Colorado Plateau of �Arizona퐼 � is mainly characterized by the Arizona Medium (>20-30 mm h-

1) rainfall regime with a thin strip along the Mogollon Rim between the White Mountain of

eastern Arizona and the San Francisco Peaks near Flagstaff characterized as Arizona High (>36-

52 mm h-1) (Moody and Martin, 2009, Figure 1). These rainfall regimes do not represent actual

109

rainfall but rather regional patterns of relatively frequent storms. Although they were developed

to broadly assess post-wildfire sediment yields, they also provide a framework for assessing

rainfall thresholds and basin flow responses.

The vegetation of the Colorado Plateau varies with elevation, ranging from desert

shrublands and semi-arid grasslands and schrublands in the lower elevations to piñon-juniper

woodlands and Ponderosa Pine forests in the intermediate elevations, to mixed conifer and

spruce-fir forests with increasing elevation, and alpine tundra at the highest elevations (Colorado

Plateau – Land Use History of North America, http://cpluhna.nau.edu). Ponderosa Pines form

extensive forests along the Mogollon Rim at the southern edge of the Plateau and also on the

Kaibab Plateau north of Grand Canyon, with mixed confer and spruce-fir forests on the upper elevations of the mountains. Arizona’s two largest wildfires, the 2002 Rodeo-Chediski and 2011

Wallow Fires, burned in these forests along the Mogollon Rim spreading into the Transition

Zone. Two fires on the Colorado Plateau were included in this study, the 2010 Schultz Fire and the 2011 Wallow Fire.

B.3.1.1 2010 Schultz Fire, San Francisco Peaks, Flagstaff, AZ

The 2010 human-caused Schultz Fire burned 6,100 ha of the eastern slopes of the San

Francisco Peaks northeast of Flagstaff between June 20th and 30th, 2010 (USDA Forest Service,

2010) (Table B.2, Figs. B.2A and B.2B). The San Francisco Peaks are a Pleistocene composite

with complex interbedded volcanic rocks overlain by glacial, colluvial and alluvial

surficial deposits (Holm, 1988). The fire burned through forests of ponderosa pine and mixed

conifer across slopes of 30% to over 100%, consuming the majority of the burned area during the

first 24 hours (USDA Forest Service, 2010) The soil burn severity map shows most of the area

110

burned at moderate (27%) and high (40%) severity (Table B.2; USDA Forest Service, 2010).

Elevations within the burned area range from approximately 2,100 m at the forest boundary to

3,370 m near at the top of the burned area. Three 1-mm tipping-bucket automated local evaluation in real time (ALERT) rain gauges were installed within the burned area below the steep, upper slopes, after the fire but prior to first rainfall (Figure B.2B). Thirty-one basins were monitored to document basin flow responses from rainfall (Table B.2, Figure B.2B).

B.3.1.2 2011 Wallow Fire, east-central Arizona and western New Mexico

The massive, human-caused Wallow Fire burned 217,740 ha in the White Mountains of

east-central Arizona and western New Mexico from May 29th to July 8th, 2011(USDA Forest

Service, 2011c) (Table B.2, Figs. B.2A and B.2C). This fire currently stands as the largest in

state history (Southwest Coordination Center, http://gacc.nifc.gov/swcc/, accessed 18 February

2013) and straddles the Colorado Plateau and the Transition Zone. Winds quickly spread the fire

across much of the area burned within the first two weeks (USDA Forest Service, 2011c). The

soil burn severity map shows 14% of the area burned at moderate severity and 16% at high

severity (USDA Forest Service, 2011c). Vegetation consumed by the fire included mountain

grasslands, piñon-juniper, ponderosa pine, mixed conifer, spruce-fir and riparian vegetation

(USDA Forest Service, 2011c). The fire burned between approximately 1,570 m to 3,325 m

(Rihs, 2011) across a diverse group of rocks with dominant lithologies of mid- and late-Tertiary

sedimentary and volcanic rocks (Richard et al., 2000).

Eight basins located on the Colorado Plateau, north of Mogollon Rim, were used to

assess the post-wildfire debris-flow models (Figure B.2C). These basins are on the cusp of the

Transition Zone and the Colorado Plateau. Some put these study basins within the

111

Transition Zone, but we consider these basins within the Colorado Plateau because they are north

of the Mogollon Rim (Nations and Stump, 1981; Trapp and Reynolds, 1995), and they are lower- relief basins typical of those found all along southern part of the Colorado Plateau. Four of the basins had been part of a USDA Rocky Mountain Research Station long-term watershed study from 1962-1984 (Heede, 1985), and were re-instrumented immediately following the fire to monitor post-wildfire runoff and erosion (Wagenbrenner and Robichaud, In Review). The other four basins are adjacent to the research basins. A continually operating USGS rain gauge (15- minuted binned data) is located at the weir on the North Fork of Thomas Creek, one of the research basins (Figure B.2C).

B.3.2 Central Highlands Transition Zone

The Central Highlands Transition Zone (Transition Zone) describes a northwest- southeast trending region connecting the Colorado Plateau with the Basin and Range (Figure

B.1). From the Mogollon Rim south this region is characterized by rugged, highly faulted mountains composed of Precambrian igneous, metamorphic, sedimentary and volcanic rocks

(Nations and Stump, 1981). Elevations within the Transition Zone range from approximately 450 m at Saguaro Lake to 2,150 m at the top of the Rim.

The climate of the Transition Zone varies by elevation with hot to warm summers and cold to mild winters. Mean annual temperatures vary from 18-22°C in the lower deserts to 8-

10°C in the upper elevation forests (Prism Climate Group, http://prism.nacse.org/normals/).

Precipitation is bimodal with one peak during the summer monsoon (July - September) and a

second peak during the winter (November - March), with snowfall contributing proportional amounts in upper elevations (Sheppard et al., 2002). Mean annual precipitation varies from 200-

112

300 mm in the lower deserts to 900-1,020 mm in the upper forests (Prism Climate Group,

http://prism.nacse.org/normals/). Wildfire season in the Transition Zone is in the warm, dry

foresummer, beginning in April or May, extending to early or mid-July, depending on the start of

-1 monsoon. The rainfall regime for Transition Zone includes Arizona High (>36-52 mm h ) 2푦푟 30 in eastern two-thirds퐼 of the Transition Zone, and Arizona Medium (>20-30 mm h-1) in the western third (Moody and Martin, 2009, Figure 1).

The vegetation of the Transition Zone is very diverse ranging from desert shrubland and

Chaparral at lower elevations to piñon-juniper woodlands in mid elevations, and Ponderosa Pine

and mixed conifer forests at the upper elevations (USDA Rocky Mountain Research Station,

http://www.rmrs.nau.edu). Three of the ten largest wildfires in Arizona are located in the

Transition Zone: 2005 Cave Creek Complex (3rd), 2004 Willow Fire (5th) and 2005 Edge

Complex (7th) (Southwest Coordination Center, http://gacc.nifc.gov/swcc/index.htm). One fire in

the Transition Zone, the 2012 Gladiator Fire, is included in this study.

B.3.2.1 Gladiator Fire, Bradshaw Mountains, central Arizona

The human-caused Gladiator Fire burned 6,560 ha near Crown King in the Bradshaw

Mountains of central Arizona between May 13th and June 13th, 2012 (USDA Forest Service,

2012). The fire burned across Chaparral, Ponderosa Pine and lesser amounts of mixed conifer

forests between the elevations of 1,241 m and 2,316 m. Half of the area was burned at moderate

(32%) and high (21%) soil burn severity (Table B.2, Figs. B.2A and B.2D; USDA Forest

Service, 2012). Geologic units within the burned area include Precambrian granite, volcanic and

metamorphic rocks (Anderson and Blacet, 1972). After the fire three tipping-bucket ALERT

gauges were installed within the burned area. One was located near Peck Mine at the north end

113

of the study basins, one was located on Lincoln Ridge at the south end of the study area (Figure

B.2D), and the third was placed on the western perimeter of the fire. Seven basins located on the

eastern perimeter of the fire, along the Crown King Road, were used to assess the performance of

the debris-flow models (Figure B.2D).

B.3.3 Basin and Range

The Basin and Range province includes all of southern and western Arizona, extending

into southern New Mexico, southeastern California, southern Nevada and portions of northern

Mexico (Nations and Stump, 1981). The Basin and Range is characterized by generally

northwest-southeast trending mountain ranges separated by broad alluvial valleys. The mountain

ranges in western Arizona are generally low elevation with sparse vegetation hence wildfires are

not common in these areas. The mountain ranges in southeastern Arizona, called Sky Islands,

extend from elevations of approximately 600 m in the lower valleys to approximately 2,800 m at

the upper elevations, and extending to 3,267 m at the highest peak in southeastern Arizona, Mt.

Graham in the Pinaleño Mountains. These mountain ranges are some of the most biologically diverse areas in the United States. The geologic units that compose the Basin and Range are also extremely diverse.

The climate of the Basin and Range varies considerably depending on location. Mean annual temperatures vary from 16-22°C in the lower deserts to 4-8°C in the upper elevation forests (Prism Climate Group, http://prism.nacse.org/normals/). The Basin and Range province of southeastern Arizona is strongly influenced by the North American Monsoon. Precipitation is strongly bimodal with approximately half of the precipitation falling during the summer monsoon (July - September) and a second peak precipitation peak during the winter (November -

114

March) with snowfall contributing proportional amounts in highest elevations (Sheppard et al.,

2002). Snow is rare in the lower valleys. Mean annual precipitation varies from approximately

300-400 mm in the lower deserts to 900-1,270 mm at the highest peaks (Prism Climate Group, http://prism.nacse.org/normals/). As with the rest of Arizona, wildfires begin in the warm, dry foresummer, generally May, and extend to the start of monsoon, typically early July. The

rainfall regimes in this area reflect the influence of the monsoon. Most of southeastern 2푦푟 30 Arizona퐼 is represented by the Arizona High rainfall regime (>36-52 mm h-1), but the Huachuca

Mountains and surrounding areas are classified as Arizona Extreme (>52-100 mm h-1) (Moody and Martin, 2009, Figure 1).

Vegetation in the Basin and Range province of southeastern Arizona is very diverse. In the Sky Islands, vegetation on the lower flanks of the mountains includes Sonoran Upland thornscrub and Chihuahuan semi-desert grasslands, steppe and upland scrub vegetation (Arizona

Firescape, http://www.azfirescape.org). Middle elevation slopes have transitional oak woodlands, Madrean oak-piñon-juniper woodlands with components of manzanita and mountain mahogany, and Madrean lower montane pine-oak forests and woodlands (Arizona Firescape, http://www.azfirescape.org). Upper elevation vegetation include Madrean upper montane conifer-oak woodlands and forests and dry mixed conifer forests, with wet mixed conifer forests and spruce-fir forests at the highest elevations (Arizona Firescape, http://www.azfirescape.org).

Two wildfires from this province were included in this study: the 2011 Horseshoe 2 Fire and

2011 Monument Fire.

115

B.3.3.1 2011 Horseshoe 2 Fire, Chiricahua Mountains, southeastern Arizona

The human-caused Horseshoe 2 Fire burned 90,226 ha of the Chiricahua Mountains in southeastern Arizona between May 8th and June 25th, 2011 (USDA Forest Service, 2011a)

(Table B.2, Figs. B.2A and B.2E). This fire burned approximately 70% of the mountain range and currently stands as the 4th largest wildfire in state history (Southwest Coordination Center, http://gacc.nifc.gov/swcc/, accessed 18 February 2013). The soil burn severity map shows 12% of the Horseshoe 2 Fire burned at high severity and 30% at moderate severity (Table B.2, Fig.

B.2E; Youberg, 2011). Elevations within the burned area range from approximately 1,300 m near the base of the range to 2,975 m at . Vegetation within the fire perimeter varies from grasslands to Madrean woodlands and forest composed of oak, piñon-juniper, pine and conifers depending on elevation (Arizona Firescape, http://www.azfirescape.org). Lithologic

units include a diverse suite of Proterozoic, Paleozoic, and Cretaceous metamorphic, sedimentary

and volcanic rocks, mid-Tertiary volcanic rocks, and Quaternary basalts and surficial deposits

(Pallister and du Bray, 1992, 1994; Drewes, 1996).

One research rain gauge (timestamped 2 mm tipping bucket) was installed in the East

Turkey Creek drainage just below Buena Vista Peak, north of Rustler Park, on July 1, 2011

(Figure B.2D, north gauge). A temporary National Weather Service weather station was installed on the drainage divide between East Turkey Creek and Pine Canyon, south of Rustler Park, on

July 26, 2011 (Figure B.2D, south gauge). The weather station recorded timestamped packets of data with variable time intervals and rainfall amounts. Twenty-five basins in Pinery, Pine and

East Turkey Creek Canyons were monitored to assess the post-wildfire debris-flow models

(Figure B.2D).

116

B.3.3.2 Monument Fire, Huachuca Mountains in southeastern Arizona

The human-caused Monument Fire burned 12,353 ha in the Huachuca Mountains in

southeastern Arizona between June 12th and July 8th, 2011 (USDA Forest Service, 2011b). The

fire burned across federal and private land, covering the southern third of the Huachuca

Mountains and extending out into adjacent grasslands. Almost half of the fire area burned at high

(7%) and moderate (39%) soil burn severity, mostly on the steeper mountain slopes (USDA

Forest Service, 2011b) (Table B.2, Figs. B.2A and B.2F). The fire burned between the elevations

of 1,335 m and 2,885 m across grasslands and Madrean oak, piñon-juniper, pine and conifers

woodlands and forests (Arizona Firescape, http://www.azfirescape.org). Lithologies include

Precambrian granite on the lower flanks of the mountain, and numerous Paleozoic sedimentary

units, Mesozoic volcanic and sedimentary rocks, and Quaternary surficial deposits (Hayes and

Raup, 1968).

Two 1-mm tipping-bucket ALERT rain gauges were installed within the burned area

during the first week of July, prior to full containment. One gauge was installed at the north end

of the fire on the ridge between Carr and Miller Canyons, and the other gauge was installed at

the south end of the fire on the lower slopes of upper Ash Canyon (Figure B.2F). Ten basins

were monitored to assess the post-wildfire debris-flow models (Figure B.2F).

B.4. Methods

In this section we describe our methods for collecting basin response data and assessing rainfall characteristics. We use ROC analysis to objectively define the critical rainfall ID thresholds for generating post-wildfire debris flows, and to assess the predictive strengths of the debris-flow probability models. Our long-term goal is to develop a rainfall intensity-duration

117

threshold for each physiographic province. In this study, however, the limited number of debris-

flow producing storms allows for only one statewide threshold.

B.4.1 Data collection

During the first summer after each fire, study basins were monitored to document basin

responses after rainfall. When flows occurred, field assessments of deposits were used to

determine the occurrence of debris flows or sediment-laden flood flows only. The downstream

extent of debris flows and, where possible, intact deposits were mapped to estimate debris-flow volumes. Basins were defined by outlets that represented a break in channel slope, often delineated by a fan. Field data was supplemented with reports, ground and aerial photographs, videos and personal communication with BAER Implementation teams, local flood control and forest service personnel, and other researchers, to help determine and verify basin responses.

Raw precipitation data was downloaded and processed from ten rain gauges within the fire perimeters and used to derive rainfall parameters for debris-flow modeling, and rainfall characteristics to calculate rainfall intensity-duration thresholds. Monsoonal rainfall is characterized by short-duration, convective thunderstorms, and several storms can occur within a

24-hour period. Here, storms are separated based on a minimum 3-hour period of no rain

(Restrepo-Posada and Eagleson, 1982).

B.4.2 Data analysis

B.4.2.1 Rainfall analysis

Rainfall storm totals (mm) and average storm intensities (mm h-1) were calculated for

each storm. Smaller storms (generally only a few mm) were not used to assess the post-wildfire

118

debris-flow models. Rainfall ID thresholds were developed from all of the rainfall data first by

calculating peak rainfall intensity for durations of 5, 10, 15, 30, and 60 minutes (I5-I60) for each

storm that occurred during the first summer. These duration intervals were selected because most

of the storms had durations of an hour or less, with only a few storms lasting longer than two

hours. To calculate the peak intensities of interest, an intensity for every minute of each storm

was calculated by backwards differencing cumulative rainfall for the various durations using the

following equation:

( ( ) ( )) ( ) = 푅 푡 − 푅 푡 − 퐷 퐼퐷 푡 -1 퐷 where ID is rainfall intensity (mm h ), t is time, R is the cumulative rainfall (mm), and D

is the duration (hours) (Staley et al., 2012). Storm data from each rain gauge were then classified

with a binary parameter to define storms that produced at least debris flow (1) or no debris flows

(0, sediment-laden floods or no responses). Because monsoonal storms are cellular, rainfall

amounts can vary dramatically across the burned area. This means that for any given storm, data

from one rain gauge may be classified as a debris-flow producing storm while data from another gauge on a different part of the same burn scar is classified as a non-debris-flow producing storm. Where possible, Level III NEXRAD radar data was used to determine which basins received rainfall representative of the nearest gauge.

B.4.2.2 Post-wildfire debris-flow modeling

To assess the post-wildfire debris-flow probability and volume models, the flow response in the basins were classified for each rainfall event as either having a debris flow (1) or not having a debris flow (0). Using a GIS, basin morphometric, soil burn severity and soils data for

119

each basin were derived from 10 m digital elevation models (DEMs), final soil burn severity

maps, and STATSGO soils data. Model parameters for each basin were entered into a

spreadsheet where probability and volume were calculated. The probabilities for each basin from

each model were paired with the binary response data and analyzed using ROC methods.

B.4.2.3 Receiver operating characteristic (ROC) analysis

ROC analysis provides a method for assessing the performance of models of classified

data, either discrete (two or more categories) or continuous (probabilities), and for visualizing

their relative performances (Fawcett, 2006). This method has long been used by the signal detection and medical communities to evaluate models of classified data (presence/absence)

(Bradley, 1997; Eng, 2005; Fawcett, 2006), and has recently been adopted by the landslide community to evaluate landslide susceptibility models and to objectively define rainfall intensity-duration thresholds (Godt et al., 2008; Frattini et al., 2009; Staley et al., 2012).

ROC analysis categorizes classified data into one of four classes using a two-by-two contingency table (Figure B.3A). The columns represent the observed class, in this case whether or not a debris flow occurred, and the rows represent the modeled class above or below a selected threshold. For example, a BAER team geologist may chose a probability threshold of

≥50% to identify basins at risk of debris flows. A true positive (TP) classification represents data where the model predicts an event (i.e. a debris flow) and one or more occurs. A false positive

(FP, false alarm) represents data where the model predicts an event yet none occur. Conversely, a false negative (FN, failed alarm) is when the model fails to predict an event yet one or more occurs. Finally, the true negative (TN) is when the model does not predict an event and none occur. The goal of using ROC analysis is to 1) assess the predictive strengths of the post-fire

120

debris-flow models, and 2) to objectively define rainfall ID thresholds that maximize true

positive outcomes while minimizing failed and false alarms (Fawcett, 2006; Staley et al., 2012).

ROC graphs provide a visual assessment of how well a model predicts between positive

and negative outcomes (Fawcett, 2006; Staley et al., 2012). This is done using an ROC utility

function which describes the relationship between the true positive rate (tprate) and the false

positive rate (fprate). The tprate measures how well a model predicts the occurrence of an event, in

this case a debris flow, and the fprate measures the rate at which the model incorrectly predicts a

debris flow. Probabilities from the debris-flow modeling are combined with the binary basin flow response data, whether a debris flow occurs (1) or not (0). Then ROC utility function calculates the true and false positive rate for the entire range of possible thresholds in increments of one. The post-wildfire debris-flow model probability thresholds range from 0-100%. The rainfall ID thresholds range from 1-100 mm h-1. At the same time a third metric, the threat score

(TS) is calculated (Figure B.3A). TS is used in optimization analysis to describe overall model

performance by reducing the score equally with the occurrence of either a failed (FN) or false

(FP) alarms (Staley et al., 2012). The highest TS is then used to identify the optimal threshold. In

cases where the highest TS occurs over more than one increment, the highest rainfall intensity is

selected as the optimal rainfall ID threshold and the lowest probability is selected for the debris-

flow model thresholds. Objectively defined thresholds yield a single point that is plotted in ROC

space, providing a means to assess all of thresholds against each other. The ROC utility function

generates a curve describing fp:tp rates over the entire range of possible threshold.

Models can then be compared either by comparing points representing single-threshold values or curves that show model performances over the entire range of possible thresholds. The

121

plotted position of a single-value threshold point indicates how well a model works with a

chosen threshold. The 1:1 line in the ROC graph provides a reference for assessing the model

performances, with the line representing the outcomes of a random model (Figure B.3B, pt B).

Points plotting higher above and to the left of the 1:1 line indicate better models. A near-perfect model plots in the upper left corner of the graph (Figure B.3B, pt A), and a model that performs worse than random plots to the right and below the 1:1 line (Figure B.3B, pt E). A more conservative model has fewer false alarms but at the expense of a lower tprate (Figure B.3B, pt

D), while a more liberal model has a higher tprate at the expense of more false alarms (Figure

B.3B, pt C).

The curve described by the ROC function provides information about the predictive

strengths of the models over the entire range of potential thresholds (Fawcett, 2006; Hand,

2012). The area under each curve (AUC) is calculated and compared to each other to provide relative measures of performance between the models (Bradley, 1997; Fawcett, 2006; Hand,

2010). An AUC closer to one indicates a better performing model (Figure B.3B, Curve F is better than Curve G), while an AUC of 0.5 or less indicates a random or worse model. AUC is a global indicator of model performance over the entire range of possible thresholds (Hand, 2010).

Curve shape indicates where models perform best and must be considered along with the AUC to understand how well each model performs compared to the other (Eng, 2005). The curve of a more conservative model is highest aboe the 1:1 line along the y-axis and minimizes false alarms at the cost of more failed alarms (Figure B.3B, Curve H). a more liberal model plots highest above the 1:1 line near the top of the graph and has fewer failed alarms at the cost of more false

122

alarms (Figure B.3B, Curve F). Thus, it is instructive to look at both the shape of the curve and

AUC values to determine predictive strengths of the models (Bradley, 1997).

B.5. Results

Rainfall and basin flow response data were compared with results from three post-fire debris-flow probability models, A, B and C, and one debris-flow volume model to assess the predictive strength of each model. Data from 82 rain storms, 12 of which produced debris flows, were used to develop objectively defined rainfall ID thresholds for use in issuing warnings in

Arizona.

B.5.1 Basin responses

Basin flow responses were recorded in 80 recently burned basins during the first summer following five wildfires. Basins outlets were selected based on the locations of the deposits which were due to a change in slope as the channel 1) transitioned from erosional to depositional processes as it exited the mouth of a canyon or mountain front, 2) crossed at a road bed, or 3) entered a larger channel at a confluence angle high enough to constrict flow and force deposition.

As a result, some smaller modeled basins on the Schultz and Horseshoe 2 Fires with outlets at roads or at tributary junctions, are nested within larger basins that are defined by a transition from erosional to depositional domains. Basin size was restricted to a minimum of at least 0.1 km2 (the minimum size of basins used to develop the models) and a maximum of 13.5 km2,

which was well within the range of the original model development dataset (Cannon et al.,

2010).

123

B.5.1.1 Colorado Plateau

Thirty-one basins were monitored within the Schultz Fire burned area, ten of which are

nested (Table B.3, Figure B.4B). Basins varied in size from 0.1-1.8 km2, with many of the basins

mostly, or completely, burned at moderate and high soil burn severity (Tables B.2 and B.3,

Figure B.2B). There were two debris-flow producing storms. A storm on July 20th, 2010, produced debris flows in 23 basins and another storm on August 16th in 17 basins (Table B.4,

Figure B.4B). Debris-flow volumes were estimated for 20 of the basins based on detailed field

mapping that documented the locations of the sediment-laden flood and debris-flow deposits.

Debris-flow deposits were rarely debris-flow only as most were modified by subsequent flood flows. For this reason, debris-flow volumes are only estimated to an order of magnitude ().

Within the Wallow Fire burned area, eight basins were selected to assess the models

(Table B.3, Figure B.2C). Due to the size and remoteness of the fire, real-time monitoring was not possible in this study. Instead, the four Rocky Mountain Research Station watersheds were selected to assess the models, along with four adjacent basins (Figs. B.2C and B.4C). On July

11th, 2011, the largest rainfall of the summer occurred (Table B.4). The four research watersheds

did not have debris flows, but an adjacent unnamed tributary to Thomas Creek did. The

responses and a debris-flow volume estimate from these basins were provided by the Rocky

Mountain Research Station (P. Robichaud, personal communication, April 2012). The other

three basins (two debris flow and one sediment-laden flood) were assessed later with verification

from the Apache-Sitgreaves National Forest watershed group (Figure B.4C, Tom Subirge,

personal communication, May 2012). While numerous other basins within the Wallow Fire had

debris flows, the data were not sufficient to include any but these watersheds.

124

B.5.1.2 Central Highlands Transition Zone

Seven basins within the Gladiator Fire burned area, along the Crown King Road, were

assessed for basin responses (Table B.4, Figure B.4D). Rain first fell on the burned area on July

13th, 2012, followed by two storms on July 14th. The second July 14th storm (Table B.4) triggered

debris flows in three basins (Figure B.4D) at approximately 17:30 (Denny Faulk, Yavapai

County Emergency Manager, personal communication, 2012). Basin responses along the Crown

King Road, on the east side of the burned area, were initially assessed and described by Kevin

Weldon, Prescott National Forest geologist, and Mark Massis, Yavapai County Flood Control

District hydrologist, and later field verified (Table B.3, Figure B.4D). Access restrictions

prevented data collection on the west side of the burned area.

B.5.1.3 Basin and Range

The Horseshoe 2 Fire burned approximately 70% of the entire Chiricahua Mountain

range. Basins in this study were limited to accessible canyons in the north central part of the burn

area (Figure B.2E). Thirty-seven basins were originally assessed for flow responses, however,

due to limited precipitation data only 25 of the 37 basins were used to assess the debris-flow models (Tables B.3, Figure B.4E). Nine basins are nested due to flow deposits located at roads or at tributary junctions, and the 10th nested basin was part of a short-term post-wildfire hillslope

and channel erosion study (DeLong et al., 2012). The first storm that caused significant erosion

and debris flows occurred on July 11th, 2011 (Table B.4). Debris flows were documented in 11 of the 25 basins (Figure B.4E). Additional debris flows occurred in six of the monitored basins on

August 15th, and in nearby basins on August 11th and 13th, 2011. These were initially

125

documented by the BAER Implementation team and later verified by field evaluations. Debris-

flow volumes were mapped in four basins.

Ten basins were monitored within the Monument Fire burned area. Five of the basins had

at least one debris flow and five had only sediment-laden flood flows (Fig, B.4F, Table B.3). The

first significant storm occurred July 10th, 2011 (Table B.4), prior to containment, in the northern

part of the burn scar. Debris flows occurred in three northern basins with significant flooding and

erosion in other channels (Figure B.4F). No rain was received on the south end of the fire on July

10th. Debris flows occurred in a southwestern basin on August 11th, and in an eastern basin on

August 13th (Table B.4, Figure B.4F).

B.5.2 Rainfall characteristics and objectively defined intensity-duration thresholds

Rainfall data from the first monsoon after each fire was collected from a total of eight

tipping-bucket rain gauges, one USGS 15-minute binned gauge and one National Weather

Service temporary weather station with data stored in variable time and rainfall amounts. These

data were analyzed for storm duration, total storm precipitation, average storm intensity, and

peak intensities (I5-I60). A total of 82 storms were assessed with 12 of those storms triggering

debris flows (Table B.4). Total rainfall and average storm intensity data for each debris-flow producing storm and for the largest non-debris-flow producing storms (minimum 3) from each fire were parameters entered into the post-wildfire debris-flow probability models.

The rainfall data were also combined with the binary basin response data to develop ROC curves and objectively define rainfall ID thresholds (OBJ_ID). A comparison of peak intensities for various durations of debris-flow triggering storms and non-debris-flow producing storms show that while the populations overlap, the means of the two populations are significantly 126

different (Figure B.5). These population differences provide the opportunity to develop

objectively defined rainfall intensity-duration thresholds using ROC analysis and threat scores.

The ROC function calculated the true and false positive rates by 1-mm h-1 increments for each duration interval. Threat scores were also calculated by 1-mm h-1 increments for each

duration interval to objectively define the rainfall ID thresholds. The single-point value of each

individual OBJ_ID threshold was plotted in ROC space to compare the models against each

other (Figure B.6A). TS values range from 0.41-0.49, and objectively defined thresholds range

-1 -1 from a high of 78 mm h for I5 to a low of 22 mm h for I60 (Figure B.6B-F). ROC curves and

AUC values were also calculated for each duration interval to assess model performances. AUC

values range from 0.86 – 0.90 (Figure B.6B-F). This analysis provides one objectively defined

ID threshold for each duration interval for the entire state. The long-term goal is to have thresholds for each physiographic province, but there was insufficient data in this study to accomplish that task.

B.5.3 Post-fire debris-flow probability and volume models assessments

To assess the predictive strength of the three post-wildfire debris-flow probability models, rainfall data were entered into the models, along with GIS-derived basin morphologic parameters, soil burn severity characteristics and soils data. Each model produced a range of debris-flow probabilities per storm per fire (ntotal = 487). For each fire, all debris-flow producing

storms were used to test model performances (Table B.4), along with three to five of the largest

rainfall events that produced significant runoff. The models were assessed by individual fires

(Figure B.7A-E).

127

B.5.3.1 Colorado Plateau

Assessments of model performance with data from the Colorado Plateau shows mixed results. Schultz Fire data show curves from all three models describe similar arcs along the y- axis and the top of the graph, with AUC numbers ranging from 0.87-0.93 (Figure B.7A, Table

B.5). Curves with data from the Wallow Fire, however, show a wide range of results with model

A tracking along the y-axis in the conservative range and then dropping to the 1:1 line in the liberal region. The curve of model C, however, tracks along the 1:1 line and the curve for model

B falls below the 1:1 line (Figure B.7B). AUC numbers range from 0.71 for model A to 0.55 for model C and 0.38 for model B (Figure B.7B, Table B.5).

B.5.3.2 Central Highlands Transition Zone

ROC curves with data from the Gladiator Fire track above the 1:1 line across the entire range of the ROC space (Figure B.7C). All of the models perform similarly in the conservative space but model C also performs well in the liberal space which is reflected in the AUC numbers. Model C AUC is 0.77 while models A and B are 0.74 and 0.73, respectively (Table

B.5).

B.5.3.3 Basin and Range

Model curves from the Horseshoe 2 Fire track above the 1:1 line in the liberal and moderate regions of the graph, but drop down to the x-axis in the conservative region (Figure

B.7D). AUC numbers range from 0.64-0.67 with models A and B performing better than C

(Figure B.7D, Table B.5). ROC curves with data from the Monument Fire track just above to moderately above the 1:1 line for all models, rising highest above the line towards the more

128

liberal region of the graph (Figure B.7E). AUC numbers range from 0.61-0.63 with model C

slightly outperforming A and B (Table B.5).

Debris-flow deposits were mapped in 26 basins, 21 basins on the Colorado Plateau

(Figure B.8, red circles) and five basin in the Basin and Range (Figure B.8, black squares).

Debris-flow deposits in most basins had been removed or significantly altered before field measurements could be made. Estimates of measured volumes were compared with volumes predicted by the model. Modeled debris-flow volumes consistently overestimated measured volumes, although most measured volumes fell within two standard errors of the predicted values.

B.6 Discussion

B.6.1 Rainfall intensity-duration thresholds

Objectively defined rainfall intensity-duration thresholds were developed for Arizona using rainfall data from five burned areas. The long-term goal is to develop rainfall intensity- duration thresholds for each physiographic province, however due to the limited number of debris-flow producing storms in this study we have currently developed one statewide threshold.

ID thresholds are well suited to post-wildfire debris flows because these debris flows are most often triggered by runoff, not saturation, and therefore antecedent soil moisture conditions are not critical to the initiation of post-wildfire debris flows. Optimally, the objectively defined

ID threshold is based on the rainfall intensity at the time of debris-flow initiation (Kean et al.,

2011; Staley et al., 2012). Since actual timing of debris-flow initiation is unknown in this study,

129

we used peak rainfall intensity for the durations of interest as the best proxy to describe the

conditions triggering debris flows.

Threat scores were used to objectively define ID thresholds from the generated ROC

curves, and AUC numbers were calculated for each curve (Figure B.6A-F). All of the objectively defined threshold values plot high and to the left of the 1:1 line in a general cluster, but the I10

and I15 models outperform the other models (Figure B.7A). This is also reflected in the AUC

values of the ROC curves (Figs. 7B-7F). These results show that, in general, all the models have

strong predictive strengths, maximizing high true positive rates while minimizing the false

positive rates, although the I10 and I15 models are the best performers. These findings are consists

with those of Staley et al. (2012).

Rainfall intensity-duration curves describe threshold intensities for triggering landslides

and debris flows and are typically defined as a power law equation as opposed to the individual

thresholds developed here. Threshold curves are most often developed for saturation-induced landslides and debris flows (Caine, 1980; Cannon and Ellen, 1985; Montgomery et al., 2000;

Godt et al., 2006; Godt and Coe, 2007; Coe et al., 2008; Guzzetti et al., 2008; Webb et al.,

2008b; Baum and Godt, 2010), although recently they have also been developed for runoff- induced debris flows (Cannon et al., 2008; Coe et al., 2008; Cannon et al., 2011; Staley et al.,

2012). Threshold curves for saturation-induced debris flows are generally developed using the

lowest rainfall intensity that generates debris flows (Caine, 1980; Godt et al., 2006). These

thresholds are set low and thus have a high false alarm rate. A second method uses the highest

rainfall intensities that did not produce debris flows to develop threshold curves (Cannon et al.,

2008). These thresholds are set high and can have high failed alarm rates. Staley et al. (2012)

130

showed that objectively defined ID thresholds based on the threat score provided better

indicators of conditions for generating debris flows while minimizing failed and false alarms.

Using the objectively defined rainfall ID thresholds developed in this study, a power-law threshold curve for post-wildfire debris flows in Arizona is described with the following equation:

I = 22D-0.50

Figure 9 shows a comparison of this threshold with both a worldwide threshold and

numerous regional thresholds (Figure B.9). Caine’s (1980) worldwide curve is shown by the

black dotted line (Figure B.9). Solid lines represent curves from non-fire-related debris flows and landslides, while the dashed lines represent fire-related debris flows, with colors representing different areas in the western U.S. (Figure B.9). Most of the non-fire-related debris flows were generated from shallow, saturation-induced slope failures. The exceptions to this are some debris flows documented in Colorado (Godt and Coe, 2007; Coe et al., 2008). The sparsely vegetated landscape of the Chalk Cliffs, derived from highly fractured and hydrothermally altered quartz monzonite (Coe et al., 2008) respond to rainfall similarly to recently burned areas. Because of sediment availability at Chalk Cliffs three to four runoff-induced debris flows are produced, on average, every year (Coe et al., 2008). Since these are runoff-induced debris flows it is not surprising that the curve for the Chalk Cliffs are similar to the curve for the nearby Coal Seam

Fire.

The threshold curve developed with data from this study plots high and to the left of the graph (Figure B.9, rad dashed line) indicating short-duration, high-intensity rainfall triggers post-

131

wildfire debris flows in Arizona. A threshold curve for southern Arizona was developed in 2006

(Webb et al., 2008b) following a week-long storm that culminated in hundreds of saturation- induced debris flows in four mountain ranges (Figure B.9, solid red line). A comparison between these threshold curves highlights the fact that post-wildfire debris flows are generated from shorter-duration storms, but with similar or higher rainfall intensity, reflecting an overall lower total rainfall required for post-wildfire debris flows compared to saturation-induced debris flows.

Return intervals (RIs) of the storms that triggered debris flows in this study were

generally five years or less with two exceptions, A July 20th, 2010, storm on the Schultz Fire had

an RI of 20 years, and the July 14th, 2012, storm on the Gladiator Fire had an RI of 10 years

(Table B.4). All of the other storms had RIs of five years or less. These RIs are similar to those

documented from storms in other burned areas (Cannon et al., 2008; Kean et al., 2011;

Friedman, 2012) and are within the range of storms used to develop the models with the one

exception of the 20-year storm. Comparing our OBJ_ID threshold values to the NOAA Online

Atlas 14 intensity data indicate that these thresholds are equivalent to approximately a one-year

return-interval storm over the Schultz, Wallow and Gladiator burn scars, and much less than a

one-year storm over the Horseshoe 2 and Monument burn areas. These results reflect the 2푦푟 30 rainfall regimes. This is not surprising since the rainfall regimes reflect regional patterns, 퐼but

these also indicate that it might be appropriate to extend the Arizona Extreme regime over more

of southeastern Arizona, where the monsoon is strongest. Additional data will be needed to

further define these relationships.

132

B.6.1.1 Practical guidelines for using the OBJ_ID threshold models

These objectively defined rainfall intensity-duration thresholds are appropriate for forecasting warnings as storms approach areas recently burned by wildfires. Forecasting centers can use these thresholds, and in particular the I10 and I15 thresholds, to compare with radar

rainfall rates as storms move towards a burned area. Assessing radar rainfall intensities using

these thresholds will likely provide sufficient warning times for evacuations or other emergency

response measures. For people living in the WUI, within or below a recently burned area, these

thresholds provide guidelines that can be compared with real-time rainfall recorded and broadcasts by the ALERT gauges. In these instances, homeowners will want to convert the

-1 normalized I10 intensity of 52 mm h to 9 mm (0.34”) in 10 minutes, and use that as their

warning threshold.

B.6.2 Post-fire debris-flow probability models

The predictive strengths of three multivariate statistical debris-flow probability models,

models A, B and C, and one debris-flow volume model were assessed for use in Arizona. Basin

flow responses from 80 recently burned watersheds and rainfall data from 82 storms were used

to assess the models. ROC analyses were used to develop the data for assessing the models. The

shapes of the ROC curves and the AUC numbers provide information about the overall

predictive strengths of the models.

B.6.2.1 Colorado Plateau

All three models were very strong predictors with data from the Schultz Fire (Figure

B.7A). The curves describe near perfect models with high AUC numbers (Table B.5). Model A

slightly outperforms models B and C, and perform best in the conservative range. Results of the

133

models with data from the Wallow Fire, on the other hand, indicate model A is moderately predictive in the conservative range, but models B and C are completely unsuccessful. The curve and the AUC of model C indicate the results of a random model (Figure B.7B, Table B.5).

Model B, which tracks below the 1:1 line and has an AUC number of 0.38, indicates a worse than random model. It is not clear why the models perform so poorly in this area. Potential explanations may be due to basin gradients, burn severity and/or soils characteristics. Data on

Table B.3 shows that the study basins from the Schultz and Wallow Fires are comparable in size, but differ in all other basin measurements. The Wallow basins have much lower relief, lower overall soil burn severity and higher clay, organic matter content and liquid limits (Table B.3).

Visually the two sets of basins are very different. The Schultz Fire basins are on the steep upper slopes of the San Francisco Peaks and may be quite similar to basins from the Missionary Ridge

Fire near Durango, Colorado. These basins were part of the model development dataset. The

Wallow basins, on the other hand, are lower gradient watersheds similar to other basins formed on the more gently sloping Colorado Plateau surface. This indicates that data collected from future fires on the Colorado Plateau should be assessed by separate sub-regions: lower gradient basins on the Colorado Plateau versus higher gradient basins on the volcanic mountain ranges.

Clearly this is an area where addition research is needed.

B.6.2.2 Central Highlands Transition Zone

ROC curves and AUC numbers with data from the Gladiator Fire indicate that the models are moderately strong to strong predictors of post-fire debris flows, with all models performing better in the conservative and moderate regions and model C outperforming A and B in the liberal region (Figure B.7C). Data from the Transition Zone is currently limited to this fire with

134

only seven basins assessed. Three of the ten largest wildfires in Arizona, however, have burned in the Transition Zone and other studies have documented post-wildfire debris flows in this province (Pearthree and Youberg, 2004; Youberg, 2008). Data from new fires are needed to better define the predictive strengths of each model in this region. The Transition Zone is a very diverse region and, with enough data, may eventually need to be separated into sub-regions.

B.6.2.3 Basin and Range

Curves and AUC number from the Horseshoe 2 Fire indicate moderately predictive models with all of the models performing similarly. The shapes of the curves indicate that none of the models perform well in the conservative range. All of the models perform similarly in the moderate and liberal regions, and are best in the moderate region when true outcomes are balances with false alarms (Figure B.7D). Models A and B slightly outperformed model C.

ROC curves with data from the Monument Fire track just above the 1:1 line in the conservative region, rising higher in the liberal region (Figure B.7E). In the liberal region models

B and C appear to outperform A. AUC numbers indicate weakly to moderately predictive models

(Table B.5), with model C outperforming A and B.

These two sets of curves from nearby but different mountain ranges show the models performed similarly with a stronger performance with data from the Horseshoe 2 Fire. One potential explanation for this is that there were more than double the number of basins monitored in the much larger Horseshoe 2 Fire than the Monument Fire (Tables B.2 and B.3). Post-wildfire debris flows have been documented in other Sky Islands (Youberg, unpublished data). As the

Sky Islands continue to burn, additional basin responses and rainfall data will help clarify which models are best. Another explanation, however, is that this regions is strongly influenced by the 135

North American Monsoon and rainfall intensities for similar storms (same RI) are higher here

than they are to the north. The models were developed with data from much farther north and

may not be properly calibrated to rainfall in this region.

B.6.2.4 Practical guidelines for using the models

Using data from Arizona, this comprehensive comparison of the predictive strengths of

three multivariate statistical post-wildfire debris-flow models shows complicated outcomes.

While the ROC curves allow comparisons between models, a BAER team geologist needs to

know which model to choose. To explore model performances further, we plotted single-value

probability thresholds for each model and each fire using probability thresholds commonly used

by BAER teams; a 50% probability (Figure B.10A) and a slightly more conservative 60%

probability (Figure B.10B). These probability thresholds are how BAER teams identify basins at

risk for post-fire debris flows that need to be assessed in the field.

These plots were created by selecting the probability thresholds (50% or 60%) and then

finding the fprate and tprate for each model on each fire. Results from the Schultz Fire (green squares) show all of the models work very well using either probability threshold. These results

coincide with the ROC curves. Results from the Wallow Fire (orange squares) show that none of

the models work using either probability threshold, contrary to the ROC curves which showed

that model A worked in the conservative region (Figure B.7B). Information in Table B.5

explains this discrepancy. As with the rainfall intensity-duration thresholds, we used threat scores to objectively define probability thresholds for each debris-flow model and fire. Although the AUC number for model A is 0.71, the objectively defined probability threshold is 7%.

Objectively defined probability thresholds for Wallow Fire models B and C were 4% and 8%,

136

respectively. These thresholds are too low to be useful. By comparison, objectively defined

thresholds for the Schultz Fire models are 66% for models A and B, and 45% for model C. This

is reflected in the plotting positions in the 50% (Figure B.10A) and 60% (Figure B.10B)

probability graphs.

Using the 50% probability threshold, results from the Gladiator Fire (Figure B.10A,

purple triangles) show that all of the models correctly predicted the debris-flow basins (high

tprate) but also had high false alarms (high fprate). By increasing the probability threshold to 60%

(Figure B.10B) models B and C provide a better balance between the fprate and tprate. Objectively

defined probability thresholds using TS indicate very high thresholds for all models (Table B.5).

Model C had the highest AUC number (0.77) and the lowest objectively defined probability

threshold (73%), while models A and B had lower AUC numbers and extremely high objectively

defined probability thresholds (95% and 87%, respectively).

Results from the Horseshoe 2 (blue diamonds) Fire using the 50% probability shows

model A plotting very high and to the right indicating high correct predictions at the expense of

many false alarms (Figs. 10A). Model C on the other hand has few false alarm and few correct

predictions. Model B is balances the fprate and tprate and is a moderately strong predictor of post-

fire debris flows. When using the more conservative 60% probability threshold, model C no

longer works, model B has a slightly reduced tprate and a significantly reduced fprate, and model A

has dropped down but still has a high fprate (Fig 10B). The objectively defined probability

thresholds identify 59% for model B, 41% for model C, and 85% for model A (Table B.5).

Using the 50% probability threshold, results from the Monument Fire (red diamonds)

show similar, but weaker, results to the Horseshoe 2 data. All of the models plot at a consist 137

height above the 1:1 line but in different regions of the graph (Figure B.10A) Models B plots in the moderate region with model B outperforming model C. Model A plots high and to the right, again indicating high correct predictions of debris flow at the expense of many false alarms.

When using the more conservative 60% probability threshold (Figure B.10B), model C no longer works, plotting almost on top of the 1:1 line. Models A and B still provide reasonable but reduced performances. The objectively defined thresholds for models within this burned area are

89% for model A, 39% for model B and 55% for model C (Table B.5).

This last analysis shows there are no straightforward choices for BAER team geologists assessing the potential risks of post-fire debris flows. Some very clear guidelines, however, can be derived from these complicated results.

Recommendations the Colorado Plateau: If the burned area being assessed in located on one of the isolated volcanic mountain ranges with steeper gradient basins, any of the models can be used to assess the probability of post-wildfire debris flows. Based on the findings here, however, model A may be the best choice. For burned areas on the lower gradient Colorado

Plateau, none of these models are appropriate for assessing the likelihood of post-fire debris flows and BAER teams should use field assessments and geologic knowledge of the area to identify basins at risk for post-fire debris flows. Future work is required to develop models appropriate for assessing these lower gradient basins.

Recommendations for the Transition Zone: Model C is the best choice, in conjunction with a more conservative probability threshold. The threshold should be at least 60% and it could be higher as the objectively defined probability threshold is 73%. Basins identified with these thresholds can then be the focus of fieldwork. A note of caution is required, as the data used to 138

make these recommendations is limited. Additional data is needed in this region to better understand how the models are working.

Recommendations for the Basin and Range: Recommendations here are a bit more complicated. In both cases, model A over-predicted debris-flow basins (high false alarms) and the objectively defined probability thresholds were quite high at 85% in the Horseshoe 2 Fire and

89% in the Monument Fire. Model B was the most moderate performer, balancing true outcomes with false alarms, with objectively defined probability threshold of 59% in the Horseshoe 2 Fire and 39% in the Monument Fire. Model C performed more conservatively but had the highest

AUC in the Monument Fire. Objectively defined probability thresholds were 41% in the

Horseshoe 2 Fire and 55% in the Monument Fire. Recommendations for the Basin and Range include three options. One option is to use model B, the moderate performer, with a lower probability threshold, 50% at the highest. Another option is to use model A with a very high threshold, 80% at least. A third option is to use model C with a lower threshold, possibly as low as 40%. The last option is to run all three models with the different appropriate thresholds and then assess the results to see if there if some basins are consistently identified as being at risk.

Those basins can be the focus of additional field work.

B.6.3 Debris-flow volume model

The debris-flow volume model consistently over predicted the measured volumes. There are several possible explanations. First, it may that the field measured volumes are incorrect.

This is extremely likely. Debris-flow deposits are re-worked very quickly after deposition in natural environments. Some of the data for the models were derived from debris-flow basins in southern California. Material in debris-flow basins are not transported away until physically

139

removed, usually by heavy equipment. Thus, there may be a discrepancy between debris-flow volumes measured from basins and those measured in natural deposits. Another possibility is that basins in Arizona are more supply-limited than the areas used to develop the models.

Determining the amount of sediment available for transport is a notoriously difficult parameter to quantify. Both measurement error and sediment availability probably contributed to the difference between the predicted and measured volumes. In general, the model can be used an approximation of potential maximum volumes from basins at risk for post-wildfire debris flows in Arizona until new models can be developed.

B.7 Guidelines for future studies

The trend of increasingly large and severe wildfires is expected to continue for the foreseeable future. With continued encroachment of development into the wildland-urban interface, the need for models that accurately predict the probability of postwildfire debris flows will also continue to increase. Recent research on runoff-induced debris flows have significantly improved our understanding of the processes generating these flows (Kean and Staley, 2011;

McCoy et al., 2012; Staley et al., 2012; Kean et al., 2013). Additional research regarding rainfall patterns, contributing area, hydrologic connectivity, sediment availability, and the role of geomorphic properties in postwildfire runoff and erosion (Moody et al., 2013) will contribute to the body of knowledge and towards the goal of developing physically based models. Based on recent research, and topics discussed at the 2013 AGU Chapman Conference, Synthesizing

Empirical Results to Improve Predictions of Post-wildfire Runoff and Erosion, I briefly outline three areas for future research to contribute towards the body of knowledge in this area.

140

B.7.1 Rainfall analyses

Rainfall intensity is one of the most important factors in the generation of postwildfire debris flows. In a study comparing timing of debris-flow surges with peak storm intensity and

peak triggering intensity Staley et al. (2012) found, in all cases, that peak triggering intensities

were lower than peak storm intensities and debris-flows surges occurred before the peak storm

intensity. Unfortunately most burned basins aren’t, and won’t be, instrumented. Currently the

best approximation we have for debris-flow triggering intensities are peak intensities for various durations (this study). An exploration of within-storm precipitation patterns may provide new

data to help inform models predicting runoff, erosion and debris flows.

Precipitation is both spatially and temporally variable, especially in cellular

thunderstorms. The pattern of rainfall influences infiltration, time-to-ponding, runoff and erosion

(Moody et al., 2013). Raw precipitation data from tipping-bucket rain gauges in this study could be used to develop rainfall profiles. The goal of this work would be to develop storm profiles for individual storms, and also grouped storms when appropriate (more than one storm in a 24 hour period) that could then be analyzed for differences of patterns between debris-flow and non- debris-flow producing storms. Because gauges are a point measurement, Level III NEXRAD data could also be incorporated to assess how well the rainfall, and hence the pattern, is reflected across study basins with different flow responses (NEXRAD analysis is complete for some storms). The goal of this work would be to tie data from the rainfall profiles into infiltration, time-to-ponding, runoff and erosion through existing equations and models.

141

B.7.2 Debris-flow probability models

The multivariate, statistical debris-flow probability models assessed in this study describe

basin morphometrics, soil burn severity and soils characteristics at the basin scale. Assessing

hazards at the basin scale makes sense since these analyses typically focus on values-at-risk

(VARs – building, infrastructure, streams with endangered species) lower in the basin. Based on

field evidence, however, postwildfire debris flows are generated high in basins on upper

hillslopes and channels (Schmidt et al., 2011; McCoy et al., 2012). Recent research in burned areas of California and at Chalk Cliff in Colorado show these runoff-induced debris flows are generated through a series of pulsed sediment deposition in the upper channel which becomes destabilized through progressive deposition and finally breached, initiating a debris flow (Kean et al., 2013). While hazard assessments need to evaluate entire basins, debris flows are being initiated in only a portion of basins. The question then becomes what is the most appropriate scale for describing these upper basins processes and how can that information be translated into model parameters for a physically based debris-flow probability model?

One way to address this question is to explore contributing areas and hydrologic connectivity using high-resolution photographs and digital elevation models. The USFS Region

3 (R3 - Arizona and New Mexico) Photogrammetry Program began collecting high-resolution

(23-cm) digital, stereo, aerial photographs over all of the forests within R3 in 2011. Coverage currently includes the 2011 Wallow, Horseshoe 2 and Monument Fires and the 2012 Gladiator

Fire. One-meter scale DEMs are being generated from these stereo photographs. The Coronado

National Forest collected stereo, aerial photographs (contact prints, ~1:6000 at the mountain crest) over the Horseshoe 2 Fire in the Chiricahua Mountains. One set of photos were taken prior

142

to the first erosive storm (east side of the range including the ridge crest) and just after (west side

of the range including the ridge crest). The second set of contact prints were taken in the spring

of 2012.

Rilling and gully heads are clear in all of these photos. Using the photo-derived, 1m

DEMs along with all of the visual photos hydrologic connectivity and variable contributing areas

could be assessed. The 1m DEMs could then be upscaled to 10m DEMs and the same

information extracted for a comparison. Results from this study would help determine if variable contributing area and hydrologic connectivity can be sufficiently described in physically based models using 10m DEMs.

B.7.2 Debris-flow volume models

The debris-flow volume models used in this study consistently over-predicted volumes

from debris basins. The availability of channel sediment for transport and wetness of the channel

bed material are important factors determining the volume of a debris flows (McCoy et al., 2010;

Moody et al., 2013). While studies have shown that these smaller storms don’t have a lasting

impact on the hillslope moisture levels, it may be that smaller storms pre-wet channel sediment

(Coe et al., 2008; Kean and Staley, 2011; Schmidt et al., 2011). Studies at Chalk Cliffs show that

wet channel beds are more erosive and contribute more volume to an eroding debris flow than

dry channels do (McCoy et al., 2010). Channel-bed moisture levels are easy to measure with

soil-moisture probes, but cost-prohibitive because probes are lost in scouring flows. Sediment

availability, on the other hand, is a notoriously difficult parameter to characterize. In burned

basins, 75-80% of total eroded material comes from channel beds and banks, so channels should

be the first assessment for sediment availability.

143

The R3 digital photographs and 1m DEMs may provide a way to assess channel sediment

availability. First, channels in basins with debris-flow volumes estimated in this study could be assessed using the photo sets and the 1m DEMs to extract an estimate of the material evacuated.

Scoured channels are clear in the photos so plannimetric area of scour can be determined. Depth of scour will be more difficult but possible. Bart Matthews (R3 Photogrammetry Program manager) estimated the width and depth of a gully in the Horseshoe 2 Fire from the digital stereo photos and 1m DEMs. We found during a field assessment that these measurements were accurate. Although most channels aren’t as large as a gully, channel-depth estimated from the high-resolution digital photographic and DEM dataset should be reasonable. Volumes from channels using this method will be an approximation, but the data will provide more information than is currently available. As new fires burn, the digital photographs will provide pre-fire channel conditions with which to assess postfire erosion. Rainfall and NEXRAD radar data from storms preceding debris-flow events could be used to get an estimate of channel bed wetness.

Results from this study would provide new volume data that would help inform how best to parameterize available channel sediment for debris-flow volume models.

B.8 Conclusions

In this paper, we characterized rainfall conditions and basin flow responses from recently burned basins during the first summer following five wildfires in Arizona. We developed objectively defined rainfall intensity-duration thresholds for various durations to use in issuing warnings and, with these thresholds, developed an ID threshold curve to compare with other regional and worldwide curves. We also assessed the predictive strengths of three multivariate

144

post-wildfire probabilistic debris-flow models, Models A, B and C, and one multivariate debris- flow volume model for use by BAER teams in Arizona.

Objectively defined rainfall ID threshold models were most successful for durations of

10- and 15-minutes, although the predictive strengths of 5-, 30- and 60-minutes were only slightly lower than the I10 and I15 thresholds. The rainfall ID curve shows that post-wildfire debris flows in Arizona are triggered by short-duration storms with higher intensities than other documented threshold curves in the western U.S.

Assessments of the predictive strengths of the three post-wildfire debris-flow models yielded mixed results. On the Colorado Plateau, any of the models are appropriate for assessing burned basins if they are located on the volcanic mountains (e.g., San Francisco Peak, White

Mountains, Kendrick Peak, ), however, none of the models are appropriate for basins on the more gently sloping Colorado Plateau. In the Transition Zone, model C appears to be most appropriate model using a higher probability threshold. And in the

Basin and Range, model B with a moderate threshold is appropriate, but additional assessment can be made using model A with a high probability threshold, and/or model C with a lower probability threshold.

145

B.9 References

Anderson, C.A., and Blacet, P.M., 1972, Geologic map of the Quadrangle, Yavapai County, Arizona: U.S. Geologcal Survey, Geologic Quadrangle Maps of the United States, Map GQ-997, scale 1:62,500. Baum, R., and Godt, J., 2010, Early warning of rainfall-induced shallow landslides and debris flows in the USA: Landslides, v. 7, no. 3, p. 259-272, doi:10.1007/s10346-009-0177-0. Bradley, A.P., 1997, The use of the area under the ROC curve in the evaluation of machine learning algorithms: Pattern Recognition, v. 30, no. 7, p. 1145-1159, doi:http://dx.doi.org/10.1016/S0031-3203(96)00142-2. Caine, N., 1980, The Rainfall Intensity: Duration Control of Shallow Landslides and Debris Flows: Geografiska Annaler. Series A, Physical , v. 62, no. 1/2, p. 23-27. Cannon, S.H., Boldt, E.M., Laber, J.L., Kean, J.W., and Staley, D.M., 2011, Rainfall intensity– duration thresholds for postfire debris-flow emergency-response planning: Natural Hazards, v. 59, no. 1, p. 209-236, doi:10.1007/s11069-011-9747-2. Cannon, S.H., and Ellen, S., 1985, Abuntdant debris avalanches, San Francisco Bay Region, California: California Geology, v. 38, no. 12, p. 267-272. Cannon, S.H., Gartner, J.E., Rupert, M.G., Michael, J.A., Rea, A.H., and Parrett, C., 2010, Predicting the probability and volume of postwildfire debris flows in the intermountain western United States: Geological Society of America Bulletin, v. 122, no. 1-2, p. 127- 144, doi:10.1130/b26459.1. 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, v. 96, no. 3-4, p. 250-269. 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, v. 96, no. 3-4, p. 270-297. Coelho, C.d.O.A., Ferreira, A.J.D., Boulet, A.-K., and Keizer, J.J., 2004, Overland flow generation processes, erosion yields and solute loss following different intensity fires: Quarterly Journal of Engineering Geology Hydrogeology, v. 37, no. 3, p. 233-240, doi:10.1144/1470-9236/03-043. DeBano, L.F., 2000, The role of fire and soil heating on water repellency in wildland environments: a review: Journal of Hydrology, v. 231-232, p. 195-206. DeLong, S.B., Henderson, W., and Youberg, A., 2012, Post wildfire landscape change in a small headwater catchment from terrestrial LiDAR, in Southwest Wildfire Hydrology & Hazard Workshop Proceedings, 2, Tucson, Arizona, Arizona Geological Survey, p. 20. Doerr, S.H., Shakesby, R.A., Blake, W.H., Chafer, C.J., Humphreys, G.S., and Wallbrink, P.J., 2006, Effects of differing wildfire severities on soil wettability and implications for hydrological response: Journal of Hydrology, v. 319, no. 1-4, p. 295-311, doi:10.1016/j.jhydrol.2005.06.038. Drewes, H., 1996, Geologic maps of the , southeast Arizona and southwest New Mexico: US Geological Survey Bulletin, p. 17-41. Ebel, B.A., Moody, J.A., and Martin, D.A., 2012, Hydrologic conditions controlling runoff generation immediately after wildfire: Water Resources Research, v. 48, no. 3, p. W03529, doi:10.1029/2011wr011470. 146

Eng, J., 2005, Receiver Operating Characteristic Analysis: A Primer1: Academic Radiology, v. 12, no. 7, p. 909-916, doi:http://dx.doi.org/10.1016/j.acra.2005.04.005. Fawcett, T., 2006, An introduction to ROC analysis: Pattern Recognition Letters, v. 27, no. 8, p. 861-874, doi:http://dx.doi.org/10.1016/j.patrec.2005.10.010. Frattini, P., Crosta, G., and Sosio, R., 2009, Approaches for defining thresholds and return periods for rainfall-triggered shallow landslides: Hydrological Processes, v. 23, no. 10, p. 1444-1460. Friedman, E., 2012, Debris-Flow Hazard Assessment for the 2010 Medano Pass Fire Burn Area, Great Sand Dunes National Park and Preserve, Colorado: Golden, CO, Colorado School of Mines, 93 p. Fule, P.Z., Yocom, L.L., Montano, C.C., Falk, D.A., Cerano, J., and Villanueva-Diaz, J., 2012, Testing a pyroclimatic hypothesis on the Mexico-United States border: Ecology, v. 93, no. 8, p. 1830-1840. Gartner, J.E., Bigio, E.R., and Cannon, S.H., 2004, Compilation of post wildfire runoff-event data from the Western United States, U. S. Geological Survey. Reston, VA, United States. Godt, J.W., Baum, R.L., and Chleborad, A.F., 2006, Rainfall characteristics for shallow landsliding in Seattle, Washington, USA: Earth Surface Processes and Landforms, v. 31, no. 1, p. 97-110. Godt, J.W., and Coe, J.A., 2007, Alpine debris flows triggered by a 28 July 1999 thunderstorm in the central Front Range, Colorado: Geomorphology, v. 84, no. 1-2, p. 80-97. Godt, J.W., Schulz, W.H., Baum, R.L., and Savage, W.Z., 2008, Modeling Rainfall Conditions For Shallow Landsliding In Seattle, Washington: Reviews in Engineering Geology Volume XX: Landslides and Engineering Geology of the Seattle, Washington, Area, v. 20, no. 0, p. 137-152. Grissino-Mayer, H.D., and Swetnam, T.W., 2000, Century-scale climate forcing of fire regimes in the American Southwest: The Holocene, v. 10, no. 2, p. 213-220. Guzzetti, F., Peruccacci, S., Rossi, M., and Stark, C., 2008, The rainfall intensity–duration control of shallow landslides and debris flows: an update: Landslides, v. 5, no. 1, p. 3-17. Hand, D.J., 2010, Evaluating diagnostic tests: The area under the ROC curve and the balance of errors: Statistics in Medicine, v. 29, no. 14, p. 1502-1510, doi:10.1002/sim.3859. ---, 2012, Assessing the Performance of Classification Methods: International Statistical Review, v. 80, no. 3, p. 400-414, doi:10.1111/j.1751-5823.2012.00183.x. Hayes, P.T., and Raup, R.B., 1968, Geologic map of the Huachuca and , southeastern Arizona: U.S. Geologica Survey, Miscellaneous Geologic Invstigations, Map I-509, scale 1:48,000. Heede, B., 1985, Channel adjustments to the removal of log steps: an experiment in a mountain stream: Environmental Management, v. 9, no. 5, p. 427-432, doi:10.1007/bf01866341. Holm, R.F., 1988, Geologic map of San Francisco Mountain, Elden Mountain, and Dry Lake Hills, Coconino County, Arizona: U.S. Geological Survey, Miscellaneous Investigations Series Map I-1663, scale 1:24000. 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, p. 2013JF002796, doi:10.1002/jgrf.20148. 147

Kean, J.W., and Staley, D.M., 2011, Direct measurement of the hydrologic conditions leading up to and during post-fire debris flow in southern California, USA, in Genevois, R., Hamilton, D.L., and Prestininzi, A., Debris‐Flow Hazards Mitigation, Mechanics, Prediction, and Assessment: Casa Editrice Univ. La Sapienza, Rome, p. 685-694. 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: J. Geophys. Res., v. 116, no. F4, p. F04019, doi:10.1029/2011jf002005. Littell, J.S., McKenzie, D., Peterson, D.L., and Westerling, A.L., 2009, Climate and wildfire area burned in western U.S. ecoprovinces, 1916–2003: Ecological Applications, v. 19, no. 4, p. 1003-1021, doi:10.1890/07-1183.1. 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, v. 38, no. 8, p. 735-738, doi:10.1130/g30928.1. McCoy, S.W., Kean, J.W., Coe, J.A., Tucker, G.E., Staley, D.M., and Wasklewicz, T.A., 2012, Sediment entrainment by debris flows: In situ measurements from the headwaters of a steep catchment: Journal of Geophysical Research: Earth Surface, v. 117, no. F3, p. F03016, doi:10.1029/2011jf002278. Montgomery, D.R., Schmidt, K.M., Greenberg, H.M., and Dietrich, W.E., 2000, Forest clearing and regional landsliding: Geology (Boulder), v. 28, no. 4, p. 311-314. Moody, J.A., and Ebel, B.A., 2012, Hyper-dry conditions provide new insights into the cause of extreme floods after wildfire: CATENA, v. 93, no. 0, p. 58-63, doi:10.1016/j.catena.2012.01.006. 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, v. 379, no. 3-4, p. 291- 303, doi:DOI: 10.1016/j.jhydrol.2009.10.015. Moody, J.A., and Martin, D.A., 2001, Initial hydrologic and geomorphic response following a wildfire in the Colorado Front Range: Earth Surface Processes and Landforms, v. 26, no. 10, p. 1049-1070. Moody, J.A., and Martin, D.A., 2009, Synthesis of sediment yields after wildland fire in different rainfall regimes in the western United States: International Journal of Wildland Fire, v. 18, no. 1, p. 96-115. Moody, J.A., Shakesby, R.A., Robichaud, P.R., Cannon, S.H., and Martin, D.A., 2013, Current Research Issues Related to Post-wildfire Runoff and Erosion Processes: Earth-Science Reviews, v. 122, p. 10-37, doi:http://dx.doi.org/10.1016/j.earscirev.2013.03.004. Nations, D., and Stump, E., 1981, Geology of Arizona: Debuque, Iowa, Kendall/Hunt Publishing Company. Neary, D.G., Ryan, K.C., and DeBano, L.F., 2005, Wildland Fire in : Effects of Fire on Soil and Water ( Revised 2008 ed.): Fort Collins, CO, USDA Forest Service, Rocky Mountain Research Station, General Technical Report RMRS-GTR-42, v. 4, p. 250. Pallister, J.S., and du Bray, E.A., 1992, Geologic map of the Fife Peak quadrangle, Cochise County, Arizona, in Survey, U.S.G.: Reston, VA, U.S. Geological Survey Open-File Report 92-0242, p. 9 p., 2 sheets, scale 1:24,000.

148

---, 1994, Geologic map of the Fife Peak quadrangle, Cochise County, Arizona: U.S. Geological Survey Quadrangle Map GQ-1708, 1 sheet, scale 1:24,000. Pearthree, P.A., and Youberg, A., 2004, Fire and Sediment Deposition: Arizona Geology, v. 34, no. 3, p. 1-2. Restrepo-Posada, P.J., and Eagleson, P.S., 1982, Identification of independent rainstorms: Journal of Hydrology, v. 55, no. 1–4, p. 303-319, doi:10.1016/0022-1694(82)90136-6. Richard, S.M., Reynolds, S.J., Spencer, J.E., and Pearthree, P.A., 2000, Geologic Map of Arizona: Arizona Geolgocial Survey Map M-35, 1 sheet, scale 1:1,000,000. Rihs, J., 2011, Wallow Wildfire BAER Resource Report, Apache-Sitgreaves National Forest, Hydrology, US Department of Agriculture, unpublished report, p. 46. Ruddy, B.C., 2011a, Probability and volume potential postwildfire debris flows in the 2011 Horseshoe II Burn Area, Southeastern Arizona: OFR 2011-1197: Reston, VA, U.S. Geological Survey, p. 10. ---, 2011b, Probability and volume potential postwildfire debris flows in the 2011 Monument Burn Area, Southeastern Arizona: OFR 2011-1181: Reston, VA, U.S. Geological Survey, p. 9. ---, 2011c, Probability and volume potential postwildfire debris flows in the 2011 Wallow Burn Area, Eastern Arizona: OFR 2011-1214: Reston, VA, U.S. Geological Survey, p. 11. Schmidt, K.M., Hanshaw, M.N., Howle, J.F., Kean, J.W., Staley, D.M., Stock, J.D., and Bawdeng, G.W., 2011, Hydrologic conditions and terrestrial laser scanning of post-fire debris flows in the San Gabriel Mountains, CA, USA, in Genevois, R., Hamilton, D.L., and Prestininzi, A., Debris Flow Hazards Mitigation, Mechanics, Prediction, and Assessment: Casa Editrice Univ. La Sapienza, Rome, p. 583-593. Schoennagel, T., Veblen, T.T., and Romme, W.H., 2004, The Interaction of Fire, Fuels, and Climate across Rocky Mountain: Bioscience, v. 54, no. 7, p. 661-676. Sheppard, P.R., Comrie, A.C., Packin, G.D., Angersbach, K., and Hughes, M.K., 2002, The climate of the US Southwest: Climate Research, v. 21, no. 3, p. 219-238. 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, p. 1-16, doi:10.1007/s10346-012-0341-9. Stein, S.M., Comas, S.J., Menakis, J.P., Carr, M.A., Stewart, S.I., Cleveland, H., Bramwell, L., and Radeloff, V.C., 2013, Wildfire, wildlands, and people: understanding and preparing for wildfire in the wildland-urban interface, USDA Forest Service, Rocky Mountain Research Station, General Technical Report RMRS-GTR-299, p. 40. Stoof, C.R., Moore, D., Fernandes, P.M., Stoorvogel, J.J., Fernandes, R.E.S., Ferreira, A.J.D., and Ritsema, C.J., 2013, Hot fire, cool soil: Geophysical Research Letters, v. 40, no. 8, p. 1534-1539, doi:10.1002/grl.50299. Swetnam, T.W., and Anderson, R.S., 2008, Fire Climatology in the western United States: introduction to special issue: International Journal of Wildland Fire, v. 17, no. 1, p. 1-7. Swetnam, T.W., and Baisan, C.H., 1996, Historical fire regime patterns in the Southwestern United States since AD 1700, in Allen, C.D., Fire Effects in Southwestern Forests, Proceedings of the Second La Mesa Fire Symposium: Las Alamos, New Mexico, USDA Forest Service, p. 10-33.

149

Swetnam, T.W., and Betancourt, J.L., 1998, Mesoscale Disturbance and Ecological Response to Decadal Climatic Variability in the American Southwest: Journal of Climate, v. 11, no. 12, p. 3128-3147. Tillery, A.C., Cannon, S.H., and Micheal, J.A., 2011a, Postwildfire preliminary debris flow hazard assessment for the area burned by the 2011 Las Conchas Fire in north-central New Mexico, U.S. Geological Survey Open-File Report 2011-1308, p. 11. Tillery, A.C., Darr, M.J., Cannon, S.H., and Micheal, J.A., 2011b, Postwildfire preliminary debris flow hazard assessment for the area burned by the 2011 Track Fire in Northeastern New Mexico and Southeastern Colorado, U.S. Geological Survey Open-File Report 2011-1257, p. 9. Trapp, R.A., and Reynolds, S.J., 1995, Map showing names and outlines of physiographic areas in Arizona used by the Arizona Geological Survey with comprehensive base map: Arizona Geological Survey. USDA Forest Service, 2010, Schultz Fire Burned Area Emergency Response Report, USDA Forest Service, , unpublished report, p. 167. ---, 2011a, Horseshoe 2 Fire Burned-Area Report, FS-2500-8, USDA Forest Service, Coronado National Forest, unpublished report, p. 21. ---, 2011b, Monument Fire Burned-Area Report, FS-2500-8, USDA Forest Service, Coronado National Forest, unpublished report. ---, 2011c, Wallow Wildfire Burned-Area Report, FS-2500-8, USDA Forest Service, Apache- Sitgreaves National Forest, unpublished report, p. 32. ---, 2012, Gladiator Fire Burned-Area Report, FS-2500-8, USDA Forest Service, Prescott National Forest, unpublished report, p. 39. Wagenbrenner, J.W., and Robichaud, P.R., In Review, Post-fire sediment delivery across spatial scales in the interior western US: Earth Surface Processes and Landforms. Webb, R.H., Magirl, C.S., Griffiths, P.G., and Boyer, D.E., 2008, Debris flows and floods in southeastern Arizona from extreme precipitation in late July 2006: Magnitude, frequency, and sediment delivery, U.S. Geological Survey Open-File Report 2008-1274: Tucson, Arizona, United States Geological Survey, p. 95. Westerling, A.L., Gershunov, A., Brown, T.J., Cayan, D.R., and Dettinger, M.D., 2003, Climate and wildfire in the western United States: Bulletin of the American Meteorological Society, v. 84, no. 5, p. 595-604. Westerling, A.L., Hidalgo, H.G., Cayan, D.R., and Swetnam, T.W., 2006, Warming and Earlier Spring Increase Western U.S. Forest Wildfire Activity: Science, v. 313, no. 5789, p. 940- 943, doi:10.1126/science.1128834. Williams, A.P., Allen, C.D., Millar, C.I., Swetnam, T.W., Michaelsen, J., Still, C.J., and Leavitt, S.W., 2010, Forest responses to increasing aridity and warmth in the southwestern United States: Proceedings of the National Academy of Sciences, v. 107, no. 50, p. 21289- 21294, doi:10.1073/pnas.0914211107. Youberg, A., 2008, Delineating post-wildfire debris-flow hazards for pre-fire mitigation, Pine and Strawberry, Arizona: A FEMA 5% Initiative Study, Open-File Report 08-08: Tucson, Arizona Geological Survey, p. 20. Youberg, A., 2011, Horseshoe II Fire, Burned Area Emergency Response (BAER) Assessment, Geologic Hazard Assessment: Specialist Report: Tucson, Coronado National Forest, p. 15. 150

B.10 Figure and Table Captions

Table B.1. Multivariate statistical models to predict postwildfire debris-flow probability and volumes (Cannon et al., 2010).

Table B.2. Fire location and dates, area burned, and final soil burn severity as a percentage of the entire burned area.

Table B.3. Number of basins monitored, and ranges of basin morphometry and soil burn severity parameters used in the models. Means are in parenthesis.

Table B.4. Debris-flow producing rainfall data.

Table B.5. AUC, objectively defined thresholds and threat scores for the three multivariate, statistical, postwildfire, debris-flow models using data from 80 basins recently burned by five Arizona wildfires.

Figure B.1. Physiographic provinces of Arizona and locations of study wildfires.

Figure B.2. Locations of study basins. Panel A: Locations of wildfires in this study. Colorado Plateau: Panel B: 2010 Schultz Fire. Panel C: 2011 Wallow Fire. Transition Zone: Panel D: 2012 Gladiator Fire. Basin and Range: Panel E: 2011 Horseshoe 2 Fire. Panel F: 2011 Monument Fire. Study basins outlined in black. Rain gauge locations shown with blue diamonds. Fire perimeters outlined in red. Soils burn severity shown by colors of dark green (unburned), light green (low soil burn severity), yellow (moderate) and high (red).

Figure B.3. ROC analysis methods and metrics. Panel A. 2x2 Contingency table matrix categorizing the possible outcomes of a bimodal model and performance metrics. Panel B. ROC space with hypothetical curves and AUC values. Point A represents a near perfect model, point B represents a random model and point E represents a worse than random model. Point C and curve I represent models with a high true positive rate at the cost a more false alarms. This is the liberal region of the ROC space. Point D and curve H represent models with a lower false alarm rate at the cost of true outcomes. This is the conservative region of ROC space. Curve G is less successful than either curve than either curve H or I. (Modified from Staley et al., 2012).

Figure B.4. Flow responses of study basins. Flow responses by basin color with brown indicating at least one debris flow and blue indicating no debris flows. Panel A: Locations of wildfires in this study. Panel B: 2010 Schultz Fire. Panel C: 2011 Wallow Fire. Panel D: 2012 Gladiator Fire. Panel E: 2011 Horseshoe 2 Fire. Panel F: 2011 Monument Fire. Study basins outlined in black. Locations of rain gauges shown with blue diamonds. Fire perimeters outlined in red.

Figure B.5. Box and whisker plots of debris-flow and flood producing rainfall populations of peak intensities for various durations. 151

Figure B.6. Objectively defined rainfall intensity-duration threshold curves for durations of 5- minutes (a), 10-minutes (b), 15-minutes (c), 30-minutes (d), and 60-minutes €, and plotted in ROC space (f).

Figure B.7. ROC curves of models A, B and C for each fire (a-e) with AUC numbers.

Figure B.8. Measured to modeled debris-flow volumes for 26 basins. Basins on the Colorado Plateau are red circles, those from the Basin and Range are black squares. No volume measurements were collected for the Transition Zone. The solid line shows a perfect fit, the dashed lines represent two standard errors of the predicted value.

Figure B.9. Rainfall intensity-duration threshold curves for non-wildfire (solid) and wildfire- related (dashed) debris flows. Rainfall intensity thresholds for AZ - Arizona, objectively defined thresholds, this study, SEAZ – southeastern Arizona (Webb et al., 2008), SWA - Seattle, Washington (Godt et al., 2006), MRO – Mettman Ridge, Oregon (Montgomery et al., 2000); RMA –Rocky Mountain Alpine, Colorado (Godt and Coe, 2007), CCC – Chalk Cliffs, Colorado (Coe et al., 2008), MRF – Missionary Ridge Fire and CSF – Coal Seam Fire, Colorado (Cannon et al., 2008), SFC – San Francisco Bay Area, California (Cannon and Ellen, 1985), OGF1 and OGF2 – Old and Grand Prix Fires, 1st and 2nd years (Cannon et al., 2008), and OGPO – objectively defined thresholds for the Old and Grand Prix Fires (Staley et al., 2012).

Figure B.10. ROC graphs for each model and each fire using a probability threshold of 50% (Panel A) and 60% (Panel B). Colorado Plateau: Schultz Fire (Sch) in green squares and Wallow Fire (WAL) in orange squares. Transition Zone: Gladiator Fire (Gld) in purple triangles. Basin and Range: Horseshoe 2 Fire (Hs2) in blue diamonds and Monument Fire (Mon) in red diamonds.

152

Table B.1. Parameters of multivariate statistical models.

Debris-flow Probability P = ex/(1+ex): Equations for x Models A x = -0.7 + 0.3S – 1.6R + 0.6B + 0.2C – 0.4L + 0.7I B x = -0.76 – 1.1R + 0.6B + 0.09C – 1.4O + 0.06I C x = 4.8 + 0.5B + 0.2C – 0.4L – 1.5H+ 0.07I Debris-flow Volume Model ln V = 7.2 + 0.6(lnA) 0 0.7(B’)½ + 0.2(T) ½ + 0.3 Model Parameters S Percent of basin with slopes ≥ 30% ½ R Ruggedness (Melton Ratio) = Basin Elev Range/(Abasin) B Percent of basin burned at moderate + high soil burn severity C Percent clay O Percent organic matter L Liquid limit H Hydrologic group I Average storm rainfall intensity (mm h-1) T Storm total (mm) A Area of the basin with slopes ≥ 30% (km2) B’ Area of the basin burned at mod+high soil burn severity (km2) 0.3 Bias correction that changes volume median value to mean

153

Table B.2. Study wildfires data.

Total Physio- Soil Burn Severity (% of Total Area) Dates Area Fire Name graphic Range Burned Burned1 Unburned Low Moderate High Province (ha) Colorado San Francisco 6/20-30 Schultz 6,100 8 25 27 40 Plateau Peaks 2010 Colorado Plateau - White 5/29-7/8 Wallow 217,740 24 48 14 16 Transition Mountains 2011 Zone Transition Bradshaw 5/13-6/13 Gladiator 6,572 18 29 32 21 Zone Mountains 2012 Basin and Chiricahua 5/8-7/20 Horseshoe 2 90,226 20 38 30 12 Range Mountains 2011 Basin and Huachuca 6/12-7/12 Monument 12,353 14 40 39 7 Range Mountains 2011 1Area burned and burn severity data from final soil burn severity images (tiffs or rasters).

154

Table B.3. Number of basins monitored, and ranges of basin morphometry and soil burn severity parameters used in the models. Means are in parenthesis.

Range and Average () of : Number of: Number Number of %Basin %Basin of Basin Debris- Fire Name Monitored Relief Rugged- Slopes Burned at %Liquid %Organic Hydrologic Flood Nested Size %Clay Flow Basins (m) ness GTE Mod+High Limit Matter Group Basins Basins (km2) Basins 30% Severity 0.1- 252- 21.3- 3.6 1160 0.4-1.0 0.4-1.0 32.8 29.4-39.3 0.1-0.1 Schultz Fire 31 10 (0.7) (492) (0.7) (0.8) 0.5-1.0 (0.9) (23.1) (31.0) (0.1) 2.7-2.8 (2.7) 7 24 0.5- 144- 31.4- Wallow 2.3 317 0.1-0.4 0.1-0.8 35.8 36.2-41.9 0.0-0.1 Fire 8 0 (1.4) (225) (0.2) (0.3) 0.1-0.8 (0.5) (33.8) (39.3) (0.05) 2.0-2.6 (2.3) 5 3 0.4- 453- 17.5- Gladiator 2.4 780 0.5-0.7 0.6-0.9 23.6 22.9-28.2 0.2-0.5 Fire 7 0 (1.0) (562) (0.6) (0.8) 0.5-1.0 (0.8) (22.3) (27.1) (0.4) 3.2-3.5 (3.3) 4 3 0.1- 177- 19.0- Horseshoe 13.5 886 0.2-1.5 0.4-1.0 21.1 25.3-27.3 0.1-0.3 2 Fire 25 10 (1.8) (584) (0.6) (0.8) 0.7-1.0 (0.8) (20.6) (26.3) (0.1) 3.0-4.0 (3.4) 10 15 0.1- 402- 21.1- Monument 5.9 1173 0.5-1.1 0.5-1.0 24.2 26.2-30.7 0.1-0.1 Fire 10 0 (3.0) (905) (0.6) (0.9) 0.5-1.0 (0.8) (21.6) (27.0) (0.1) 3.4-3.4 (3.4) 5 5

155

Table B.4. Debris-flow producing rainfall data.

Peak Peak Physio- Storm Total Storm Average I10 I30 graphic Start of Duration Rainfall Frequency Intensity (mm (mm Province Fire Gage Storm (hours) (mm) (yrs) (mm h-1) h-1) h-1) 7/20/2010 CP Schultz 3470 13:18 0.75 41 20 55 132 70 8/16/2010 CP Schultz 3470 10:44 0.78 27 5 34 90 36 7/20/2010 CP Schultz 3460 13:19 0.77 45 20 59 120 80 8/16/2010 CP Schultz 3460 11:11 0.52 26 5 50 84 7/20/2010 CP Schultz 3450 13:18 0.87 37 10 43 84 60 7/11/2011 CP Wallow 9489082 14:45 1.50 42 5 28 63 54 Horseshoe 7/11/2011 BR 2 Chiri_B2 12:11 1.63 54 5 33 82 72 Horseshoe 8/11/2011 BR 2a Chiri_B2 19:31 2.68 25 <1 9 78 46 Horseshoe 8/11/2011 BR 2a KC2CPZ1 19:36 3.38 29 <1 9 38 25 Horseshoe 8/13/2011 BR 2a Chiri_B2 14:00 1.42 18 <1 13 62 33 Horseshoe 8/13/2011 BR 2a KC2CPZ1 14:03 1.24 24 <1 19 31 27 Horseshoe 8/15/2011 BR 2 Chiri_B2 14:09 2.95 24 <1 8 53 40 Horseshoe 8/15/2011 BR 2 KC2CPZ1 13:59 0.85 24 <1 29 48 36 7/10/2011 BR Monument 3051 13:31 1.07 41 2 38 102 70 10/1/2011 BR Monument 3051 15:04 0.65 16 <1 25 54 28 8/11/2011 BR Monumentb 3050 13:10 1.15 7 <1 6 18 10 8/13/2011 BR Monumentb 3050 16:52 1.65 13 <1 8 60 24 7/14/2012 TZ Gladiator 3815 16:09 2.56 61 10 24 115 55 7/14/2012 TZ Gladiator 3820 16:14 2.62 55 10 21 65 35 aDebris-flow producing storms in nearby basins not included in model testing. Rain data was used to develop intensity-duration thresholds. bGauge 3050 is on the lower slopes of upper Ash Canyon and do not appear to reflect actual total rainfall in Copper (11th) and Lutz (13th) Canyons. Gauge and upper basins of Lutz and Copper are in a radar shadow and cannot be assessed. Basins used in testing debris-flow models but rainfall not used to develop intensity-duration thresholds.

156

Table B.5. AUC, objectively defined thresholds and threat scores.

Model A Model B Model C OBJ_ Threat OBJ_ Threat OBJ_ Threat Fire AUC Probability Score AUC Probability Score AUC Probability Score Schultz 0.93 66% 0.73 0.87 66% 0.63 0.93 45% 0.69 Wallow 0.71 7% 0.33 0.38 4% 0.10 0.53 8% 0.20 Gladiator 0.73 95% 0.25 0.73 87% 0.25 0.78 73% 0.25 Horseshoe 2 0.67 85% 0.21 0.68 59% 0.22 0.64 41% 0.21 Monument 0.61 89% 0.11 0.61 39% 0.12 0.63 55% 0.11

157

B.11 Figures

Figure B.1. Physiographic provinces of Arizona and locations of study wildfires.

158

Figure B.2. Locations of study basins.

159

Figure B.3. ROC analysis methods and metrics.

160

Figure B.4. Flow responses of study basins.

161

Figure B.5. Box and whisker plots of peak intensities for various durations.

162

Figure B.6. Objectively defined rainfall intensity-duration threshold curves.

163

Figure B.7. ROC curves of models.

164

Figure B.8. Measured to modeled debris-flow volumes.

165

Figure B.9. Rainfall intensity-duration threshold curves.

166

Figure B.10. ROC graphs by model and fire.

167

APPENIDX C: BASIN PHYSIOGRAPHIC AND SOIL BURN SEVERITY FACTORS INFLUENCING THE GENERATION OF POST-WILDFIRE DEBRIS FLOWS

Ann Youberg, Victor R. Baker, Philip A. Pearthree

Manuscript intended for submission to Natural Hazards

168

C.1 Abstract

Post-wildfire debris-flows hazards impose increasing risks as wildfires increase in size

and severity. Here we analyze physiographic and soil burn severity data from 86 basins recently

burned by wildfires in Arizona, 56% of which produced debris flows. The data come from 3

distinctive physiographic provinces, and they correspond to multivariate statistical model

parameters that have been widely applied to post-wildfire debris-flow hazard assessments in the intermountain western U.S. We extracted morphometric factors (basin ruggedness, slopes, etc.) in a GIS framework from 10 m DEMs and soil burn severity parameters from 30 m dNBR continuous and 10 m categorical data. Physiographic model factors did discriminate non-debris- flow (nD) from debris-flow (D) producing basins, though only weakly, for storms that followed

5 major 2010-2012 fires (Schultz, Wallow, Gladiator, Horseshoe 2, and Monument). However, slope-area plots for the basins revealed a consistent difference in slopes between nD and D basins. By studying 12 discrete intervals of slope values ranging between 14° and >42° we found that a very effective discriminating criterion is the percent basin area with slopes above a threshold of 30°. Moreover, the discrimination between nD and D basins becomes stronger with a combination of percent basin areas with high soil burn severity on slopes above this threshold value. Thus, a combination of steep slopes and high soil burn severity can be used to identify basins subject to post-wildfire debris flows according to specific criteria identified in this study.

Another criterion shown to discriminate nD from D basins is mean elevation, which probably is important because of the zonal vegetation and related soil organic and root cohesion factors that characterize Arizona. Finally, we found that post-fire channel heads mapped for the 2011

Monument Fire formed at essentially the same slope range (~30-400) as 2006 saturation induced

169

debris flows, a result that is both useful for identifying potential post-fire debris flow hazards and may provide information to guide the future development of physically based models.

170

C.3 Introduction

Flooding and debris-flow generation have long been known to result from the interactions

of both meteorological and physiographic factors. The former include storm characteristics that

combine with other transient factors (interception, evaporation, infiltration, etc.) to generate the

form of flood hydrographs, while the latter include the more permanent characteristics of

channels, the drainage network, and basin factors (slopes, relief, aspects, etc.) (Rodda, 1969).

One of most critical needs for operational hydrology is the prediction of hydrological response in

small, ungagged basins (Hrachowitz et al., 2013). Moreover, the distinction between debris

flows and water floods, and their associated hazards, is a pressing problem for flood hydrology

and hazard mitigation (Costa, 1988; Pierson, 2005b).

In a companion study (Appendix B) objectively defined rainfall intensity-duration thresholds for Arizona were identified that specify the rainfall conditions under which post- wildfire debris flows are likely to be generated. An Arizona dataset of 80 recently burned basins impacted by 82 different storms developed during the companion study was used to assess the predictive strengths of three published multivariate statistical post-wildfire debris-flow probability models that had been developed for the intermountain western U.S. using data from the northern portion of that region (Cannon et al., 2010). All three probability models performed adequately but with variable results in most cases across Arizona, including in applications to rainfall intensity return intervals greater than those in the datasets from which the models were derived. In one burned area, however, all three models completely failed (Appendix B). Results of that study provide important operational guidelines for the quick assessments that must be made by Burned Area Emergency Response (BAER) teams of the potential for damaging debris

171

flows associated with post-fire rainfall. These assessments are essential in choosing appropriate

emergency response and mitigation measures.

Whereas the emphasis in Appendix B was on assessing rainfall factors in conjunction

with evaluating the three probability models for post-wildfire debris-flow prediction in Arizona,

the present study complements Appendix B by focusing on the physiographic factors in

combination with burn severity measures. Physiographic factors have generally received less

attention than meteorological factors, so this study is necessarily explorative. However, if criteria for the recognition of basins likely to generate debris flows can be identified on a physiographic basis, then potentially hazardous areas may be identified in advance by considering combinations

of future wildfire and rainfall conditions that would result in debris flows. Such identification

can be used to prioritize precautionary efforts that might be needed immediately after a wildfire

and prior to a debris-flow triggering rainfall event. The results presented in Appendix B can be

combined with results from the present study, which seeks to identify those physiographic

criteria that, in combination with burn severity factors, make particular basins more susceptible

to debris-flow generation.

In this study, we assess several basin morphometric factors and burn severity parameters

(Appendix B; Cannon et al., 2010), in addition to various slope measurements, using our Arizona

dataset of recently burned basins to identify characteristics that may better indicate the potential

for generating post-fire debris-flows. It might seem likely in a qualitative sense that a

combination of steep slopes and high burn severity will lead to an increased tendency for debris-

flow generating processes. This can be considered to be a working hypothesis. However, it is not

clear exactly how to translate this insight into specific, quantitative criteria that can be applied in

172

a practical operational sense to prioritize possible problem areas following wildfires. Moreover,

it is also not clear if the same practical criteria can be applied to the kinds of diverse

physiographic regions represented by recent Arizona wildfire locations. These practical

considerations also form part of our motivation for the study, in addition to (1) the need to

evaluate how well the model parameters identify basins likely to produce debris-flow responses,

and (2) to use a large data set of post-wildfire basin flow responses to identify patterns in

specific, easily measured criteria that may help discriminate between non-debris-flow and debris- flow producing basins.

We hypothesize that there are differences in slopes and soil burn severity between debris-

flow and non-debris-flow producing basins that can be identified through geomorphic and GIS

analysis. The expectation is that, in an analysis of burned basins, slope-scale patterns of upper basins hillslopes will be identified that can inform future efforts towards (1) improving the existing multivariate statistical models used operationally to predict post-wildfire debris flow probabilities (Cannon et al., 2010), (2) identifying potential debris-flow producing basins before they burn to target mitigation efforts (Abramson et al., 2009), and (3) possibly discovering key variables and/or thresholds that can contribute to the future development of physically based models for post-wildfire debris-flow hazard prediction (Parrett et al., 2004; Gabet and Bookter,

2008; Kean et al., 2013).

In the post-wildfire landscape, soil burn severity is the key factor increasing runoff and is

the disturbance mechanism that allows for the generation of post-wildfire debris flows. Soil burn

severity can be portrayed as categorical (low, moderate, high) or continuous (discrete values for

each pixel). Arguments have been made that burn severity as described by continuous data is a

173

better metric than categorical burn severity data because the data are not smoothed and subtle

trends in burn severity remain. While this may be important for assessing post-wildfire runoff

from small hillslope plots, the expectation has been that during hazard assessments of burned basins either measure is adequate. In this study we assess and compare both categorical and continuous data in the context of a hazard assessment. A critical component of debris-flow

hazards assessments is determining the sediment available for transport, which directly impacts

debris-flow magnitudes. Currently, extensive, time-consuming fieldwork is required to assess sediment available for transport. However, the relatively rapidly generated physiographic

measures discussed in this paper, when combined with burn severity measures, may be able to

identify areas of high potential for post-fire debris-flow generation, so that any needed field investigation of sediment accumulation can be more effectively focused on the areas of highest potential.

C.3.1 Burn severity measures

Post-wildfire hydrologic responses are strongly related to soil burn severity (e.g. DeBano,

2000; Neary et al., 2005; Moody et al., 2008; Nyman et al., 2011; Shakesby, 2011). Quantifying soil burn severity is important for understanding potential changes to hydrologic conditions that influence the occurrence of post-wildfire debris flows. Areas classified as high soil burn severity indicate near complete consumption of surface vegetation, litter and duff (Parsons et al., 2010), which results in decreased surface roughness and increased hydrologic connectivity leading to increased runoff velocities (Larsen et al., 2009; Cawson et al., 2013). Soils burned at high severity have disaggregated soil structure due to consumption of soil organics and fine roots

(Cannon and Reneau, 2000; Moody and Nyman, 2012), and reduced infiltration capacity due to

174

hyper-dry soil conditions, fire-induced soil-water repellency, decreased hydraulic conductivity,

surface sealing and air entrapment (Shakesby et al., 2000; Moody et al., 2005; Doerr et al., 2006;

Moody et al., 2009; Ebel et al., 2012), These changes result in increased runoff volumes with

higher boundary shear stresses, and increased soil erodibility and lower erosional thresholds

(Moody et al., 2005; Nyman et al., 2013). Thus, critical shear stress thresholds for sediment

entrainment are crossed with less contributing area than in pre-fire conditions (Istanbulluoglu et

al., 2004; Gabet and Bookter, 2008; Wohl, 2013). Consequently, rills develop more easily very

high on the upper portions of hillslopes, concentrating fluid flow and transporting pulses of

sediment downslope into coalescing flow areas, valley heads or channels, where runoff-induced

debris flows can form as described by the sediment capacitor model (Kean et al., 2013), the

firehose effect or other mechanisms (Johnson and Rodine, 1984; Gabet and Bookter, 2008; Santi

et al., 2008; Parise and Cannon, 2012).

Burn severity is characterized using pre- and post-fire LANDSAT images taken about a

year apart to reflect similar growing conditions (McKinley, 2013). Each image is made using an

algorithm called normalized burn ratio (NBR)

= (C.1) 푅4−푅7 푁퐵푅 �푅4+푅7� that employs bands 4 and 7 (Miller and Thode, 2007). Band 4 is a near-infrared wavelength that is sensitive to chlorophyll of living plants, and band 7 is a short wave infrared (SWIR) wavelength that is sensitive to water content in both soils and vegetation and in separating dead wood from soil, ash and charred wood (Miller and Thode, 2007). The NBR images are then differenced by

175

dNBR = NBRpre-fire – NBRpost-fire (C.3)

to create a continuous-value dNBR burn severity image (McKinley, 2013) Continuous dNBR

values are converted to an 8-bit (256 values) image and divided into four ranges of burn severity

to create a Burned Area Reflectance Classification (BARC) map (McKinley, 2013). This raw- data map is adjusted by BAER teams based on field assessments and on-the-ground soils tests to create a final soil burn severity map that is classified into four categories of unburned (1), low

(2), moderate (3) and high (4) (Parsons et al., 2010). While the final soil burn severity data have the benefit of incorporating field observations, the data are categorical and hence smooth variability within each category. Because subtle trends can be seen that may help delineate areas with greater potential for erosion Moody et al. (2009) argue that the continuous data of dNBR offer a better way to assess soil burn severity as the latter influences runoff at the hillslope scale.

However, these subtle burn severity trends may not be as critical in a hazard assessment. The post-wildfire debris-flow probability models use the metric “percent of basin area burned at moderate plus high soil burn severity” (Cannon et al., 2010).

C.3.2 Debris-flow processes

Debris flows are unsteady, non-uniform, very poorly sorted, sediment-rich slurries that are generated by saturated soil-induced hillslope failures (Costa and Williams, 1984; Iverson,

2003a), or by runoff-induced mechanisms (Coe et al., 2008; McCoy et al., 2010; Kean et al.,

2011; Kean et al., 2013). Saturation-induced debris flows are initiated by unusually prolonged or intense precipitation falling onto saturated hillslopes until pore pressures increase and shear strengths decrease to the critical point of failure (Costa, 1984; Cannon and Ellen, 1985; Iverson et al., 1997; Webb et al., 2008b; Montgomery et al., 2009), Initiation occurs for these debris

176

flows by slumping or rotational failures, or by shallow translational failures of thin soil over

impervious surfaces, such as bedrock, that liquefies and rapidly mobilizes into viscous slurries

through liquefaction or dilatancy (Costa, 1984; Iverson et al., 1997; Iverson et al., 2010). The initiation points for these saturation-induced debris flows can clearly be identified by the failure scarps remaining on the hillslopes.

Runoff-induced debris flows occur in disturbed or sparsely vegetated, high-sediment yield basins by shorter-duration, higher-frequency, lower-magnitude storms (Cannon, 2001;

Griffiths et al., 2004; Cannon et al., 2008; Coe et al., 2008; Webb et al., 2008b; McCoy et al.,

2010). Runoff-induced debris flows predominate in the post-fire environment, typically occurring within a year following wildfires during relatively common (2- to 10-year frequency) rainfall events (Cannon, 2001; Schmidt et al., 2011; Parise and Cannon, 2012) when antecedent soil moisture is low or absent (Cannon et al., 2008; Kean and Staley, 2011; Friedman, 2012).

Unlike saturation-induced debris flows, initiation points of runoff-induced debris flows are difficult to locate. Initiation of runoff-induced debris flows may occur through a variety of processes, including the firehose effect in which runoff entrains colluvial material below cliffs

(Johnson and Rodine, 1984; Griffiths et al., 2004; Godt and Coe, 2007), progressive sediment entrainment of channel material by surface flow (Wohl and Pearthree, 1991; Cannon et al., 2001;

Gabet and Bookter, 2008; Santi et al., 2008; McCoy et al., 2012), mass failure of channel bed or bank materials (Takahashi, 2007), or by centimeter-scale soil slips (Gabet, 2003). Runoff- induced debris flows, however, probably initiate most commonly through a series of processes where pulses of sediment-laden flows originating as hillslope rills are transported to a channel or concentrated flow area, deposited and de-watered at an in-channel step or subtle change in

177

topography, building up progressively until the deposit becomes destabilized, at which point a

subsequent pulse of sediment breaks through the deposit creating a dam breach and initiating a

debris flow (Kean et al., 2013). This process, modeled as a sediment capacitor, has been documented at the USGS Chalk Cliffs Observatory in Colorado and in basins burned by wildfires in southern California (Kean et al., 2013).

Debris-flow initiation zones have been documented on slopes ranging from 14° (25%), to

>40° (84%) (Takahashi, 1981; Rickenmann and Zimmermann, 1993; Coe et al., 2008; Kean et al., 2013). Slopes of in-channel, runoff-induced debris flows can occur between 17.5° to >40° depending on boundary conditions (Kean et al., 2013). In the Grand Canyon, the minimum slope necessary for debris-flow initiation is 20° (36%) (Webb et al., 2008a), while work in southeastern Arizona suggests slopes where debris flows initiate range between 28.5° - 40° (54-

84%) (Melton, 1965a).

In July, 2006, a five-day extreme precipitation event resulted in 113 shallow, translational hillslope failures that coalesced into 64 debris flows in the southern Huachuca Mountains

(Youberg et al., 2006). The hillslope failures, mapped in the field and using 0.6 m resolution imagery, were located on slopes ranging from 17°-55° (31-143%) with a mean, and median, slope of 35° (70%) (Youberg, unpublished data). The multivariate probability debris-flow models consider slopes within basins using the metric “percent basin area with slopes ≥30%”

(16.6°) (Cannon et al., 2010). In this study, we assess different ranges of slope values and burn severity to identify patterns for delineating potential post-wildfire debris-flow basins.

178

C.4 Study Area

This study uses post-wildfire basin-flow responses data collected from 86 burned basins during the first summer following five recent wildfires located in Arizona’s three physiographic provinces (Appendix B). The fires used in this study include the 2010 Schultz Fire and 2011

Wallow Fire on the Colorado Plateau in northern Arizona, the 2012 Gladiator Fire in the Central

Highlands Transition Zone, and the 2011 Horseshoe 2 Fire and 2011 Monument Fire in the Basin and Range Province in southeastern Arizona (Figure C.1; for a complete description of each burned area study site, see Appendix B). While each physiographic zone has distinct geologic,

geomorphic, and vegetation characteristics that reflect the topography and the climate of that

zone, there are commonalities between burned areas.

The forests of Arizona are generally found in the higher elevations of mountainous areas,

and on the Mogollon Rim where extensive Ponderosa Pine forests grow. Vegetation

communities within the forested regions are related to elevation and latitude. The Schultz Fire

burned 6100 ha on the eastern slopes of the San Francisco Peaks between elevations of

approximately 2,100 m to 3,370 m, with vegetation communities of Ponderosa Pine forests in the

lower elevations, mixed conifer forests in the upper elevations, and spruce-fir forests at the highest elevations (USDA Forest Service, 2010). Primary lithologic units are Pleistocene volcanic rocks and Quaternary surficial deposits (Holm, 1988). The Wallow Fire burned 217,740 ha in the White Mountains between elevations of 1,570m to 3,325 m (Rihs, 2011). The study basins in the Wallow Fire burned area were located between elevations of 2,490 m to 2,845 m with vegetation communities of mountain grasslands, Ponderosa Pine forests, and mixed conifer forests (USDA Forest Service, 2011c). Primary lithologic units are Neogene sedimentary and

179

volcanic rocks (Richard et al., 2000). The Gladiator Fire burned 6,560 ha in the Bradshaw

Mountains between elevations of 1,241 m to 2,316 m, with vegetation communities of chaparral in the lower and mid elevations, and Ponderosa Pine forests with small amounts of mixed conifer forests at the upper elevations (USDA Forest Service, 2012). Primary lithologic units are

Precambrian granite, volcanic and metamorphic rocks (Anderson and Blacet, 1972). The

Horseshoe 2 Fire burned 90,226 ha in the Chiricahua Mountains between elevations of 1,300 m to 2,975 m with vegetation communities of grasslands and oak savannas and woodlands at the lower elevations, piñon-juniper and pine-oak woodlands at lower to mid elevations, and Madrean pine forests and mixed conifer forests at mid to upper elevations (USDA Forest Service, 2011a;

Arizona Firescape, http://www.azfirescape.org). Lithologic units include a diverse suite of

Proterozoic, Paleozoic, and Cretaceous metamorphic, sedimentary and volcanic rocks, Oligocene volcanic rocks, and Quaternary basalts and surficial deposits (Pallister and du Bray, 1992, 1994;

Drewes, 1996). The Monument Fire burned 12,353 ha in the Huachuca Mountains between elevations of 1,335 m to 2,885 m, with vegetation communities of desert grasslands below approximately 1,525 m, chaparral, oak woodlands, and pine-oak woodlands in the mid elevations, and small amounts of mixed conifer forests above approximately 2,135 m (USDA

Forest Service, 2011b). Lithologies include Precambrian granite on the lower flanks of the mountain, and numerous Paleozoic sedimentary units, Mesozoic volcanic and sedimentary rocks, and Quaternary surficial deposits (Hayes and Raup, 1968).

Rainfall varies widely across Arizona and with elevation. Precipitation is bimodal with approximately half of the annual precipitation falling as rain during the summer monsoon (July -

September) and a second peak of precipitation during the winter (November - March), with

180

snowfall contributing proportional amounts in upper elevations (Sheppard et al., 2002). The

North American Monsoon is stronger in the southern part of the state and average rainfall intensities are higher (Sheppard et al., 2002; Moody and Martin, 2009). In Arizona, monsoon immediately follows wildfire season and often is what ultimately extinguishes wildfires.

C.5 Methods

Basin morphometric and burn severity parameters from recently burned basins are analyzed to identify characteristics and patterns that may better indicate the likelihood of a basin generating post-fire debris-flows. In the companion study, 80 basins from five burned areas were used to assess rainfall factors and to evaluate post-wildfire debris-flow models (Appendix B). Six basins in the Horseshoe 2 Fire burned area that were not included in the first study, due to distances from rain gauges, are included here for a total of 86 basins. Burned basin flow responses were monitored during the first summer after each fire. Field reconnaissance of deposits, along with photographs, videos and other field reports, was used to determine the occurrence of debris flows. The downstream extents of debris flows were mapped to define basin outlets, typically delineated by an alluvial fan or similar flow expansion reach.

There are several factors that limit field observations, and these, in turn affect the accuracy of the data and propagate throughout the work. Identifying and discriminating debris- flow deposits from flood deposits may be challenging due to reworking by subsequent floods

(Melis et al., 1997, type III flows). Debris flows and floods are part of a continuum of fluid- sediment flows, thus in some cases it is difficult to confidently assess whether deposits were emplaced by floods or debris flows (Pierson, 2005a). The timing of field observations sometimes occurred after road crews had removed all material so deposits could not be assessed. And

181

finally, limited exposures of deposits make it difficult to assess sorting, bedding, and other important deposit characteristics. In all cases when it was not possible to verify that a debris flow had occurred or that the deposit was debris-flow related the basin flow was classified as non- debris-flow.

For each monitored basin, morphometric factors, continuous and categorical soil burn severity data, and slope interval data were extracted using GIS. We compiled a geodatabase for each study fire containing information on basin outlet locations, the presence (D) or absence

(nD) of debris flow(s) and corresponding debris-flow date(s), as well as basin morphometric factors, burn severity characteristics and other factors. The data for all fires were combined and classified by basin responses; those that produced at least one debris flow were labeled as D, and those that did not were labeled as nD. GIS data were then exported as two populations for statistical analyses, with n values of 38 for nD and 48 for D for an N of 86 basins.

All statistical analyses were conducted in SigmaStats, which is part of the SigmaPlot software. Data were first explored through descriptive statistics, by testing for population distributions and by visual assessments using scatterplots and box plots. Populations were statistically compared using a two-tailed t-test for normally distributed populations, and the

Mann-Whitney (Wilcoxan) Rank Sum test for non-normally distributed populations. Non- parametric tests like the Mann-Whitney Rank Sum test compare population medians as opposed to means tested by the t-test. For small N values this is a problem, but for larger N values the

Mann-Whitney Rank Sum test provides a p-value that is the same or similar to that provided by the t-test. P-values are provided for both t-test and the Mann-Whitney Rank Sum tests for comparisons when available. Tests were based on a 95% confidence interval (P = 0.05).

182

This study was done in phases with each phase informing the next. During the first phase, basin morphometric factors and basin-wide burn severity were assessed for all basins. In the second phase, nD and D basins were assessed using slope-area plots for each fire area. During the third phase, we assessed various slope intervals, compared categorical and continuous soil burn severities, and then assessed slopes with measures high soil burn severity. During the last phase, we used a test case in the Huachuca Mountains (2011 Monument Fire) to compare slopes where post-fire channel heads were observed to slopes where non-fire related, saturation-induced hillslope failures occurred in 2006.

Basin morphometric factors assessed in this study include average basin gradient, percent of basin area with slopes ≥30% or ≥50%, basin ruggedness (change in basin elevation divided by the square root of the basin area; Melton, 1965b), basin relief, a basin relief ratio (change in basin elevation divided by the longest flow path), and percent of basin area burned at moderate plus high soil burn severity. These factors have all either been incorporated into the Cannon et al.

(2010) post-wildfire debris-flow models used in Arizona (see Appendix B), or they were considered for use in formulating such models (Cannon et al., 2010). We consider them all here in order to see the relative contributions by these factors to the accuracy of the models for predicting debris flows. Other basin physiographic/morphometric factors assessed for their possible contribution to predicting potential post-wildfire debris flows include basin shape (basin area divided by basin length), elongation ratio (basin width divided by basin length), a hypsometric index (mean minus minimum elevation divided by basin relief), mean elevation, and basin area. Basin morphometrics were extracted in Esri’s ArcMap software using 10 m DEMs.

183

Soil burn severity data were extracted in ArcMap first from 10 m categorical final soil burn

severity data, and later from 30 m dNBR continuous data.

Slope-area plots of nD and D basins for each fire were developed from data generated

using TauDEM tools in ArcMap. Specific contributing area, which is equivalent to contributing

area per unit width (contributing area normalized to the grid resolution), and local slope grids

were created. Both grids were derived using the d-infinity algorithm. A point feature dataset was

created with a point located in the center of each grid cell. The specific contributing area and

slope data were extracted to the point layer and then exported into nD and D populations by fire.

The data were averaged using 0.1 log bins and plotted. These log-log plots suggested a consistent

difference in slopes between the nD and D populations, thus motivating a more in-depth

assessment of the slope factor.

Slope intervals were extracted for each basin in ArcMap using 10 m DEMs, beginning

with the model parameters slopes ≥30% and ≥50%. We used these values because metrics of

percent basin area with these two slope ranges were chosen for operational reasons during the

development of the debris-flow probability models (Susan Cannon, written communication, Nov.

8, 2013). A more detailed set of slope intervals was then developed. A literature search provided

boundary slope conditions of 14° to 40° for generating debris flows and slopes generally on the

cusp of stability (Van Burkalow, 1945; Melton, 1965a; Rickenmann and Zimmermann, 1993;

Coe et al., 2008; Webb et al., 2008a; Kean et al., 2013; Lamb et al., 2013). Twelve slope intervals were then developed to reflect model constraints and previous research (Table C.1).

The lower slope intervals, A-D, were divided based on the operational boundaries of 30% (16.6°) and 50% (26.5°) in order to assess how well those slopes distinguished nD from D basins.

184

Interval E was based on Melton’s (1965a) work that suggested 28.5° (54%) as the lower boundary of semi-stable slopes. Starting at interval F (30°), slopes are in the zone of the threshold angle of stability of debris covered slopes (Melton, 1965a), and intervals was consistently increased by increments of 2° (Table C.1). Percent basin area in each zone were extracted from slope grids and statistically analyzed. Finally, categorical and continuous soil burn severity data on selected slopes were analyzed.

The final phase of this study consisted of mapping post-fire channel heads in the 2011

Monument Fire burned area, Huachuca Mountains, using high-resolution (0.23 m) digital aerial photographs collected by the U.S. Forest Service in late August of 2011 (Figure C.2). Channel heads were located based on the highest point where incision could be identified, with a threshold channel width of approximately 1 m (Figure C.2, Panel B). Broad, shallowly scoured rills were not mapped, nor were areas obscured by vegetation or shadows. Channel heads were mapped only in the northern part of the fire to avoid the area where non-fire-related, saturation-

induced hillslope failures transitioned into debris flows in 2006 (Figure C.2, Panel A; Youberg et

al., 2006). Most of the hillslope failure scarps were field mapped during August and September

of 2006 (Figure C.2, Panel C), with inaccessible scarps mapped using an October, 2006, Digital

QuickBird image (resolution 0.6 m). The post-fire channel heads are point features in the

Monument Fire geodatabase, and the hillslope failure scarps are from a separate geodatabase.

Specific contributing area and local slope data were extracted to the post-fire channel heads and the saturation-induced failure scarps point data and then plotted on 10 m derived slope-area graphs.

185

C.6 Results and Discussion

C.6.1 Basin morphometric factors

We assessed basin morphometric factors for differences between basins that did not produce debris flows (nD) and those that did (D) (Table C.2). Three slope metrics and three relief metrics are statistically different (Figure C.3, Table C.2). Box plots indicate more separation between the two populations for mean percent slopes, percent basin area with slopes

≥50%, and basin relief (Figure C.3). Basin shape, elongation ratio, and basin areas between the two populations are not statistically different, but mean basin elevations and the hypsometric index are (Table C.2).

Basin morphometric measures of slope and relief consistently indicate that debris-flow basins are steeper than the non-debris-flow basins. This relationship between debris-flow basins and steepness becomes stronger when considering the steeper slopes (i.e. ≥50% vs ≥30%) in watersheds (Figure C.3). Mean basin elevation and the hypsometric index are also statistically different (Table C.2). While these metrics provide evidence of slope differences between nD and

D basin populations, they provide no information regarding where the differences may occur or the importance of these differences.

C.6.2 Slope-area plots

Log-binned slope-area plots were constructed for each basin type (nD/D) by fire to explore basin-wide patterns (Figure C.4). Plots for nD and D basins from the Schultz and

Wallow Fires (Figure C.4, Panels A and B) show distinct slope separations between the nD and

D slope-area plots. This slope separation is consistent but weaker on the Gladiator (Figure C.4,

186

Panel C) and Horseshoe 2 Fires(Figure C.4, Panel D), and slightly stronger on the Monument

Fire (Figure C.4, Panel E).

Slope-areas plots provide information about dominant hydrologic processes and erosional thresholds in basins (Montgomery and Foufoula-Georgiou, 1993; Istanbulluoglu et al., 2002).

The relationship between specific contributing area and local slope changes from positive to negative as hydrological processes transition from diffusive hillslope to fluvial (Willgoose et al.,

1991; Montgomery, 2001). The shape of the curve also provides information about the location of dominant processes and erosional thresholds within basins (Montgomery and Foufoula-

Georgiou, 1993; Stock and Dietrich, 2003). In non-glaciated regions, slope-area plots can show signature non-linear curvatures of slope that indicate transitioning of dominant processes from diffusive hillslope (Region I) to unincised valley heads (Region II) to debris-flow dominated channels (Region III) and then to fluvially dominated channels (Region IV) (Ijjasz-Vasquez and

Bras, 1995; Montgomery, 2001; Stock and Dietrich, 2003; Stock and Dietrich, 2006; Tarolli and

Dalla Fontana, 2009).

Some interesting trends can be seen in the slope-area plots for the fires in this study

(Figure C.4). First, as already discussed, there is a separation in slopes between nD and D basins, with the strongest separation between basins on the Colorado Plateau (FigureC. 3, Panels A and

B). Second, the overall steepness of the basins varies among the five study fires, and hence geographic area, with the lowest slopes located in the Wallow Fire burn area (Figure C.4, Panel

B). Third, the contributing areas that are associated with the transition from one region to the next, say Region I to Region II, vary by burned area but show similar patterns (widths of regions) by physiographic area. The Wallow Fire basins have the largest diffusive hillslope

187

region (Figure C.4, Panel B, Region I), which is to be expected for lower gradient basins. The

debris-flow dominated region (Region III) is smallest for basins on the Colorado Plateau (Figure

C.4, Panels A and B), and is similar for basins in the Transition Zone (Figure C.4, Panel C) and

the Basin and Range (Figure C.4, Panels D and E), suggesting the minimum contributing areas

for transitioning from debris-flow (Region III) to fluvially dominated (Region IV) channels

varies for different geographic areas. However, this conclusion should be viewed as more

preliminary than definitive because the slope-area plots were developed with 10 m DEM data, which has a coarse enough resolution to shift leftward contributing area thresholds between regions (Tarolli and Dalla Fontana, 2009). Slope-area plots created with higher resolution data would provide a better assessment of contributing area for each region (Tarolli and Dalla

Fontana, 2009). Nevertheless, these slope-area plots do show that slope is a factor that helps distinguish, to varying degrees, between nD and D basins, again suggesting that slopes be assessed in more detail to achieve the goals of the study.

C.6.3 Slope intervals

Data for 12 different slope intervals were extracted and analyzed using either a two-tailed t-test or the Mann-Whitney Rank Sum test (Table C.4). Data for each slope interval are plotted to provide a visual analysis of the population distributions (Figures C.5 and C.6). Scatterplots and box plots show mean basin elevation (x-axis) against the percent basin area in each slope interval

(y-axis). The box plots at the tops of the graphs show the population distribution of mean basin elevation between nD (gray boxes) and D (white boxes) basins, and are consistent between each graphs, while the box plots on the right sides of the graphs, representing the percent of basin area within each slope interval, change for each slope interval as do the scales of the y-axes. The two

188

lowest slope intervals, A and B, are statistically different, but the next two intervals, C and D, are

not (Table C.3). In slope intervals A-C (14-26.5°), nD basins have comparatively more area than

D basins (Figure C.5). The population distribution transitions over intervals D-E (26.5-30°) from a lesser to a greater percentage basin area for D basins (Figure C.5). Slope interval E (28.5-30°) and higher intervals are all statistically different (Table C.3), with the greatest separation between the population distributions of nD and D basins occurring over slope intervals G (32-

34°) and H (34-36°) (Figure C.6).

The two lowest slope intervals (A-B) are statistically different, and are noteworthy due to

the smaller percentage of basin area representing D basins as compared to nD basins (Figure C.5,

Panels A-B), A reversal in these population distributions occurs over a slope range between 20-

30° (C-E) with D basins occupying a larger percent basin area on slopes ≥30° (Figure C.5, Panel

F). Although the greatest separation occurs over a slope range between 32-36° (Figure C.6,

Panels G and H), this trend continues over all of the upper slope intervals. Thus, beginning with

slopes approximately ≥30° there is a separation between nD and D basins, and for operational

purposes during hazard assessments it would be reasonable to assess burned basins using all

slopes ≥30°.

C.6.3 Soil burn severity

Categorical moderate (Figure C.7) and high (Figure C.8) soil burn severity data were

assessed for percent of basin area (Panel A), on slopes ≥16.6° (30%, Panel B), ≥26.5° (50%,

Panel C) and ≥30° (58%, Panel D). Basin-wide moderate soil burn severity data show a

statistically different pattern, with the debris-flow (D) basins having less percent area of moderate soil burn severity than the non-debris-flow (nD) basins (Figure C.7, Panel A). The

189

three comparisons of nD to D basins with moderate soil burn severity on different measures of slope are not statistically different (Figure C.7, Panels B-D). On slopes ≥16.6° and ≥26.5° nD basins have more areas of moderate soil burn severity than D basins, and moderate soil burn severity areas are approximately equal on slopes ≥30° (Figure C.7).

The opposite trend is true for measures of high soil burn severity for basin area (Figure

C.8, Panel A), and on slopes ≥16.6° (30%, Figure 7, Panel B), ≥26.5° (50%, Figure C.8, Panel C) and ≥30° (58%, Figure C.8, Panel D). Although there are no statistical differences basin-wide, the percent of basin area with high soil burn severity is greater for D basins than nD basins

(Figure C.8, Panel A). D basins have a higher percentage of basin area with high soil burn severity than nD basins with all three measures of slope, and each slope measure is statistically different (Figure C.8, Panels B-D).

A comparison between categorical (FSBS) and continuous (dNBR) soil burn severity data was made to determine if differences between these data types might impact a hazard assessment. Using a combination scatterplots and box plots, data representing percent basin area for FSBS were plotted first against dNBR (Figure C.9, Panel A), and then against percent basin area on slopes ≥30° with high soil burn severity (Figure C.9, Panel B). Both sets of data plot closely to the 1:1 line, but the basin-wide populations are not statistically different (P = 0.099) while the combination of slope and high burn severity data are (P = <0.001).

To assess the combined impacts of slope and high soil burn severity, three graphs were generated to show a comparison of percent of basin area with slopes ≥30° (Figure C.10, Panel

A), and percent of basin area with high soil burn severity on slopes ≥30° using continuous dNBR

190

data (Figure C.10, Panel B) and categorical FSBS data (Figure C.10, Panel C). Each measure is

plotted against mean basin elevation.

Basin-wide and slope-limited (≥16.6°, ≥26.5°, ≥30°) assessments of moderate and high

categorical soil burn severity data (Figures C.7 and C.8) show that no measures of moderate soil

burn severity, nor does basin-wide high soil burn severity provide a discrimination between nD

and D basins. High soil burn severity combined with slope steepness, however, does discriminate

between nD and D basins. This separation becomes stronger with increasing slopes, suggesting

that slopes coupled with burn severity can be used to identify potential post-wildfire debris-flow basins.

Comparisons of the continuous dNBR data to the categorical FSBS data show that both data sets provide similar results and plot close to the 1:1 line (Figure C.9). Thus, either measure would be appropriate for a hazard assessment. A basin-wide comparison between the two datasets doesn’t reveal any trends (Figure C.9, Panel A). When comparing high soil burn severity on slopes ≥30°, nD basins appear to have a maximum area of approximately 20% high soil burn severity, while D basins range up to 50% (Figure C.9, Panel B). The combined impacts of slope and high soil burn severity can be assessed first by considering only slopes ≥30° (Figure C.10,

Panel A), and then slopes ≥30° with high soil burn severity using continuous dNBR data (Figure

C.10, Panel B) and categorical FSBS data (Figure C.10, Panel C). All of the graphs show some discrimination between nD and D basins, but the trend becomes stronger when high soil burn severity is included (Figure C.10, Panels B and C). Again both burn severity data sources provide similar results.

191

Another trend that is apparent in the scatterplots of slopes (Figures C.5 and C.6) and

slopes with a burn severity component (Figure C.10) is that nD and D basins populations are also

somewhat segregated by mean basin elevation. This relationship might seem counterintuitive

given that the data come from all across Arizona. However, basin elevation in this region acts as a kind of proxy for forest type and density. Lower elevation basins in Arizona have more open, drier forests, while higher basins tend to have denser, moister forests. This is supported by the findings of Istanbullouglu et al. (2004) that forest density is directly related to surface roughness, root cohesion, and the magnitude of episodic erosion after disturbances such as wildfire. Higher elevation forests have higher root densities and thus can provide greater cohesion to soils on steeper slopes. When these areas are burned at high soil burn severity, fine roots in the upper few centimeters are consumed, disaggregating soil structure and lowering erosional thresholds

(Parsons et al., 2010; Moody and Nyman, 2012). The resulting lowered thresholds result in post- wildfire erosion higher on hillslopes, with less contributing area than would be found under undisturbed conditions (Moody and Martin, 2001; Istanbulluoglu et al., 2002; Nyman et al.,

2013). Indeed, channel heads that were mapped pre- and post-wildfire in the Front Range of

Colorado showed a two-fold decrease in contributing area of channel heads eroded after a fire as opposed to before the fire (Wohl, 2013).

Istanbullouglu et al. (2002) suggest that slope-area plots provide information about long- term evolution of the landscape and may not identify thresholds in short-term disturbance environments (wildfires). Therefore, channel heads that form during the high erosion/low threshold period after fires may develop much higher in the landscape, with lower contributing area, than would be identified on a slope-area plots (Istanbulluoglu et al., 2002). Although Stock

192

and Dietrich (2003; 2006) showed that debris flow regions can be defined on slope-area plots by non-linear slope curvatures, Tarolli and Dalla Fontana (2009) showed that the DEM resolution is critical to correctly identifying the slope-area region and contributing area of channel heads.

These researchers also showed that, while 10 m DEMs do not have sufficient resolution to accurately assess contributing areas, the slope measurements remain fairly consistent between

DEM resolutions (Tarolli and Dalla Fontana, 2009). Therefore, while it may not be possible to robustly assess contributing areas to channel heads, it should be possible to compare slopes on which post-fire channel heads develop with slopes identified in this study and with slopes on which saturation-induced debris flows were initiated (Figure C.2).

C.6.6 Slopes and erosional thresholds

To compare these slopes, we plotted post-fire channel heads on the Monument Fire slope- area plot from Figure C.4 (Figure C.11, Panel A), and the saturation-induced hillslope failures on a slope-area graph developed with basin data from the 2006 debris flows using the same 10 m

DEM (Figure C.11, Panel B). Local slopes on which the post-fire channel heads were mapped range from 0.23-1.64 m/m with a mean of 0.71 m/m and median of 0.69 m/m (range:13-63°, mean and median 35°). The saturation-induced hillslope failures developed on local slopes with a range of 0.31-1.38 m/m with a mean of 0.72 m/m and a median of 0.70 m/m (range: 17-55°, mean and median 35°). These two populations are not statistically different (two-tailed P-value =

0.980). The 25-75 percentiles for the channel heads range from 0.57-0.82 m/m and 0.59-0.80 m/m for the hillslope failures (approximately 30° to 40° for both datasets).

Post-fire channel heads are plotted with corresponding basin slope-area plots (Figure

C.11, Panels A and B). In both of these plots, the channel heads and the hillslope failures are

193

mostly located in the areas considered to be valley heads (Region II) or debris-flow dominated

channels (Region III) (Montgomery and Foufoula-Georgiou, 1993; Stock and Dietrich, 2003;

Stock and Dietrich, 2006). There are a few points on each graph that plot in the hillslope

diffusive area (Region I), but this is likely a DEM resolution issue (Tarolli and Dalla Fontana,

2009), and these plots cannot be used to confidently assess contributing area. Statistical analyses

show the slopes on which post-fire channel heads and saturated hillslope failures occurred are

essentially the same.

Although it is unlikely that every mapped post-fire channel head produced debris flows, some portion of the channel heads did result in debris flows, and this limited assessment indicates that runoff-induced debris flows and saturation-induced failures reflect similar slope

thresholds. Although the slopes are similar, the erosional thresholds for channel heads and

saturation-induced failures have different influencing factors. Saturation-induced failures are

driven by high pore pressures (Iverson, 2000) that can be influenced by surface infiltration and

by subsurface flows through joints, fractures and faults (Montgomery et al., 2009). Runoff-

induced debris flows entrain sediment from hillslope rills and channels and the driving force is

runoff (Kean et al., 2013; Nyman et al., 2013). High soil burn severity reduces soil and root

cohesion due to the consumption of soil organics and fine roots (Parsons et al., 2010; Moody and

Nyman, 2012) and decreases surface roughness and increases hydrologic connectivity due to the loss of surface vegetation, litter and duff (Larsen et al., 2009; Cawson et al., 2013), leading to increased runoff volumes and velocities that can more easily exceed boundary shear stresses

(Nyman et al., 2013). Rill erosion can form with modest runoff and supply sediment packets to coalescing flow areas where debris flows can initiate as described by the sediment capacitor

194

model (Kean et al., 2013), the firehose effect (Johnson and Rodine, 1984)or other mechanisms

(Wohl and Pearthree, 1991; Gabet and Bookter, 2008; Santi et al., 2008; Parise and Cannon,

2012).

Based on the work presented here, the slopes on which these processes are likely to occur are ≥30°, with the critical slopes approximately 30-40°. The findings in this study also show, that for this area, post-fire channel heads and saturation-induced hillslope failures formed at

essentially the same slope range (30-40°). Moreover, this is a result that is useful for identifying

potential post-fire debris-flow hazards, and may provide information to guide the future

development of physically based models. Additional work with higher resolution DEMs and

field mapping of post-fire channel heads will be required to make more accurate assessments of

contributing areas.

C.8 Conclusions

In a companion study (Appendix B) objectively defined rainfall intensity-duration

thresholds were developed for Arizona that specify the rainfall conditions under which post-

wildfire debris flows are likely to be generated. Results of that study provide important

operational guidelines for the quick assessments that must be made by BAER teams of the

potential for damaging debris flows associated with post-fire rainfall. Whereas the emphasis in

Appendix B was on assessing rainfall factors in conjunction with evaluating the three probability models for post-wildfire debris-flow prediction in Arizona, the present study complements

Appendix B by focusing on the physiographic factors that act in combination with burn severity

measures.

195

In this study, we assessed basin morphometric factors, burn severity characteristics and

slope data from 86 recently burned basins to identify hillslope-scale patterns useful in assessing

potential post-fire debris-flow basins. Basin morphometric factors with a component of slope or

relief weakly discriminated between nD and D basins (Table C.2, Figure C.3). The statistical

relationships did show that basin shape, elongation ratio, and basin area were not significant

(Table C.2), but mean basin slope, percent basin areas with slopes ≥50%, and basin relief had

some significance in regard to separating debris-flow from non-debris-flow basins (Figure C.3).

These observations show that basin morphometric parameters used in the multivariate debris- flow probability models (Cannon et al., 2010) do discriminate between nD and D basin in

Arizona, but the basin-wide burn severity measures incorporated in the models do not (Table

C.2, Figures C.7 and C.8).

Results from this study show that a combination of two metrics, as described by percent of basin area with high soil burn severity on slopes ≥30°, can be used to help discriminate between basins likely to produce post-wildfire runoff-induced debris flows and those that are not

(Figure C.10). No significant differences were found between categorical and continuous soil burn severity for the purposes of hazard assessments (Figure C.9). Slopes on which post-fire channel heads and non-fire related, saturation-induced failures occurred are consistent, with critical slopes for erosional thresholds between 30-40° (Figure C.11). Additional work is required to achieve a more accurate definition of contributing areas for these erosional thresholds.

In this study, criteria for the recognition of basins likely to generate debris flows were identified on the basis of both physiographic and soil burn severity factors. Results from this

196

study can be combined with results presented in Appendix B to identify basins likely to generate

post-wildfire debris flows. For example, results in Appendix B showed that all three post-

wildfire debris-flow probability models (Cannon et al., 2010) considered for use in Arizona

provided varying degrees of predictive strength throughout most of the state, and thus require a

complicated assessment of which model, and model threshold, to choose. The findings presented

here provide a framework for additional assessments to help identify areas likely to produce

post-wildfire debris flows. During a BAER assessment, results from the debris-flow probability

model(s) can be compared with GIS-derived metrics of (1) percent basin areas with high soil

burn severity on slopes ≥30°, (2) other physiographic measures found to have some relevance in

distinguishing D- from nD- producing areas (mean basin slope, basin relief; Figure C.2), and (3)

mean basin elevation (Figure C.10) or forest type and density. The combination of some or all of

these factors can be used to identify potentially hazardous areas where field efforts can be

focused to assess potential sediment volumes, depositional zones and values-at-risk. Moreover, these same criteria can be used to identify potentially hazardous areas during pre-disaster mitigation assessments by considering combinations of future wildfire and objectively defined rainfall conditions identified in Appendix B that would result in devastating debris flows. The data presented here also may inform the development of physically based models and the slopes that are critical to the generation of post-wildfire debris flows.

197

C.9 References

Abramson, L., Chennell, M., Eisch, E., Glassco, A., and Holley, T., 2009, Post-Fire Sedimentation and Flood Risk Potential in the Mission Creek Watershed of Santa Barbara: Santa Barbara, California, University of California, Santa Barbara, 85 p. Anderson, C.A., and Blacet, P.M., 1972, Geologic map of the Mount Union Quadrangle, Yavapai County, Arizona: U.S. Geologcal Survey, Geologic Quadrangle Maps of the United States, Map GQ-997, scale 1:62,500. Cannon, S.H., 2001, Debris-flow generation from recently burned watersheds: Environmental and Engineering Geoscience, v. 7, no. 4, p. 321-341. Cannon, S.H., and Ellen, S., 1985, Abuntdant debris avalanches, San Francisco Bay Region, California: California Geology, v. 38, no. 12, p. 267-272. Cannon, S.H., Gartner, J.E., Rupert, M.G., Michael, J.A., Rea, A.H., and Parrett, C., 2010, Predicting the probability and volume of postwildfire debris flows in the intermountain western United States: Geological Society of America Bulletin, v. 122, no. 1-2, p. 127- 144, doi:10.1130/b26459.1. 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, v. 96, no. 3-4, p. 250-269. Cannon, S.H., Kirkham, R.M., and Parise, M., 2001, Wildfire-related debris-flow initiation processes, Storm King Mountain, Colorado: Geomorphology, v. 39, no. 3-4, p. 171-188. Cannon, S.H., and Reneau, S.L., 2000, Conditions for generation of fire-related debris flows, Capulin Canyon, New Mexico: Earth Surface Processes and Landforms, v. 25, no. 10, p. 1103-1121. Cawson, J.G., Sheridan, G.J., Smith, H.G., and Lane, P.N.J., 2013, Effects of fire severity and burn patchiness on hillslope-scale surface runoff, erosion and hydrologic connectivity in a prescribed burn: Forest Ecology and Management, v. 310, no. 0, p. 219-233, doi:http://dx.doi.org/10.1016/j.foreco.2013.08.016. 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, v. 96, no. 3-4, p. 270-297. Costa, J.E., 1984, Physical geomorphology of debris flows, in Costa John, E., and Fleisher, P.J., eds., Developments and applications of geomorphology.: Berlin, Federal Republic of Germany, Springer-Verlag, p. 268-317. ---, 1988, Rheologic, geomorphic, and sedimentologic differentiation of water floods, hyperconcentrated flows, and debris flows, in Baker Victor, R., Kochel, R.C., and Patton Peter, C., eds., Flood geomorphology.: New York, NY, United States, John Wiley & Sons, p. 113-122. Costa, J.E., and Williams, G.P., 1984, Debris-flow dynamics: Open-File Report 84-606 (video): Reston, U.S. Geological Survey, Water Resources Division. DeBano, L.F., 2000, The role of fire and soil heating on water repellency in wildland environments: a review: Journal of Hydrology, v. 231-232, p. 195-206. Doerr, S.H., Shakesby, R.A., Blake, W.H., Chafer, C.J., Humphreys, G.S., and Wallbrink, P.J., 2006, Effects of differing wildfire severities on soil wettability and implications for hydrological response: Journal of Hydrology, v. 319, no. 1-4, p. 295-311, doi:10.1016/j.jhydrol.2005.06.038. 198

Drewes, H., 1996, Geologic maps of the Coronado National Forest, southeast Arizona and southwest New Mexico: US Geological Survey Bulletin, p. 17-41. Ebel, B.A., Moody, J.A., and Martin, D.A., 2012, Hydrologic conditions controlling runoff generation immediately after wildfire: Water Resources Research, v. 48, no. 3, p. W03529, doi:10.1029/2011wr011470. Friedman, E., 2012, Debris-Flow Hazard Assessment for the 2010 Medano Pass Fire Burn Area, Great Sand Dunes National Park and Preserve, Colorado: Golden, CO, Colorado School of Mines, 93 p. Gabet, E.J., 2003, Post-fire thin debris flows: sediment transport and numerical modeling: Earth Surface Processes and Landforms, v. 28, no. 12, p. 1341-1348. 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, v. 96, no. 3-4, p. 298-309. Godt, J.W., and Coe, J.A., 2007, Alpine debris flows triggered by a 28 July 1999 thunderstorm in the central Front Range, Colorado: Geomorphology, v. 84, no. 1-2, p. 80-97. Griffiths, P.G., Webb, R.H., and Melis, T.S., 2004, Frequency and initiation of debris flows in Grand Canyon, Arizona: J. Geophys. Res., v. 109, no. F4, p. F04002, doi:10.1029/2003jf000077. Hayes, P.T., and Raup, R.B., 1968, Geologic map of the Huachuca and Mustang Mountains, southeastern Arizona: U.S. Geologica Survey, Miscellaneous Geologic Invstigations, Map I-509, scale 1:48,000. Holm, R.F., 1988, Geologic map of San Francisco Mountain, Elden Mountain, and Dry Lake Hills, Coconino County, Arizona: U.S. Geological Survey, Miscellaneous Investigations Series Map I-1663, scale 1:24000. Hrachowitz, M., Savenije, H.H.G., Blöschl, G., McDonnell, J.J., Sivapalan, M., Pomeroy, J.W., Arheimer, B., Blume, T., Clark, M.P., Ehret, U., Fenicia, F., Freer, J.E., Gelfan, A., Gupta, H.V., Hughes, D.A., Hut, R.W., Montanari, A., Pande, S., Tetzlaff, D., Troch, P.A., Uhlenbrook, S., Wagener, T., Winsemius, H.C., Woods, R.A., Zehe, E., and Cudennec, C., 2013, A decade of Predictions in Ungauged Basins (PUB)—a review: Hydrological Sciences Journal, v. 58, no. 6, p. 1198-1255, doi:10.1080/02626667.2013.803183. Ijjasz-Vasquez, E.J., and Bras, R.L., 1995, Scaling regimes of local slope versus contributing area in digital elevation models: Geomorphology, v. 12, no. 4, p. 299-311, doi:http://dx.doi.org/10.1016/0169-555X(95)00012-T. Istanbulluoglu, E., Tarboton, D.G., Pack, R.T., and Luce, C., 2002, A probabilistic approach for channel initiation: Water Resour. Res., v. 38, no. 12, p. 1325, doi:10.1029/2001wr000782. Istanbulluoglu, E., Tarboton, D.G., Pack, R.T., and Luce, C.H., 2004, Modeling of the interactions between forest vegetation, disturbances, and sediment yields: J. Geophys. Res., v. 109, no. F1, p. F01009, doi:10.1029/2003jf000041. Iverson, R.M., 2000, Landslide triggering by rain infiltration: Water Resources Research, v. 36, no. 7, p. 1897-1910. Iverson, R.M., 2003, The debris-flow rheology myth, in Rickenmann, D., and Chen, C.-l., 3rd International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction, and Assessment: Davos, Switzerland, American Society of Civil Engineers, v. 1, p. 303-314. 199

Iverson, R.M., Logan, M., LaHusen, R.G., and Berti, M., 2010, The perfect debris flow? Aggregated results from 28 large-scale experiments: J. Geophys. Res., v. 115, no. F3, p. F03005, doi:10.1029/2009jf001514. Iverson, R.M., Reid, M.E., and LaHusen, R.G., 1997, Debris-flow mobilization from landslides: Annual Review of Earth and Planetary Sciences, v. 25, p. 85-138. Johnson, A.M., and Rodine, J.R., 1984, Debris flow, in Brunsden, D., and Prior, D.B., eds., Slope Instability: Chichester, Sussex England, John Wiley and Sons Ltd., p. 257-361. 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, p. 2013JF002796, doi:10.1002/jgrf.20148. Kean, J.W., and Staley, D.M., 2011, Direct measurement of the hydrologic conditions leading up to and during post-fire debris flow in southern California, USA, in Genevois, R., Hamilton, D.L., and Prestininzi, A., Debris‐Flow Hazards Mitigation, Mechanics, Prediction, and Assessment: Casa Editrice Univ. La Sapienza, Rome, p. 685-694. 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: J. Geophys. Res., v. 116, no. F4, p. F04019, doi:10.1029/2011jf002005. Lamb, M.P., Levina, M., DiBiase, R.A., and Fuller, B.M., 2013, Sediment storage by vegetation in steep bedrock landscapes: Theory, experiments, and implications for postfire sediment yield: Journal of Geophysical Research: Earth Surface, v. 118, no. 2, p. 1147-1160, doi:10.1002/jgrf.20058. Larsen, I.J., MacDonald, L.H., Brown, E., Rough, D., Welsh, M.J., Pietraszek, J.H., Libohova, Z., de Dios Benavides-Solorio, J., and Schaffrath, K., 2009, Causes of Post-Fire Runoff and Erosion: Water Repellency, Cover, or Soil Sealing?: Soil Sci. Soc. Am. J., v. 73, no. 4, p. 1393-1407, doi:10.2136/sssaj2007.0432. 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, v. 38, no. 8, p. 735-738, doi:10.1130/g30928.1. McCoy, S.W., Kean, J.W., Coe, J.A., Tucker, G.E., Staley, D.M., and Wasklewicz, T.A., 2012, Sediment entrainment by debris flows: In situ measurements from the headwaters of a steep catchment: Journal of Geophysical Research: Earth Surface, v. 117, no. F3, p. F03016, doi:10.1029/2011jf002278. McKinley, R.A., 2013, Wildfire burn severity assessments from satellite data: a review of available products and user considerations, in Moody, J.A., and Martin, D.A., Synthesizing Empirical Results to Improve Predictions of Post-wildfire Runoff and Erosion Responses: Estes Park, Colorado, AGU Chapman Conference. Melis, T.S., Webb, R.H., and Griffiths, P.G., 1997, Debris flows in Grand Canyon National Park; peak discharges, flow transformations, and hydrographs, in Debris-flow hazards mitigation; mechanics, prediction and assessment., San Francisco, California, American Society of Civil Engineers, p. 727-736.

200

Melton, M.A., 1965a, Debris-covered hillslopes of the southern Arizona desert: consideration of their stability and sediment contribution: The Journal of Geology, v. 73, no. 5, p. 715- 729. ---, 1965b, The geomorphic and paleoclimatic significance of alluvial deposits in southern Arizona: The Journal of Geology, v. 73, no. 1, p. 1-38. Miller, J.D., and Thode, A.E., 2007, Quantifying burn severity in a heterogeneous landscape with a relative version of the delta Normalized Burn Ratio (dNBR): of Environment, v. 109, no. 1, p. 66-80, doi:http://dx.doi.org/10.1016/j.rse.2006.12.006. Montgomery, D.R., 2001, Slope Distributions, Threshold Hillslopes, and Steady-state Topography: American Journal of Science, v. 301, no. 4-5, p. 432-454, doi:10.2475/ajs.301.4-5.432. Montgomery, D.R., and Foufoula-Georgiou, E., 1993, Channel network source representation using digital elevation models: Water Resour. Res., v. 29, no. 12, p. 3925-3934, doi:10.1029/93wr02463. Montgomery, D.R., Schmidt, K.M., Dietrich, W.E., and McKean, J., 2009, Instrumental record of debris flow initiation during natural rainfall: Implications for modeling slope stability: J. Geophys. Res., v. 114. 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, v. 379, no. 3-4, p. 291- 303, doi:DOI: 10.1016/j.jhydrol.2009.10.015. Moody, J.A., and Martin, D.A., 2001, Initial hydrologic and geomorphic response following a wildfire in the Colorado Front Range: Earth Surface Processes and Landforms, v. 26, no. 10, p. 1049-1070. Moody, J.A., and Martin, D.A., 2009, Synthesis of sediment yields after wildland fire in different rainfall regimes in the western United States: International Journal of Wildland Fire, v. 18, no. 1, p. 96-115. Moody, J.A., Martin, D.A., Haire, S.L., and Kinner, D.A., 2008, Linking runoff response to burn severity after a wildfire: Hydrological Processes, v. 22, no. 13, p. 2063-2074. Moody, J.A., and Nyman, P., 2012, Variations in Soil Detachment Rates after Wildfire as a Function of Soil Depth, Flow Properties and Root Properties, U.S. Geological Survey Scientific Investigations Report 2012-5233, p. 40. 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. Neary, D.G., Ryan, K.C., and DeBano, L.F., 2005, Wildland Fire in Ecosystems: Effects of Fire on Soil and Water ( Revised 2008 ed.): Fort Collins, CO, USDA Forest Service, Rocky Mountain Research Station, General Technical Report RMRS-GTR-42, v. 4, p. 250. Nyman, P., Sheridan, G.J., Moody, J.A., Smith, H.G., Noske, P.J., and Lane, P.N.J., 2013, Sediment availability on burned hillslopes: Journal of Geophysical Research: Earth Surface, p. 2012JF002664, doi:10.1002/jgrf.20152. Nyman, P., Sheridan, G.J., Smith, H.G., and Lane, P.N.J., 2011, Evidence of debris flow occurrence after wildfire in upland catchments of south-east Australia: Geomorphology, v. 125, no. 3, p. 383-401, doi:10.1016/j.geomorph.2010.10.016. Pallister, J.S., and du Bray, E.A., 1992, Geologic map of the Fife Peak quadrangle, Cochise County, Arizona, in Survey, U.S.G.: Reston, VA, U.S. Geological Survey Open-File Report 92-0242, p. 9 p., 2 sheets, scale 1:24,000. 201

---, 1994, Geologic map of the Fife Peak quadrangle, Cochise County, Arizona: U.S. Geological Survey Quadrangle Map GQ-1708, 1 sheet, scale 1:24,000. Parise, M., and Cannon, S., 2012, Wildfire impacts on the processes that generate debris flows in burned watersheds: Natural Hazards, v. 61, no. 1, p. 217-227, doi:10.1007/s11069-011- 9769-9. Parrett, C., Cannon, S.H., and Pierce, K.L., 2004, Wildfire-related floods and debris flows in Montana in 2000 and 2001, U. S. Geological Survey. [Reston, VA], United States. Pages: 22. 2004. Parsons, A., Robichaud, P.R., Lewis, S.A., Napper, C., and Clark, J.T., 2010, Field guide for mapping post-fire soil burn severity. Gen. Tech. Rep. RMRS-GTR-243: Fort Collins, CO, US Department of Agriculture, Forest Service, Rocky Mountain Research Station. Pierson, T.C., 2005a, Distinguishing between debris flows and floods from field evidence in small watersheds: Reston, VA, U. S. Geological Survey, 4 p. ---, 2005b, Hyperconcentrated flow; transitional process between water flow and debris flow, in Jakob, M., and Hungr, O., eds., Debris-flow hazards and related phenomena: Berlin, Springer. Richard, S.M., Reynolds, S.J., Spencer, J.E., and Pearthree, P.A., 2000, Geologic Map of Arizona: Arizona Geolgocial Survey Map M-35, 1 sheet, scale 1:1,000,000. Rickenmann, D., and Zimmermann, M., 1993, The 1987 debris flows in Switzerland: documentation and analysis: Geomorphology, v. 8, no. 2-3, p. 175-189. Rihs, J., 2011, Wallow Wildfire BAER Resource Report, Apache-Sitgreaves National Forest, Hydrology, US Department of Agriculture, unpublished report, p. 46. Rodda, J.C., 1969, The flood hydrograph, in Chorley, R.J., and Barry, R.G., eds., Water, earth and man: a synthesis of hydrology, geomorphology and socio-economic geography: London, Methuen and Company, p. 405-418. 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, v. 96, no. 3-4, p. 310-321. Schmidt, K.M., Hanshaw, M.N., Howle, J.F., Kean, J.W., Staley, D.M., Stock, J.D., and Bawdeng, G.W., 2011, Hydrologic conditions and terrestrial laser scanning of post-fire debris flows in the San Gabriel Mountains, CA, USA, in Genevois, R., Hamilton, D.L., and Prestininzi, A., Debris Flow Hazards Mitigation, Mechanics, Prediction, and Assessment: Casa Editrice Univ. La Sapienza, Rome, p. 583-593. Shakesby, R.A., 2011, Post-wildfire soil erosion in the Mediterranean: Review and future research directions: Earth-Science Reviews, v. 105, no. 3-4, p. 71-100, doi:10.1016/j.earscirev.2011.01.001. Shakesby, R.A., Doerr, S.H., and Walsh, R.P.D., 2000, The erosional impact of soil hydrophobicity: current problems and future research directions: Journal of Hydrology, v. 231-232, p. 178-191. Sheppard, P.R., Comrie, A.C., Packin, G.D., Angersbach, K., and Hughes, M.K., 2002, The climate of the US Southwest: Climate Research, v. 21, no. 3, p. 219-238. Stock, J., and Dietrich, W.E., 2003, Valley incision by debris flows: Evidence of a topographic signature: Water Resources Research, v. 39, no. 4, p. 1089-1114. Stock, J.D., and Dietrich, W.E., 2006, Erosion of steepland valleys by debris flows: Geol Soc Am Bull, v. 118, no. 9-10, p. 1125-1148, doi:10.1130/b25902.1.

202

Takahashi, T., 1981, Estimation of potential debris flows and their hazardous zones: soft countermeasures for a disaster: Natural disaster science, v. 3, no. 1, p. 57-89. ---, 2007, Debris flow: mechanics, prediction and countermeasures: London, Taylor & Francis. Tarolli, P., and Dalla Fontana, G., 2009, Hillslope-to-valley transition morphology: New opportunities from high resolution DTMs: Geomorphology, v. 113, no. 1-2, p. 47-56. USDA Forest Service, 2010, Schultz Fire Burned Area Emergency Response Report, USDA Forest Service, Coconino National Forest, unpublished report, p. 167. ---, 2011a, Horseshoe 2 Fire Burned-Area Report, FS-2500-8, USDA Forest Service, Coronado National Forest, unpublished report, p. 21. ---, 2011b, Monument Fire Burned-Area Report, FS-2500-8, USDA Forest Service, Coronado National Forest, unpublished report. ---, 2011c, Wallow Wildfire Burned-Area Report, FS-2500-8, USDA Forest Service, Apache- Sitgreaves National Forest, unpublished report, p. 32. ---, 2012, Gladiator Fire Burned-Area Report, FS-2500-8, USDA Forest Service, Prescott National Forest, unpublished report, p. 39. Van Burkalow, A., 1945, Angle of repose and angle of sliding friction: an experimental study: Geological Society of America Bulletin, v. 56, no. 6, p. 669-707, doi:10.1130/0016- 7606(1945)56[669:aoraao]2.0.co;2. Webb, R.H., Griffiths, P.G., and Rudd, L.P., 2008a, Holocene debris flows on the Colorado Plateau: The influence of clay mineralogy and chemistry: Geological Society of America Bulletin, v. 120, no. 7-8, p. 1010-1020, doi:10.1130/b26055.1. Webb, R.H., Magirl, C.S., Griffiths, P.G., and Boyer, D.E., 2008b, Debris flows and floods in southeastern Arizona from extreme precipitation in late July 2006: Magnitude, frequency, and sediment delivery, U.S. Geological Survey Open-File Report 2008-1274: Tucson, Arizona, United States Geological Survey, p. 95. Willgoose, G., Bras, R.L., and Rodriguez-Iturbe, I., 1991, A physical explanation of an observed link area-slope relationship: Water Resources Research, v. 27, no. 7, p. 1697-1702, doi:10.1029/91wr00937. Wohl, E., 2013, Migration of channel heads following wildfire in the Colorado Front Range, USA: Earth Surface Processes and Landforms, v. 38, no. 9, p. 1049-1053, doi:10.1002/esp.3429. Wohl, E.E., and Pearthree, P.A., 1991, Debris flows as geomorphic agents in the Huachuca Mountains of southeastern Arizona: Geomorphology, v. 4, no. 3-4, p. 273-292. Youberg, A., Pearthree, P.A., and Baker, V.R., 2006, Comparison of Debris Flows Generated in Adjacent Unburned and Recently-Burned Areas, Coronado National Memorial, Arizona, Eos Transactions AGU: San Francisco, v. 87(52), no. Fall Meeting Supplement, p. Abstact 12661.

203

C.10 Figure and Table Captions

Table C1. Description and rational of slope intervals extracted in ArcMap to test hypothesis that slopes in debris-flow producing basins are steeper, and to explore which slopes may be most important. Table C2. Statistical tests of differences between nD and D basin areas for basin morphometric and burn severity factors. P-values less that P = 0.05 are indicate statistically different populations. Differences in populations were tested using the two-tailed t-test for normally distributed populations and the Mann-Whitney (Wilcoxon) Rank Sum test for populations that fail the normality test. Table C3. Statistical tests of differences between nD and D basin areas in selected slope intervals. Figure C.1. State of Arizona with physiographic provinces and locations of wildfires in this study. Wildfires are described in Appendix B. (Figure from Appendix B). Figure C.2. Panel A: locations of post-fire channel heads (blue diamonds) in the northern part of the Monument Fire burned area, Huachuca Mountains. Panel B: an example of mapped channel heads using 0.23 m resolution stereo digital photographs. Black double-arrowed line indicates region with unincised rills. Blue diamonds are the mapped channel heads. White double-arrowed line indicates region with incised channels. Panel C: an example of the field-mapped hillslope failures in the southern Huachuca Mountains. White arrows point to saturation-induced failure scarp. Figure C.3. Box plots of slope and basin relief metrics. A: Mean percent slopes of basins. B: Percent of basin area with slopes greater than or equal to 30%. C: Percent of basin area with slopes greater than or equal to 50%. D: Basin relief. E: Melton Ratio. F: Relief Ratio. Table 2 describes how the parameters were derived, statistical results are provided on Table 4. Figure C.4. Slope area plots of nD (black dots) and D (white dots) basins (A-E), plotted from 0.1 log-bin averages, to conduct a first-order assessment of slope differences between basins with and without debris flows. A: 2010 Schultz Fire. B: 2011 Wallow Fire. C: 2012 Gladiator Fire. D: 2011 Horseshoe 2 Fire. E: 2011 Monument Fire. Figure C.5. Scatterplots of non-debris-flow (nD, black dots) and debris-flow (D, white dos) producing basins showing percent basin area for different slope intervals plotted against mean basin elevation. Border box plots show population distributions of nD (gray) and D (white) basins for mean elevation (top) and slope intervals (sides). These figures show lower slopes, between 14-32°. See Tables 4 for an explanation of the slope intervals chosen and Table 6 for P values. Figure C.6. Scatterplots of non-debris-flow (nD, black dots) and debris-flow (D, white dos) producing basins showing percent basin area for different slope intervals plotted against mean basin elevation. Border box plots show population distributions of nD (gray) and D (white) basins for mean elevation (top) and slope intervals (sides). These figures show upper slopes,

204

above 32°. See Tables 4 for an explanation of the slope intervals chosen and Table 6 for P values. Figure C. 7. Box plots of categorical moderate final soil burn severity in basins without debris flows (nD, dark boxes and circles) and basins with at least one debris flow (D, white boxes and circles). Comparing moderate soil burn severity basin wide (A), on slopes ≥16.6° (30%, B), ≥26.5° (50%, C) and ≥30° (58%, D), with P values. Figure C.8. Box plots of categorical high final soil burn severity in basins without debris flows (nD, dark boxes and circles) and basins with at least one debris flow (D, white boxes and circles). Comparing high soil burn severity basin wide (A), on slopes ≥16.6° (30%, B), ≥26.5° (50%, C) and ≥30° (58%, D), with P values. Figure C.9. Comparison of percent basin area occupied by high soil burn severity calculated from categorical final soil burn severity data and continuous dNBR data basin wide (A) and on slopes ≥30° (B). Scatterplots show points plot near the 1:1 line (dashed). Box plots show population distributions of nD (gray) and D (white) basins. Note axis scales changes between panels A and B. Figure C.10. Comparison between nD and D basins of percent basin with slopes ≥30° (A), and percent basin area with high soil burn severity in slopes ≥30° using continuous dNBR data (B) and categorical final soil burn severity data (C). (All P-values = <0.001). Figure C.11. Log-binned slope-are plots from 10 m DEMs with 2011 post-Monument Fire channel heads (A, blue diamonds) and 2006 saturation-induced hillslope failures (B, brown triangles).

205

Table C1. Description and rational of slope intervals.

Slope Slope – Slope – Rational Interval Degrees Percent A 14° - 25% - 14° lowest slopes at Chalk Cliffs, Colorado, generating runoff- <16.6° <30% induced debris flows (Coe et al., 2008), and in Japan (Takahashi, 1981). B 16.6° - 30% - 16.6° (30% percent) used by models (Cannon et al., 2010). 17.5° <20° <36% are the lowest slopes that developed runoff-induced, in-channel debris-flow surges at Chalk Cliffs, Colorado and Arroyo Seco, California (Kean et al., 2013)17°(31%) are the lowest slopes of the 2006 hillslopes failures in Coronado National Memorial, Huachuca Mountains (Youberg, unpublished data). C 20° - 36% - Work in the Grand Canyon and in Utah suggest a minimum of 20° <26.5° <50% is needed for debris flows to occur (Webb et al., 2008). Debris flows generated in debris-filled gullies were documented between slopes of 24° to 35° (Rickenmann and Zimmermann, 1993). D 26.5° - 50% - 50% considered during model development (Cannon et al., 2010). <28.5° <54% 27° was identified as the lower slope limit of debris flow generated in weakly consolidated colluvium in Switzerland (Rickenmann and Zimmermann, 1993). E 28.5° - 54% - 28.5° describes the lower range of where debris covered slopes in <30° <58% Arizona start to become unstable (Melton, 1965). F 30° - <32° 58% - Two degree interval, no specific reference for a break. <63% G 32° - <34° 63% - Two degree interval, no specific reference for a break. <68% H 34° - <36° 68% - Debris flows generated in debris-filled gullies were documented <73% between slopes of 24° to 35° (Rickenmann and Zimmermann, 1993). 35° (70%) are the mean slopes of the 2006 hillslopes failures in Coronado National Memorial, Huachuca Mountains (Youberg, unpublished data). I 36° - <38° 73% - 36° is the upper range of slopes that maintain a debris cover in <78% Arizona (Melton, 1965). J 38° - <40° 78% - 38° slope upper limit for debris-flow generation in steep, weakly <84% consolidated talus slopes (Rickenmann and Zimmermann, 1993). K 40° - <42° 84% - 40° upper range of angle of repose (Van Burkalow, 1945; Lamb et <90% al., 2013). Maximum slopes that generated debris flows at Chalk Cliffs, Colorado are 45° (Coe et al., 2008). Greater than 40° are bare rock resembling cliffs (Melton, 1965). L ≥42° ≥90% Remaining hillslope area. Diffusive hillslope region? 55° (143%) are the highest slopes of the 2006 hillslopes failures in Coronado National Memorial, Huachuca Mountains (Youberg, unpublished data).

206

Table C2. Statistical tests of differences between nD and D basin areas for basin morphometric and burn severity factors. Basin Metrics t-test, 2-tailed Mann-Whitney Rank Sum Mean Percent Slopes P = <0.001 Percent of basin with slopes ≥ 30% P = <0.001 Percent of basin with slopes ≥ 50% P = <0.001 P = <0.001 Basin Relief P = 0.003 Melton Ratio P 0.007 P = 0.007 Relief Ratio P = 0.028 Shape P 0.431 P = 0.477 Elongation Ratio P = 0.055 Hypsometric Index P 0.009 P = 0.007 Mean Elevation P = <0.001 Basin Area P 0.462 Percent basin moderate soil burn severity P = 0.033 Percent basin high soil burn severity P = 0.119 Percent basin moderate + high soil burn P = 0.929 severity Percent basin moderate soil burn severity P = 0.171 on slopes ≥30% Percent basin high soil burn severity on P = 0.004 slopes ≥30% Percent basin moderate soil burn severity P = 0.271 on slopes ≥50% Percent basin high soil burn severity on P = <0.001 slopes ≥50% Percent basin moderate soil burn severity P = 0.983 on slopes ≥30° (58%) Percent basin moderate soil burn severity P = <0.001 on slopes ≥30° (58%)

207

Table C3. Statistical tests of differences between nD and D basin areas in selected slope intervals. Slope Slope – Slope _ Percent t-test, 2-tailed Mann-Whitney Intervals Degrees Rank Sum A 14° - < 16.6° 25% - <30% P =< 0.0001 B 16.6° - <20° P = 0.002 C 20° - <26.5° P = 0.690 D 26.5° - <28.5° 50% - <54% P = 0.168 E 28.5° - <30° 54% - <58% P = 0.003 F 30° - <32° 58% - <63% P =< 0.001 G 32° - <34° 63% - <68% P =<0.001 H 34° - <36° 68% - <73% P= <0.001 I 36° - <38° 73% - <78% P= 0.002 J 38° - <40° 78% - <84% P = 0.007 K 40° - 42° 84% - 90% P = 0.011 L ≥ 42° ≥ 90% P = 0.028 G_up ≥30° ≥58% P = <0.001

208

C.11 Figures

Figure C.1. Location of study wildfires.

209

Figure C.2. Mapped post-fire channel heads and saturation-induced hillslope scarps.

210

Figure C.3. Box plots of slope and basin relief metrics.

211

Figure C.4. Slope area plots of nD (black dots) and D (white dots) basins for each study fire.

212

Figure C.5. Scatterplots of nD and D basins for slope intervals A-F.

213

Figure C.6. Scatterplots of nD and D basins for slope intervals G-L.

214

Figure C.7. Box plots of categorical moderate final soil burn severity.

215

Figure C.8. Box plots of categorical high final soil burn severity.

216

Figure C.9. Percent basin area occupied by high soil burn severity calculated from categorical final soil burn severity data and continuous dNBR data.

217

Figure C.10. Percent basin area on slopes ≥30°, and with burn severity metrics.

218

Figure C.11. Slope-area plots with post-fire channel heads and saturation-induced hillslope failures.

219

REFERENCES

Abrahams, A.D., Parsons, A.J., and Wainwright, J., 1995, Effects of vegetation change on interrill runoff and erosion, Walnut Gulch, southern Arizona: Geomorphology, v. 13, no. 1-4, p. 37-48. Abramson, L., Chennell, M., Eisch, E., Glassco, A., and Holley, T., 2009, Post-Fire Sedimentation and Flood Risk Potential in the Mission Creek Watershed of Santa Barbara: Santa Barbara, California, University of California, Santa Barbara, 85 p. Anderson, C.A., and Blacet, P.M., 1972, Geologic map of the Mount Union Quadrangle, Yavapai County, Arizona: U.S. Geologcal Survey, Geologic Quadrangle Maps of the United States, Map GQ-997, scale 1:62,500. Anderson, R.S., Repka, J.L., and Dick, G.S., 1996, Explicit treatment of inheritance in dating depositional surfaces using in situ 10Be and 26Al: Geology, v. 24, no. 1, p. 47-51. Anderson, S.A., and Sitar, N., 1995, Analysis of Rainfall-Induced Debris Flows: Journal of Geotechnical Engineering, v. 121, no. 7, p. 544. Arundel, S.T., 2002, Modeling Climate Limits of Plants Found in Sonoran Desert Packrat Middens: Quaternary Research, v. 58, no. 2, p. 112-121. Balco, G., Stone, J.O., Lifton, N.A., and Dunai, T.J., 2008, A complete and easily accessible means of calculating surface exposure ages or erosion rates from 10Be and 26Al measurements: Quaternary Geochronology, v. 3, no. 3, p. 174-195, doi:doi:10.1016/j.quageo.2007.12.001. Ballenger, J.A.M., Holliday, V.T., Kowler, A.L., Reitze, W.T., Prasciunas, M.M., Miller, D.S., and Windingstad, J.D., 2011, Evidence for Younger Dryas global climate oscillation and human response in the American Southwest: Quaternary International, v. 242, no. 2, p. 502-519, doi:10.1016/j.quaint.2011.06.040. Banks, N.G., 1976, Reconnaissance geologic map of the Mount Lemmon quadrangle, Arizona: U.S. Geological Survey Miscellanious Field Studies Map MF-747, scale1:62,500. Bardou, E., and Jaboyedoff, M., 2008, Debris flows as a factor of hillslope evolution controlled by a continuous or a pulse process?: Geological Society, London, Special Publications, v. 296, no. 1, p. 63-78, doi:10.1144/sp296.5. Baum, R., and Godt, J., 2010, Early warning of rainfall-induced shallow landslides and debris flows in the USA: Landslides, v. 7, no. 3, p. 259-272, doi:10.1007/s10346-009-0177-0. Berti, M., and Simoni, A., 2007, Prediction of debris flow inundation areas using empirical mobility relationships: Geomorphology, v. 90, no. 1-2, p. 144-161. Betancourt, J.L., 2003, The Pleistocene-Holocene transition in subtropical North and South America, XVI INQUA congress; Shaping the Earth; a Quaternary perspective: Congress of the International Union for Quaternary Research: Reno, Nevada, Desert Research Institute, v. 16, p. 190. Bierman, P.R., 1994, Using in situ produced cosmogenic isotopes to estimate rates of landscape evolution: A review from the geomorphic perspective: Journal of Geophysical Research, v. 99, no. B7, p. 13885-13896, doi:10.1029/94JB00459. Bierman, P.R., Gillespie, A.R., and Caffee, M.W., 1995, Cosmogenic Ages for Earthquake Recurrence Intervals and Debris Flow Fan Deposition, Owens Valley, California: Science, v. 270, no. 5235, p. 447-450. Birkland, P.R., 1999, Soils and Geomorphology (Third ed.): New York, Oxford University Press, 430 p. Birkland, P.R., Machette, M.N., and Haller, K.M., 1991, Soils as a tool for applied Quaternary geology, Utah Geological and Mineral Survey Miscellaneous Publication 91-3, p. 63.

220

Bookhagen, B., 2007, Chemical separation of Al and Be from quartz-bearing rocks: Stanford, California, Stanford University, Department of Geosciences, p. 48. Bovis, M.J., and Jakob, M., 1999, The role of debris supply conditions in predicting debris flow activity: Earth Surface Processes and Landforms, v. 24, no. 11, p. 1039-1054. Bowers, J.E., Webb, R.H., and Pierson, E.A., 1997, Succession of desert plants on debris flow terraces, Grand Canyon, Arizona, U.S.A: Journal of Arid Environments, v. 36, no. 1, p. 67-86. Bradley, A.P., 1997, The use of the area under the ROC curve in the evaluation of machine learning algorithms: Pattern Recognition, v. 30, no. 7, p. 1145-1159, doi:http://dx.doi.org/10.1016/S0031-3203(96)00142-2. Bras, R.L., Tucker, G.E., and Teles, V., 2003, Six myths about mathematical modeling in geomorphology, in Wilcock Peter, R., and Iverson Richard, M., eds., Prediction in geomorphology.: Washington, DC, American Geophysical Union, p. 63-79. Bull, W.B., 1991, Geomorphic Response to Climate Change: New York, Oxford University Press. Caine, N., 1980, The Rainfall Intensity: Duration Control of Shallow Landslides and Debris Flows: Geografiska Annaler. Series A, Physical Geography, v. 62, no. 1/2, p. 23-27. Cannon, S.H., 2001, Debris-flow generation from recently burned watersheds: Environmental and Engineering Geoscience, v. 7, no. 4, p. 321-341. Cannon, S.H., Boldt, E.M., Laber, J.L., Kean, J.W., and Staley, D.M., 2011, Rainfall intensity–duration thresholds for postfire debris-flow emergency-response planning: Natural Hazards, v. 59, no. 1, p. 209-236, doi:10.1007/s11069-011-9747-2. Cannon, S.H., and Ellen, S., 1985, Abuntdant debris avalanches, San Francisco Bay Region, California: California Geology, v. 38, no. 12, p. 267-272. Cannon, S.H., and Gartner, J.E., 2005, Wildfire-related debris flow from a hazards perspective, in Jakob, M., and Hungr, O., eds., Debris-flow hazards and related phenomena: Berlin, Springer, p. 363- 385. Cannon, S.H., Gartner, J.E., Rupert, M.G., Michael, J.A., Rea, A.H., and Parrett, C., 2010, Predicting the probability and volume of postwildfire debris flows in the intermountain western United States: Geological Society of America Bulletin, v. 122, no. 1-2, p. 127-144, doi:10.1130/b26459.1. 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, v. 96, no. 3-4, p. 250-269. Cannon, S.H., Kirkham, R.M., and Parise, M., 2001, Wildfire-related debris-flow initiation processes, Storm King Mountain, Colorado: Geomorphology, v. 39, no. 3-4, p. 171-188. Cannon, S.H., and Reneau, S.L., 2000, Conditions for generation of fire-related debris flows, Capulin Canyon, New Mexico: Earth Surface Processes and Landforms, v. 25, no. 10, p. 1103-1121. Cawson, J.G., Sheridan, G.J., Smith, H.G., and Lane, P.N.J., 2013, Effects of fire severity and burn patchiness on hillslope-scale surface runoff, erosion and hydrologic connectivity in a prescribed burn: Forest Ecology and Management, v. 310, no. 0, p. 219-233, doi:http://dx.doi.org/10.1016/j.foreco.2013.08.016. Cerling, T.E., Webb, R.H., Poreda, R.J., Rigby, A.D., and Melis, T.S., 1999, Cosmogonic 3He ages and frequency of late Holocene debris flows from Prospect Canyon, Grand Canyon, USA: Geomorphology, v. 27, no. 1-2, p. 93-111. 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, v. 96, no. 3-4, p. 270-297.

221

Coelho, C.d.O.A., Ferreira, A.J.D., Boulet, A.-K., and Keizer, J.J., 2004, Overland flow generation processes, erosion yields and solute loss following different intensity fires: Quarterly Journal of Engineering Geology Hydrogeology, v. 37, no. 3, p. 233-240, doi:10.1144/1470-9236/03-043. Costa, J.E., 1984, Physical geomorphology of debris flows, in Costa John, E., and Fleisher, P.J., eds., Developments and applications of geomorphology.: Berlin, Federal Republic of Germany, Springer-Verlag, p. 268-317. ---, 1988, Rheologic, geomorphic, and sedimentologic differentiation of water floods, hyperconcentrated flows, and debris flows, in Baker Victor, R., Kochel, R.C., and Patton Peter, C., eds., Flood geomorphology.: New York, NY, United States, John Wiley & Sons, p. 113-122. Costa, J.E., and Williams, G.P., 1984, Debris-flow dynamics: Open-File Report 84-606 (video): Reston, U.S. Geological Survey, Water Resources Division. DeBano, L.F., 2000, The role of fire and soil heating on water repellency in wildland environments: a review: Journal of Hydrology, v. 231-232, p. 195-206. DeLong, S.B., Henderson, W., and Youberg, A., 2012, Post wildfire landscape change in a small headwater catchment from terrestrial LiDAR, in Southwest Wildfire Hydrology & Hazard Workshop Proceedings, Biosphere 2, Tucson, Arizona, Arizona Geological Survey, p. 20. Dickenson, W.R., 1991, Tectonic setting of faulted Tertiary strata associated with the Catalina core complex in southern Arizona: Boulder, Colorado, Geological Society of America Special Paper 264, v. Special Paper 264, p. 106. ---, 1999, Geologic framework of the Catalina foothills, outskirts of Tucson (Pima County, Arizona): Tucson, Arizona Geological Survey Contributed Map CM-99-B, p. 31. Doerr, S.H., Shakesby, R.A., Blake, W.H., Chafer, C.J., Humphreys, G.S., and Wallbrink, P.J., 2006, Effects of differing wildfire severities on soil wettability and implications for hydrological response: Journal of Hydrology, v. 319, no. 1-4, p. 295-311, doi:10.1016/j.jhydrol.2005.06.038. Drewes, H., 1996, Geologic maps of the Coronado National Forest, southeast Arizona and southwest New Mexico: US Geological Survey Bulletin, p. 17-41. Dühnforth, M., Densmore, A.L., Ivy-Ochs, S., Allen, P.A., and Kubik, P.W., 2007, Timing and patterns of debris flow deposition on Shepherd and Symmes creek fans, Owens Valley, California, deduced from cosmogenic 10Be: J. Geophys. Res., v. 112, no. F3, p. F03S15, doi:10.1029/2006jf000562. Ebel, B.A., Moody, J.A., and Martin, D.A., 2012, Hydrologic conditions controlling runoff generation immediately after wildfire: Water Resources Research, v. 48, no. 3, p. W03529, doi:10.1029/2011wr011470. Eng, J., 2005, Receiver Operating Characteristic Analysis: A Primer1: Academic Radiology, v. 12, no. 7, p. 909-916, doi:http://dx.doi.org/10.1016/j.acra.2005.04.005. Eppes, M.C., and McFadden, L., 2008, The influence of bedrock weathering on the response of drainage basins and associated alluvial fans to Holocene climates, San Bernardino Mountains, California, USA: The Holocene, v. 18, no. 6, p. 895-905. Falk, D.A., 2013, Are Madrian ecosystems approaching tipping points? Anticipating interactions of landscape disturbance and climate change, in Merging science and management in a rapidly changing : biodiversity and management of the Madrean Archipelago III and 7th Conference on Research adn Resource Management in the Southwestern Deserts, Tucson, AZ, Proceedings. RMRS-P_67. Fort Collins, CO: Department of Agriculture, Forest Service, Rocky Mountain Research Station, p. 40-47. Fawcett, T., 2006, An introduction to ROC analysis: Pattern Recognition Letters, v. 27, no. 8, p. 861-874, doi:http://dx.doi.org/10.1016/j.patrec.2005.10.010.

222

Fenton, C.R., and Pelletier, J.D., 2013, Cosmogenic 3He age estimates of Plio-Pleistocene alluvial-fan surfaces in the Lower Colorado River Corridor, Arizona, USA: Quaternary Research, v. 79, no. 1, p. 86-99, doi:10.1016/j.yqres.2012.10.006. Ferguson, C.A., Cook, J.P., and Pearthree, P.A., 2012, Geologic map of the White Hills West 7 ½' Quadrangle, Mohave County, Arizona: Arizona Geological Survey DGM-97, v 1.0, map scale 1:24,000 Fiorillo, F., and Wilson, R.C., 2004, Rainfall induced debris flows in pyroclastic deposits, Campania (southern Italy): Engineering Geology, v. 75, no. 3-4, p. 263-289. Force, E.R., 1997, Geology and mineral resources of the Santa Catalina Mountains, southeastern Arizona: Tucson, Center for Mineral Resources, 135 p. Frattini, P., Crosta, G., and Sosio, R., 2009, Approaches for defining thresholds and return periods for rainfall-triggered shallow landslides: Hydrological Processes, v. 23, no. 10, p. 1444-1460. Frechette, J.D., and Meyer, G.A., 2009, Holocene fire-related alluvial-fan deposition and climate in ponderosa pine and mixed-conifer forests, Sacramento Mountains, New Mexico, USA: The Holocene, v. 19, no. 4, p. 639-651, doi:10.1177/0959683609104031. Friedman, E., 2012, Debris-Flow Hazard Assessment for the 2010 Medano Pass Fire Burn Area, Great Sand Dunes National Park and Preserve, Colorado: Golden, CO, Colorado School of Mines, 93 p. Fule, P.Z., Yocom, L.L., Montano, C.C., Falk, D.A., Cerano, J., and Villanueva-Diaz, J., 2012, Testing a pyroclimatic hypothesis on the Mexico-United States border: Ecology, v. 93, no. 8, p. 1830-1840. Gabet, E.J., 2003, Post-fire thin debris flows: sediment transport and numerical modeling: Earth Surface Processes and Landforms, v. 28, no. 12, p. 1341-1348. 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, v. 96, no. 3-4, p. 298-309. Garfin, G., Jardine, A., Merideth, R., Black, M., and LeRoy, S., eds., 2013, Assessment of Climate Change in the Southwestern United States; A Report Prepared for the National Climate Assessment. A report by the Southwest Climate Alliance: Washington, DC, Island Press. Gartner, J.E., Bigio, E.R., and Cannon, S.H., 2004, Compilation of post wildfire runoff-event data from the Western United States, U. S. Geological Survey. Reston, VA, United States. Gerber, M.B., and Roering, J.J., 2003, Decline in root strength following wildfire in the Oregon Coast Range, Abstracts with Programs - Geological Society of America, Geological Society of America, p. 352. Gile, L.H., Hawley, J.W., and Grossman, R.B., 1981, Soils and geomorphology in the basin and range area of southern New Mexico - guidebook to the Desert Project: Soccoro, New Mexico Bureau of Mines and Mineral Resources, 222 p. Giraud, R.E., 2005, Guidelines for the geologic evaluation of debris-flow hazards on alluvial fans in Utah: Utah Geological Survey Miscellaneous Publication 05-6, 16 p. Glade, T., 2005, Linking debris-flow hazard assessments with geomorphology: Geomorphology, v. 66, no. 1-4, p. 189-213. Godt, J.W., Baum, R.L., and Chleborad, A.F., 2006, Rainfall characteristics for shallow landsliding in Seattle, Washington, USA: Earth Surface Processes and Landforms, v. 31, no. 1, p. 97-110. Godt, J.W., and Coe, J.A., 2007, Alpine debris flows triggered by a 28 July 1999 thunderstorm in the central Front Range, Colorado: Geomorphology, v. 84, no. 1-2, p. 80-97. Godt, J.W., Schulz, W.H., Baum, R.L., and Savage, W.Z., 2008, Modeling Rainfall Conditions For Shallow Landsliding In Seattle, Washington: Reviews in Engineering Geology Volume XX: Landslides and Engineering Geology of the Seattle, Washington, Area, v. 20, no. 0, p. 137-152.

223

Gosse, J.C., and Phillips, F.M., 2001, Terrestrial in situ cosmogenic nuclides: theory and application: Quaternary Science Reviews, v. 20, no. 14, p. 1475-1560, doi:10.1016/s0277-3791(00)00171-2. Griffiths, P.G., Magirl, C.S., Webb, R.H., Pytlak, E., Troch, P.A., and Lyon, S.W., 2009, Spatial distribution and frequency of precipitation during an extreme event: July 2006 mesoscale convective complexes and floods in southeastern Arizona: Water Resourses Research, v. 45, p. W07419, doi:10.1029/2008wr007380. Griffiths, P.G., Webb, R.H., and Melis, T.S., 1996, Initiation and frequency of debris flows in Grand Canyon, Arizona, U. S. Geological Survey Open-File Report 96-491: Reston, VA, United States, U. S. Geological Survey, p. 35. ---, 1997, Shales and debris-flow frequency on the Colorado Plateau, USA, in Geological Society of America, 1997 annual meeting, Abstracts with Programs - Geological Society of America, Geological Society of America, p. 139. ---, 2004, Frequency and initiation of debris flows in Grand Canyon, Arizona: J. Geophys. Res., v. 109, no. F4, p. F04002, doi:10.1029/2003jf000077. Grissino-Mayer, H.D., and Swetnam, T.W., 2000, Century-scale climate forcing of fire regimes in the American Southwest: The Holocene, v. 10, no. 2, p. 213-220. Griswold, J.P., and Iverson, R.M., 2008, Mobility statistics and automated hazard mapping for debris flows and rock avalanches [electronic resource], U.S Geological Survey Scientific Investigations Report 2007-5276 (version 1.0 ed.): Reston, Virginia, U.S. Geological Survey, p. 59. Guzzetti, F., Peruccacci, S., Rossi, M., and Stark, C., 2008, The rainfall intensity–duration control of shallow landslides and debris flows: an update: Landslides, v. 5, no. 1, p. 3-17. Hand, D.J., 2010, Evaluating diagnostic tests: The area under the ROC curve and the balance of errors: Statistics in Medicine, v. 29, no. 14, p. 1502-1510, doi:10.1002/sim.3859. ---, 2012, Assessing the Performance of Classification Methods: International Statistical Review, v. 80, no. 3, p. 400-414, doi:10.1111/j.1751-5823.2012.00183.x. Hanks, T.C., and Webb, R.H., 2006, Effects of tributary debris on the longitudinal profile of the Colorado River in Grand Canyon: J. Geophys. Res., v. 111. Harvey, A.M., Wigand, P.E., and Wells, S.G., 1999, Response of alluvial fan systems to the late Pleistocene to Holocene climatic transition: contrasts between the margins of pluvial Lakes Lahontan and Mojave, Nevada and California, USA: koCatena, v. 36, no. 4, p. 255-281, doi:10.1016/s0341-8162(99)00049-1. Hayes, P.T., and Raup, R.B., 1968, Geologic map of the Huachuca and Mustang Mountains, southeastern Arizona: U.S. Geologica Survey, Miscellaneous Geologic Invstigations, Map I-509, scale 1:48,000. Heede, B., 1985, Channel adjustments to the removal of log steps: an experiment in a mountain stream: Environmental Management, v. 9, no. 5, p. 427-432, doi:10.1007/bf01866341. Hereford, R., Thompson, K.S., and Burke, K.J., 1998, Numerical Ages of Holocene Tributary Debris Fans Inferred from Dissolution Pitting on Carbonate Boulders in the Grand Canyon of Arizona: Quaternary Research, v. 50, no. 2, p. 139-147, doi:10.1006/qres.1998.1987. Hereford, R., Thompson, K.S., Burke, K.J., and Fairley, H.C., 1996, Tributary debris fans and the late Holocene alluvial chronology of the Colorado River, eastern Grand Canyon, Arizona: Geological Society of America Bulletin, v. 108, no. 1, p. 3-19. Holm, R.F., 1988, Geologic map of San Francisco Mountain, Elden Mountain, and Dry Lake Hills, Coconino County, Arizona: U.S. Geological Survey, Miscellaneous Investigations Series Map I- 1663, scale 1:24000.

224

Holmgren, C.A., Betancourt, J.L., and Rylander, K.A., 2006, A 36,000-yr vegetation history from the Peloncillo Mountains, southeastern Arizona, USA: Palaeogeography, Palaeoclimatology, Palaeoecology. , v. 240, no. 3-4, p. 405-422. Holmgren, C.A., Norris, J., and Betancourt, J.L., 2007, Inferences about winter temperatures and summer rains from the late Quaternary record of C-4 perennial grasses and C-3 desert shrubs in the northern Chihuahuan Desert: Journal of Quaternary Science, v. 22, no. 2, p. 141-161, doi:10.1002/jqs.1023. Hrachowitz, M., Savenije, H.H.G., Blöschl, G., McDonnell, J.J., Sivapalan, M., Pomeroy, J.W., Arheimer, B., Blume, T., Clark, M.P., Ehret, U., Fenicia, F., Freer, J.E., Gelfan, A., Gupta, H.V., Hughes, D.A., Hut, R.W., Montanari, A., Pande, S., Tetzlaff, D., Troch, P.A., Uhlenbrook, S., Wagener, T., Winsemius, H.C., Woods, R.A., Zehe, E., and Cudennec, C., 2013, A decade of Predictions in Ungauged Basins (PUB)—a review: Hydrological Sciences Journal, v. 58, no. 6, p. 1198-1255, doi:10.1080/02626667.2013.803183. Hungr, O., 1995, A model for the runout analysis of rapid flow slides, debris flows, and avalanches: Canadian Geotechnical Journal, v. 32, no. 4, p. 610-623, doi:10.1139/t95-063. Hungr, O., McDougall, S., Wise, M., and Cullen, M., 2008, Magnitude-frequency relationships of debris flows and debris avalanches in relation to slope relief: Geomorphology, v. 96, no. 3-4, p. 355- 365. Ijjasz-Vasquez, E.J., and Bras, R.L., 1995, Scaling regimes of local slope versus contributing area in digital elevation models: Geomorphology, v. 12, no. 4, p. 299-311, doi:http://dx.doi.org/10.1016/0169- 555X(95)00012-T. Istanbulluoglu, E., Tarboton, D.G., Pack, R.T., and Luce, C., 2002, A probabilistic approach for channel initiation: Water Resour. Res., v. 38, no. 12, p. 1325, doi:10.1029/2001wr000782. Istanbulluoglu, E., Tarboton, D.G., Pack, R.T., and Luce, C.H., 2004, Modeling of the interactions between forest vegetation, disturbances, and sediment yields: J. Geophys. Res., v. 109, no. F1, p. F01009, doi:10.1029/2003jf000041. Iverson, R.H., 1997, The physics of debris flows: Reviews of , v. 35, no. 3, p. 245-296. Iverson, R.M., 2000, Landslide triggering by rain infiltration: Water Resources Research, v. 36, no. 7, p. 1897-1910. Iverson, R.M., 2003a, The debris-flow rheology myth, in Rickenmann, D., and Chen, C.-l., 3rd International Conference on Debris-Flow Hazards Mitigation: Mechanics, Prediction, and Assessment: Davos, Switzerland, American Society of Civil Engineers, v. 1, p. 303-314. Iverson, R.M., 2003b, How should mathematical models of geomorphic processes be judged?, in Wilcock Peter, R., and Iverson Richard, M., eds., Prediction in geomorphology.: Washington, DC, American Geophysical Union, p. 83-94. Iverson, R.M., Logan, M., LaHusen, R.G., and Berti, M., 2010, The perfect debris flow? Aggregated results from 28 large-scale experiments: J. Geophys. Res., v. 115, no. F3, p. F03005, doi:10.1029/2009jf001514. Iverson, R.M., Reid, M.E., and LaHusen, R.G., 1997, Debris-flow mobilization from landslides: Annual Review of Earth and Planetary Sciences, v. 25, p. 85-138. Iverson, R.M., Reid, M.E., Logan, M., LaHusen, R.G., Godt, J.W., and Griswold, J.P., 2011, Positive feedback and momentum growth during debris-flow entrainment of wet bed sediment: Nature Geosci, v. 4, no. 2, p. 116-121, doi:http://www.nature.com/ngeo/journal/v4/n2/abs/ngeo1040.html#supplementary- information.

225

Iverson, R.M., Schilling, S.P., and Vallance, J.W., 1998, Objective delineation of lahar-inundation hazard zones: Geological Society of America Bulletin, v. 110, no. 8, p. 972-984, doi:10.1130/0016-7606. Iverson, R.M., and Vallance, J.W., 2001, New views of granular mass flows: Geology, v. 29, no. 2, p. 115- 118, doi:10.1130/0091-7613(2001)029<0115:nvogmf>2.0.co;2. Jakob, M., 2005a, Debris-flow hazard analysis, in Jakob, M., and Hungr, O., eds., Debris-flow Hazards and Related Phenomena: Springer-Praxis, p. 411-443. ---, 2005b, A size classification for debris flows: Engineering Geology, v. 79, no. 3-4, p. 151-161. Jenkins, S.E., 2007, Fire-Related Deposition at Kendrick Mountain, Arizona: Characterization and Implications for Fire History Reconstructions: Flagstaff, Northern Arizona University, 139 p. Jenkins, S.E., Sieg, C.H., Anderson, D.E., Kaufman, D.S., and Pearthree, P.A., 2011, Late Holocene geomorphic record of fire in ponderosa pine and mixed-conifer forests, Kendrick Mountain, northern Arizona, USA: International Journal of Wildland Fire, v. 20, no. 1, p. 125-141. Johnson, A.M., and Rodine, J.R., 1984, Debris flow, in Brunsden, D., and Prior, D.B., eds., Slope Instability: Chichester, Sussex England, John Wiley and Sons Ltd., p. 257-361. Johnson, B.J., Pearthree, P.A., and Ferguson, C.A., 2011, Geologic map of the Middle Water Spring 7 ½' Quadrangle, Mohave County, Arizona: Arizona Geological Survey DGM-96, v 1.0, map scale 1:24,000 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, p. 2013JF002796, doi:10.1002/jgrf.20148. Kean, J.W., and Staley, D.M., 2011, Direct measurement of the hydrologic conditions leading up to and during post-fire debris flow in southern California, USA, in Genevois, R., Hamilton, D.L., and Prestininzi, A., Debris‐Flow Hazards Mitigation, Mechanics, Prediction, and Assessment: Casa Editrice Univ. La Sapienza, Rome, p. 685-694. 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: J. Geophys. Res., v. 116, no. F4, p. F04019, doi:10.1029/2011jf002005. Keith, S.B., Reynolds, S.J., Damon, P.E., Shafiqullah, M., Livingston, D.E., and Pushkar, P.D., 1980, Evidence for multiple intrusion and deformation within the Santa Catalina-Rincon-Tortolita crystalline complex, southeastern Arizona, in Crittenden, M.D., Jr., Coney, P.J., and Davis, G.H., eds., Cordilleran Metamorphic Core Complexes: Boulder, CO, Geological Society of America, Memoir 153, p. 217-267. Kirchner, J.W., Finkel, R.C., Riebe, C.S., Granger, D.E., Clayton, J.L., King, J.G., and Megahan, W.F., 2001, Mountain erosion over 10 yr, 10 k.y., and 10 m.y. time scales: Geology, v. 29, no. 7, p. 591-594, doi:10.1130/0091-7613(2001)029<0591:meoyky>2.0.co;2. Klawon, J.E., Dickinson, W.R., and Pearthree, P.A., 1999, Surficial Geology and Geologic Hazards of the Northern Tucson Basin, Pima County, Arizona, Tucson North and Sabino Canyon Quadrangles: Arizona Geological Survey, 28 p, scale 1:24,000. Kowler, A.L., Bright, J., Quade, J., and Holliday, V.T., 2012, Latest Pleistocene paleohydrology of the Willcox Basin, southeastern Arizona, and implications for Southwest paleoclimate, Geological Society of America Abstracts with Programs: Charlotte, NC, v. 44, no. 7, p. 187. Lamb, M.P., Levina, M., DiBiase, R.A., and Fuller, B.M., 2013, Sediment storage by vegetation in steep bedrock landscapes: Theory, experiments, and implications for postfire sediment yield: Journal of Geophysical Research: Earth Surface, v. 118, no. 2, p. 1147-1160, doi:10.1002/jgrf.20058.

226

Larsen, I.J., MacDonald, L.H., Brown, E., Rough, D., Welsh, M.J., Pietraszek, J.H., Libohova, Z., de Dios Benavides-Solorio, J., and Schaffrath, K., 2009, Causes of Post-Fire Runoff and Erosion: Water Repellency, Cover, or Soil Sealing?: Soil Sci. Soc. Am. J., v. 73, no. 4, p. 1393-1407, doi:10.2136/sssaj2007.0432. Lifton, N.A., Bieber, J.W., Clem, J.M., Duldig, M.L., Evenson, P., Humble, J.E., and Pyle, R., 2005, Addressing solar modulation and long-term uncertainties in scaling secondary cosmic rays for in situ cosmogenic nuclide applications: Earth and Planetary Science Letters, v. 239, p. 140-161. Littell, J.S., McKenzie, D., Peterson, D.L., and Westerling, A.L., 2009, Climate and wildfire area burned in western U.S. ecoprovinces, 1916–2003: Ecological Applications, v. 19, no. 4, p. 1003-1021, doi:10.1890/07-1183.1. Lyle, M., Heusser, L., Ravelo, C., Yamamoto, M., Barron, J., Diffenbaugh, N.S., Herbert, T., and Andreasen, D., 2012, Out of the Tropics: The Pacific, Great Basin Lakes, and Late Pleistocene Water Cycle in the Western United States: Science, v. 337, no. 6102, p. 1629-1633, doi:10.1126/science.1218390. Machette, M.N., 1985, Calcic soils of the southwestern United States, in Weide David, L., and Faber Marianne, L., eds., Soils and Quaternary geology of the Southwestern United States.: Boulder, CO, United States, Geological Society of America (GSA), p. 1-21. Magirl, C.S., Griffiths, P.G., and Webb, R.H., 2010, Analyzing debris flows with the statistically calibrated empirical model LAHARZ in southeastern Arizona, USA: Geomorphology, v. 119, no. 1-2, p. 111- 124, doi:DOI: 10.1016/j.geomorph.2010.02.022. Magirl, C.S., Webb, R.H., Griffiths, P.G., Schaffner, M., Shoemaker, C., Pytlak, E., Yatheendradas, S., Lyon, S.W., Troch, P.A., Dislets, S.L.E., Goodrich, D.C., Unkrich, C.L., Youberg, A., and Pearthree, P.A., 2007, Impacts of recent extreme Arizona Storms: Eos, Transactions, American Geophysical Union, v. 88, no. 17. Matmon, A., Nichols, K., and Finkel, R., 2006, Isotopic insights into smoothening of abandoned fan surfaces, Southern California: Quaternary Research, v. 66, no. 1, p. 109-118, doi:DOI: 10.1016/j.yqres.2006.02.010. Matmon, A., Schwartz, D.P., Finkel, R., Clemmens, S., and Hanks, T., 2005, Dating offset fans along the Mojave section of the San Andreas fault using cosmogenic 26Al and 10Be: Geological Society of America Bulletin, v. 117, no. 5-6, p. 795-807, doi:10.1130/b25590.1. 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, v. 38, no. 8, p. 735-738, doi:10.1130/g30928.1. McCoy, S.W., Kean, J.W., Coe, J.A., Tucker, G.E., Staley, D.M., and Wasklewicz, T.A., 2012, Sediment entrainment by debris flows: In situ measurements from the headwaters of a steep catchment: Journal of Geophysical Research: Earth Surface, v. 117, no. F3, p. F03016, doi:10.1029/2011jf002278. McDonald, E.V., McFadden, L.D., and Wells, S.G., 2003, Regional response of alluvial fans to the Pleistocene-Holocene climatic transition, Mojave Desert, California: Special Paper 368: Paleoenvironments and paleohydrology of the Mojave and southern Great Basin deserts, v. 368, no. 0, p. 189-205. McKinley, R.A., 2013, Wildfire burn severity assessments from satellite data: a review of available products and user considerations, in Moody, J.A., and Martin, D.A., Synthesizing Empirical Results to Improve Predictions of Post-wildfire Runoff and Erosion Responses: Estes Park, Colorado, AGU Chapman Conference.

227

Melis, T.S., Webb, R.H., and Griffiths, P.G., 1997, Debris flows in Grand Canyon National Park; peak discharges, flow transformations, and hydrographs, in Debris-flow hazards mitigation; mechanics, prediction and assessment., San Francisco, California, American Society of Civil Engineers, p. 727-736. Melis, T.S., Webb, R.H., Griffiths, P.G., and Wise, T.W., 1994, Magnitude and frequency data for historic debris flows in Grand Canyon National Park and vicinity, Arizona, U. S. Geological Water- Resources Investigations Report 94-4214: Reston, VA, United States, U. S. Geological Survey, p. 285. Melis, T.S., Webb, R.H., and Wise, T.W., 1993, Effect of historic debris flows on the Colorado River in Grand Canyon, in Geological Society of America, 1993 annual meeting., Abstracts with Programs - Geological Society of America, Geological Society of America, p. 142. Melton, M.A., 1965a, Debris-covered hillslopes of the southern Arizona desert: consideration of their stability and sediment contribution: The Journal of Geology, v. 73, no. 5, p. 715-729. ---, 1965b, The geomorphic and paleoclimatic significance of alluvial deposits in southern Arizona: The Journal of Geology, v. 73, no. 1, p. 1-38. Meyer, G.A., Wells, S.G., and Jull, A.J.T., 1995, Fire and alluvial chronology in Yellowstone National Park; climatic and intrinsic controls on Holocene geomorphic processes: Geological Society of America Bulletin, v. 107, no. 10, p. 1211-1230. Miller, J.D., and Thode, A.E., 2007, Quantifying burn severity in a heterogeneous landscape with a relative version of the delta Normalized Burn Ratio (dNBR): Remote Sensing of Environment, v. 109, no. 1, p. 66-80, doi:http://dx.doi.org/10.1016/j.rse.2006.12.006. Montgomery, D.R., 2001, Slope Distributions, Threshold Hillslopes, and Steady-state Topography: American Journal of Science, v. 301, no. 4-5, p. 432-454, doi:10.2475/ajs.301.4-5.432. Montgomery, D.R., and Dietrich, W.E., 1994, A physically based model for the topographic control on shallow landsliding: Water Resources Research, v. 30, no. 4, p. 1153-1171. Montgomery, D.R., and Foufoula-Georgiou, E., 1993, Channel network source representation using digital elevation models: Water Resour. Res., v. 29, no. 12, p. 3925-3934, doi:10.1029/93wr02463. Montgomery, D.R., Schmidt, K.M., Dietrich, W.E., and McKean, J., 2009, Instrumental record of debris flow initiation during natural rainfall: Implications for modeling slope stability: J. Geophys. Res., v. 114. Montgomery, D.R., Schmidt, K.M., Greenberg, H.M., and Dietrich, W.E., 2000, Forest clearing and regional landsliding: Geology (Boulder), v. 28, no. 4, p. 311-314. Moody, J.A., and Ebel, B.A., 2012, Hyper-dry conditions provide new insights into the cause of extreme floods after wildfire: CATENA, v. 93, no. 0, p. 58-63, doi:10.1016/j.catena.2012.01.006. ---, 2013, Infiltration and runoff generation processes in fire-affected soils: Hydrological Processes, p. n/a-n/a, doi:10.1002/hyp.9857. 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, v. 379, no. 3-4, p. 291-303, doi:DOI: 10.1016/j.jhydrol.2009.10.015. Moody, J.A., and Martin, D.A., 2001, Initial hydrologic and geomorphic response following a wildfire in the Colorado Front Range: Earth Surface Processes and Landforms, v. 26, no. 10, p. 1049-1070. Moody, J.A., and Martin, D.A., 2009, Synthesis of sediment yields after wildland fire in different rainfall regimes in the western United States: International Journal of Wildland Fire, v. 18, no. 1, p. 96- 115.

228

Moody, J.A., Martin, D.A., Haire, S.L., and Kinner, D.A., 2008, Linking runoff response to burn severity after a wildfire: Hydrological Processes, v. 22, no. 13, p. 2063-2074. Moody, J.A., and Nyman, P., 2012, Variations in Soil Detachment Rates after Wildfire as a Function of Soil Depth, Flow Properties and Root Properties, U.S. Geological Survey Scientific Investigations Report 2012-5233, p. 40. Moody, J.A., Shakesby, R.A., Robichaud, P.R., Cannon, S.H., and Martin, D.A., 2013, Current Research Issues Related to Post-wildfire Runoff and Erosion Processes: Earth-Science Reviews, v. 122, p. 10-37, doi:http://dx.doi.org/10.1016/j.earscirev.2013.03.004. 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. Nations, D., and Stump, E., 1981, Geology of Arizona: Debuque, Iowa, Kendall/Hunt Publishing Company. Neary, D.G., Ryan, K.C., and DeBano, L.F., 2005, Wildland Fire in Ecosystems: Effects of Fire on Soil and Water ( Revised 2008 ed.): Fort Collins, CO, USDA Forest Service, Rocky Mountain Research Station, General Technical Report RMRS-GTR-42, v. 4, p. 250. Nichols, K.K., Bierman, P.R., Foniri, W.R., Gillespie, A.R., Caffee, M.W., and Finkel, R., 2006, Dates and rates of arid region geomorphic processes: GSA Today, v. 16, no. 8, p. 4-11, doi:10.1130/GSAT01608.1. Nyman, P., Sheridan, G.J., Moody, J.A., Smith, H.G., Noske, P.J., and Lane, P.N.J., 2013, Sediment availability on burned hillslopes: Journal of Geophysical Research: Earth Surface, p. 2012JF002664, doi:10.1002/jgrf.20152. Nyman, P., Sheridan, G.J., Smith, H.G., and Lane, P.N.J., 2011, Evidence of debris flow occurrence after wildfire in upland catchments of south-east Australia: Geomorphology, v. 125, no. 3, p. 383-401, doi:10.1016/j.geomorph.2010.10.016. O'Brien, J.S., Julien, P.Y., and Fullerton, W.T., 1993, Two-Dimensional Water Flood and Mudflow Simulation: Journal of Hydraulic Engineering, v. 119, no. 2, p. 244-261. Pallister, J.S., and du Bray, E.A., 1992, Geologic map of the Fife Peak quadrangle, Cochise County, Arizona, in Survey, U.S.G.: Reston, VA, U.S. Geological Survey Open-File Report 92-0242, p. 9 p., 2 sheets, scale 1:24,000. ---, 1994, Geologic map of the Fife Peak quadrangle, Cochise County, Arizona: U.S. Geological Survey Quadrangle Map GQ-1708, 1 sheet, scale 1:24,000. Parise, M., and Cannon, S., 2012, Wildfire impacts on the processes that generate debris flows in burned watersheds: Natural Hazards, v. 61, no. 1, p. 217-227, doi:10.1007/s11069-011-9769-9. Parrett, C., Cannon, S.H., and Pierce, K.L., 2004, Wildfire-related floods and debris flows in Montana in 2000 and 2001, U. S. Geological Survey. [Reston, VA], United States. Pages: 22. 2004. Parsons, A., Robichaud, P.R., Lewis, S.A., Napper, C., and Clark, J.T., 2010, Field guide for mapping post- fire soil burn severity. Gen. Tech. Rep. RMRS-GTR-243: Fort Collins, CO, US Department of Agriculture, Forest Service, Rocky Mountain Research Station. Parsons, A.J., Abrahams, A.D., and Wainwright, J., 1996, Responses of interrill runoff and erosion rates to vegetation change in southern Arizona: Geomorphology, v. 14, no. 4, p. 311-317, doi:10.1016/0169-555x(95)00044-6. Pearthree, P.A., 2004, Geologic Map of the Huachuca City 7½' Quadrangle, Cochise County, Arizona: Arizona Geological Survey, scale 1:24000. Pearthree, P.A., and Youberg, A., 2004, Fire and Sediment Deposition: Arizona Geology, v. 34, no. 3, p. 1- 2. ---, 2006, Recent debris flows and floods in southern Arizona: Arizona Geology, v. 36, no. 3, p. 1-5.

229

Pearthree, P.A., Young, J.J., and Cook, J.P., 2012, Surficial Geology and Flood Hazards on the Western Piedmont of the Maricopa Mountains and the Southern Piedmont of the Buckeye Hills, Maricopa County, Arizona: Arizona Geological Survey Open-File Report-12-07, 23 p. Pelletier, J.D., Barron-Gafford, G.A., Breshears, D.D., Brooks, P.D., Chorover, J., Durcik, M., Harman, C.J., Huxman, T.E., Lohse, K.A., Lybrand, R., Meixner, T., McIntosh, J.C., Papuga, S.A., Rasmussen, C., Schaap, M., Swetnam, T.L., and Troch, P.A., 2013, Coevolution of nonlinear trends in vegetation, soils, and topography with elevation and slope aspect: A case study in the sky islands of southern Arizona: Journal of Geophysical Research: Earth Surface, v. 118, no. 2, p. 741-758, doi:10.1002/jgrf.20046. Pelletier, J.D., Engelder, T., Comeau, D., Hudson, A., Leclerc, M., Youberg, A., and Diniega, S., 2009, Tectonic and structural control of fluvial channel morphology in metamorphic core complexes: The example of the Catalina-Rincon core complex, Arizona: Geosphere, v. 5, no. 4, p. 1-22, doi:doi:10.1130/GES00221.1. Pelletier, J.D., and Rasmussen, C., 2009a, Geomorphically based predictive mapping of soil thickness in upland watersheds: Water Resources Research, v. 45, no. 9, p. W09417, doi:10.1029/2008wr007319. ---, 2009b, Quantifying the climatic and tectonic controls on hillslope steepness and erosion rate: , v. 1, no. 2, p. 73-80, doi:10.1130/l3.1. Pierce, J.L., Meyer, G.A., and Jull, A.J.T., 2004, Fire-induced erosion and millennial-scale climate change in northern ponderosa pine forests: Nature (London), v. 432, no. 7013, p. 87-90. Pierson, T.C., 2005a, Distinguishing between debris flows and floods from field evidence in small watersheds: Reston, VA, U. S. Geological Survey, 4 p. ---, 2005b, Hyperconcentrated flow; transitional process between water flow and debris flow, in Jakob, M., and Hungr, O., eds., Debris-flow hazards and related phenomena: Berlin, Springer. Pierson, T.C., and Costa, J.E., 1987, A rheologic classification of subaerial sediment-water flows, in Costa John, E., and Wieczorek Gerald, F., eds., Debris flows/ avalanches; process, recognition, and mitigation: Boulder, Geological Society of America, p. 1-12. Pigati, J.S., Bright, J.E., Shanahan, T.M., and Mahan, S.A., 2009, Late Pleistocene paleohydrology near the boundary of the Sonoran and Chihuahuan Deserts, southeastern Arizona, USA: Quaternary Science Reviews, v. 28, no. 3-4, p. 286-300. Pitzl, M.J., Loomis, B., and Dempsey, M., 2013, In Harm's Way, The Arizona Republic: , AZ. Restrepo-Posada, P.J., and Eagleson, P.S., 1982, Identification of independent rainstorms: Journal of Hydrology, v. 55, no. 1–4, p. 303-319, doi:10.1016/0022-1694(82)90136-6. Richard, S.M., Reynolds, S.J., Spencer, J.E., and Pearthree, P.A., 2000, Geologic Map of Arizona: Arizona Geolgocial Survey Map M-35, 1 sheet, scale 1:1,000,000. Rickenmann, D., 1999, Empirical Relationships for Debris Flows: Natural Hazards, v. 19, no. 1, p. 47-77. ---, 2005, Runout prediction methods, in Jakob, M., and Hungr, O., eds., Debris-flow Hazards and Related Phenomena: Chichester, UK Praxis. Rickenmann, D., and Zimmermann, M., 1993, The 1987 debris flows in Switzerland: documentation and analysis: Geomorphology, v. 8, no. 2-3, p. 175-189. Rihs, J., 2011, Wallow Wildfire BAER Resource Report, Apache-Sitgreaves National Forest, Hydrology, US Department of Agriculture, unpublished report, p. 46. Rodda, J.C., 1969, The flood hydrograph, in Chorley, R.J., and Barry, R.G., eds., Water, earth and man: a synthesis of hydrology, geomorphology and socio-economic geography: London, Methuen and Company, p. 405-418.

230

Roering, J.J., Schmidt, K.M., Stock, J.D., Dietrich, W.E., and Montgomery, D.R., 2003, Shallow landsliding, root reinforcement, and the spatial distribution of trees in the Oregon Coast Range: Canadian Geotechnical Journal v. 40, no. 2, p. 237-253. Ruddy, B.C., 2011a, Probability and volume potential postwildfire debris flows in the 2011 Horseshoe II Burn Area, Southeastern Arizona: OFR 2011-1197: Reston, VA, U.S. Geological Survey, p. 10. ---, 2011b, Probability and volume potential postwildfire debris flows in the 2011 Monument Burn Area, Southeastern Arizona: OFR 2011-1181: Reston, VA, U.S. Geological Survey, p. 9. ---, 2011c, Probability and volume potential postwildfire debris flows in the 2011 Wallow Burn Area, Eastern Arizona: OFR 2011-1214: Reston, VA, U.S. Geological Survey, p. 11. 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, v. 96, no. 3-4, p. 310-321. Schaffner, M., and Reed, W.B., 2005, Effects of wildfire in the mountainous terrain of southeast Arizona: post-burn hydrologic response of nine watersheds, NOAA National Weather Service Western Regional Technical Attacment 05-01, Available online: http://www.wrh.noaa.gov/wrh/05TAs/ta0501.pdf. Schilling, S.P., 1998, LAHARZ: GIS Programs for Automated Mapping of Lahar-Inundation Hazard Zones: U.S. Geological Survey, 80 p. Schmidt, K.M., Hanshaw, M.N., Howle, J.F., Kean, J.W., Staley, D.M., Stock, J.D., and Bawdeng, G.W., 2011, Hydrologic conditions and terrestrial laser scanning of post-fire debris flows in the San Gabriel Mountains, CA, USA, in Genevois, R., Hamilton, D.L., and Prestininzi, A., Debris Flow Hazards Mitigation, Mechanics, Prediction, and Assessment: Casa Editrice Univ. La Sapienza, Rome, p. 583-593. Schoennagel, T., Veblen, T.T., and Romme, W.H., 2004, The Interaction of Fire, Fuels, and Climate across Rocky Mountain: Bioscience, v. 54, no. 7, p. 661-676. Shakesby, R.A., 2011, Post-wildfire soil erosion in the Mediterranean: Review and future research directions: Earth-Science Reviews, v. 105, no. 3-4, p. 71-100, doi:10.1016/j.earscirev.2011.01.001. Shakesby, R.A., Doerr, S.H., and Walsh, R.P.D., 2000, The erosional impact of soil hydrophobicity: current problems and future research directions: Journal of Hydrology, v. 231-232, p. 178-191. Sheppard, P.R., Comrie, A.C., Packin, G.D., Angersbach, K., and Hughes, M.K., 2002, The climate of the US Southwest: Climate Research, v. 21, no. 3, p. 219-238. Spencer, J.E., 2000, Possible origin and significance of extension-parallel drainages in Arizona's metamorphic core complexes: Geological Society of America Bulletin, v. 112, no. 5, p. 727-735, doi:10.1130/0016-7606(2000)112<727:poasoe>2.0.co;2. 1 Spencer, J.E., and Pearthree, P.A., 2004, Geologic map of the Oro Valley 7 /2' Quadrangle and the Pusch Peak area, northeastern Pima County, Arizona, scale 1:24000. Spencer, J.E., and Reynolds, S.J., 1989, Middle Tertiary tectonics of Arizona and adjacent areas, in Jenney, J.P., and Reynolds, S.J., eds., Geologic Evolution of Arizona: Tucson, Arizona Geological Society Digest 17, p. 539-574. 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, p. 1-16, doi:10.1007/s10346-012-0341-9. Stein, S.M., Comas, S.J., Menakis, J.P., Carr, M.A., Stewart, S.I., Cleveland, H., Bramwell, L., and Radeloff, V.C., 2013, Wildfire, wildlands, and people: understanding and preparing for wildfire in the wildland-urban interface, USDA Forest Service, Rocky Mountain Research Station, General Technical Report RMRS-GTR-299, p. 40.

231

Stock, J., and Dietrich, W.E., 2003, Valley incision by debris flows: Evidence of a topographic signature: Water Resources Research, v. 39, no. 4, p. 1089-1114. Stock, J.D., and Dietrich, W.E., 2006, Erosion of steepland valleys by debris flows: Geol Soc Am Bull, v. 118, no. 9-10, p. 1125-1148, doi:10.1130/b25902.1. Stone, J.O., 2000, Air pressure and cosmogenic isotope production: J. Geophys. Res., v. 105, p. 23753– 23759. Stoof, C.R., Moore, D., Fernandes, P.M., Stoorvogel, J.J., Fernandes, R.E.S., Ferreira, A.J.D., and Ritsema, C.J., 2013, Hot fire, cool soil: Geophysical Research Letters, v. 40, no. 8, p. 1534-1539, doi:10.1002/grl.50299. Swanson, F.J., 1981, Fire and geomorphic processes, in Hagen, J., Johnston, D., Bonnicksen, T.M., Christensen, N.L., Lotan, J.E., and Reiners, W.A., eds., Fire regimes and ecosystem properties: USDA Forest Service, p. 401-420. Swetnam, T.W., and Anderson, R.S., 2008, Fire Climatology in the western United States: introduction to special issue: International Journal of Wildland Fire, v. 17, no. 1, p. 1-7. Swetnam, T.W., and Baisan, C.H., 1996, Historical fire regime patterns in the Southwestern United States since AD 1700, in Allen, C.D., Fire Effects in Southwestern Forests, Proceedings of the Second La Mesa Fire Symposium: Las Alamos, New Mexico, USDA Forest Service, p. 10-33. Swetnam, T.W., and Betancourt, J.L., 1998, Mesoscale Disturbance and Ecological Response to Decadal Climatic Variability in the American Southwest: Journal of Climate, v. 11, no. 12, p. 3128-3147. Takahashi, T., 1981, Estimation of potential debris flows and their hazardous zones: soft countermeasures for a disaster: Natural disaster science, v. 3, no. 1, p. 57-89. ---, 2007, Debris flow: mechanics, prediction and countermeasures: London, Taylor & Francis. Tarboton, D.G., 2005, Terrain Analysis Using Digital Elevation Models (TauDEM): Logan, Utah State University. Tarolli, P., and Dalla Fontana, G., 2009, Hillslope-to-valley transition morphology: New opportunities from high resolution DTMs: Geomorphology, v. 113, no. 1-2, p. 47-56. Tillery, A.C., Cannon, S.H., and Micheal, J.A., 2011a, Postwildfire preliminary debris flow hazard assessment for the area burned by the 2011 Las Conchas Fire in north-central New Mexico, U.S. Geological Survey Open-File Report 2011-1308, p. 11. Tillery, A.C., Darr, M.J., Cannon, S.H., and Micheal, J.A., 2011b, Postwildfire preliminary debris flow hazard assessment for the area burned by the 2011 Track Fire in Northeastern New Mexico and Southeastern Colorado, U.S. Geological Survey Open-File Report 2011-1257, p. 9. Trapp, R.A., and Reynolds, S.J., 1995, Map showing names and outlines of physiographic areas in Arizona used by the Arizona Geological Survey with comprehensive base map: Arizona Geological Survey. USDA Forest Service, 2010, Schultz Fire Burned Area Emergency Response Report, USDA Forest Service, Coconino National Forest, unpublished report, p. 167. ---, 2011a, Horseshoe 2 Fire Burned-Area Report, FS-2500-8, USDA Forest Service, Coronado National Forest, unpublished report, p. 21. ---, 2011b, Monument Fire Burned-Area Report, FS-2500-8, USDA Forest Service, Coronado National Forest, unpublished report. ---, 2011c, Wallow Wildfire Burned-Area Report, FS-2500-8, USDA Forest Service, Apache-Sitgreaves National Forest, unpublished report, p. 32. ---, 2012, Gladiator Fire Burned-Area Report, FS-2500-8, USDA Forest Service, Prescott National Forest, unpublished report, p. 39.

232

Van Burkalow, A., 1945, Angle of repose and angle of sliding friction: an experimental study: Geological Society of America Bulletin, v. 56, no. 6, p. 669-707, doi:10.1130/0016- 7606(1945)56[669:aoraao]2.0.co;2. Van Devender, T.R., 1990, Late Quaternary vegetation and climate of the Sonoran Desert, United States and Mexico, in Betancourt, J.L., Van Devender, T.R., and Martin, P.S., eds., Packrat middens; the last 40,000 years of biotic change: Tucson, AZ, Univ. Ariz. Press, p. 134-165. Van Devender, T.R., Toolin, L.J., and Burgess, T.L., 1990, The ecology and paleoecology of grasses in selected Sonoran Desert plan communities, in Betancourt, J.L., Van Devender, T.R., and Martin, P.S., eds., Packrat middens; the last 40,000 years of biotic change: Tucson, AZ, Univiversity of Arizona Press, p. 326-349. Wagenbrenner, J.W., and Robichaud, P.R., In Review, Post-fire sediment delivery across spatial scales in the interior western US: Earth Surface Processes and Landforms. Wagner, J.D.M., 2006, Speleothem record of southern Arizona paleoclimate, 54 to 3.5 ka: Tucson, University of Arizona, Unpublished dissertation, 129 p. Wagner, J.D.M., Cole, J.E., Beck, J.W., Patchett, P.J., Henderson, G.M., and Barnett, H.R., 2010, Moisture variability in the southwestern United States linked to abrupt glacial climate change: Nature Geosci, v. 3, no. 2, p. 110-113, doi:http://www.nature.com/ngeo/journal/v3/n2/suppinfo/ngeo707_S1.html. Waters, M.R., 1989, Late Quaternary lacustrine history and paleoclimatic significance of pluvial Lake Cochise, southeastern Arizona: Quaternary Research, v. 32, no. 1, p. 1-11. Webb, R., Griffiths, P.G., and Melis, T.S., 2005, Initiation of debris flows in bedrock canyons in the Colorado River drainage, USA: Abstracts with Programs - Geological Society of America, Geological Society of America v. 37, p. 36. Webb, R.H., and Betancourt, J.L., 1992, Climatic variability and flood frequency of the Santa Cruz River, Pima County, Arizona, U. S. Geological Survey Water-Supply Paper 2379: Reston, VA, United States, U. S. Geological Survey, p. 40. Webb, R.H., Griffiths, P.G., and Rudd, L.P., 2008a, Holocene debris flows on the Colorado Plateau: The influence of clay mineralogy and chemistry: Geological Society of America Bulletin, v. 120, no. 7- 8, p. 1010-1020, doi:10.1130/b26055.1. Webb, R.H., Magirl, C.S., Griffiths, P.G., and Boyer, D.E., 2008b, Debris flows and floods in southeastern Arizona from extreme precipitation in late July 2006: Magnitude, frequency, and sediment delivery, U.S. Geological Survey Open-File Report 2008-1274: Tucson, Arizona, United States Geological Survey, p. 95. Webb, R.H., Melis, T.S., Griffiths, P.G., Elliott, J.G., Cerling, T.E., Poreda, R.J., Wise, T.W., and Pizzuto, J.E., 1999, Lava Falls Rapid in Grand Canyon: effect of late Holocene debris flows in the Colorado River, U. S. Geological Survey Professional Paper 1591: Reston, VA, United States, U. S. Geological Survey, p. 90. Webb, R.H., Pringle, P.T., Reneau, S.L., and Rink, G.R., 1988, The 1984 Monument Creek debris flow: Implications for the formation of rapids on the Colorado River in Grand Canyon National Park: Geology, v. 16, p. 50-54. Webb, R.H., Pringle, P.T., and Rink, G.R., 1989, Debris flows from tributaries of the Colorado River, Grand Canyon National Park, Arizona, U. S. Geological Survey Professional Paper 1492: Reston, VA, United States, U. S. Geological Survey, p. 39. Wells, S.G., and Harvey, A.M., 1987, Sedimentologic and geomorphic variations in storm-generated alluvial fans, Howgill Fells, Northwest England Geological Society of America Bulletin, v. 98, no. 2, p. 182-198.

233

Weng, C., and Jackson, S.T., 1999, Late Glacial and Holocene vegetation history and paleoclimate of the Kaibab Plateau, Arizona: Palaeogeography, Palaeoclimatology, Palaeoecology, v. 153, p. 179- 201. Westerling, A.L., Gershunov, A., Brown, T.J., Cayan, D.R., and Dettinger, M.D., 2003, Climate and wildfire in the western United States: Bulletin of the American Meteorological Society, v. 84, no. 5, p. 595-604. Westerling, A.L., Hidalgo, H.G., Cayan, D.R., and Swetnam, T.W., 2006, Warming and Earlier Spring Increase Western U.S. Forest Wildfire Activity: Science, v. 313, no. 5789, p. 940-943, doi:10.1126/science.1128834. Wieczorek, G.F., and Glade, T., 2005, Climatic factors influencing occurrence of debris flows, in Jakob, M., and Hungr, O., eds., Debris-flow hazards and related phenomena: Berlin, Federal Republic of Germany, Springer, p. 325-362. Wilcock, P.R., and Iverson, R.M., 2003, Prediction in geomorphology, in Wilcock Peter, R., and Iverson Richard, M., eds., Prediction in geomorphology.: Washington, DC, American Geophysical Union, p. 3-11. Willgoose, G., Bras, R.L., and Rodriguez-Iturbe, I., 1991, A physical explanation of an observed link area- slope relationship: Water Resources Research, v. 27, no. 7, p. 1697-1702, doi:10.1029/91wr00937. Williams, A.P., Allen, C.D., Macalady, A.K., Griffin, D., Woodhouse, C.A., Meko, D.M., Swetnam, T.W., Rauscher, S.A., Seager, R., Grissino-Mayer, H.D., Dean, J.S., Cook, E.R., Gangodagamage, C., Cai, M., and McDowell, N.G., 2013, Temperature as a potent driver of regional forest drought stress and tree mortality: Nature Clim. Change, v. advance online publication, doi:http://www.nature.com/nclimate/journal/vaop/ncurrent/abs/nclimate1693.html#suppleme ntary-information. Williams, A.P., Allen, C.D., Millar, C.I., Swetnam, T.W., Michaelsen, J., Still, C.J., and Leavitt, S.W., 2010, Forest responses to increasing aridity and warmth in the southwestern United States: Proceedings of the National Academy of Sciences, v. 107, no. 50, p. 21289-21294, doi:10.1073/pnas.0914211107. Wohl, E., 2013, Migration of channel heads following wildfire in the Colorado Front Range, USA: Earth Surface Processes and Landforms, v. 38, no. 9, p. 1049-1053, doi:10.1002/esp.3429. Wohl, E.E., and Pearthree, P.A., 1991, Debris flows as geomorphic agents in the Huachuca Mountains of southeastern Arizona: Geomorphology, v. 4, no. 3-4, p. 273-292. Wolkowinsky, A.J., and Granger, D.E., 2004, Early Pleistocene incision of the San Juan River, Utah, dated with 26Al and 10Be: Geology, v. 32, no. 9, p. 749-752, doi:10.1130/g20541.1. Youberg, A., 2008, Delineating post-wildfire debris-flow hazards for pre-fire mitigation, Pine and Strawberry, Arizona: A FEMA 5% Initiative Study, Open-File Report 08-08: Tucson, Arizona Geological Survey, p. 20. Youberg, A., 2011, Horseshoe II Fire, Burned Area Emergency Response (BAER) Assessment, Geologic Hazard Assessment: Specialist Report: Tucson, Coronado National Forest, p. 15. Youberg, A., Cline, M.L., Cook, J.P., Pearthree, P.A., and Webb, R.H., 2008, Geologic mapping of debris- flow deposits in the Santa Catalina Mountains, Pima County, Arizona: Arizona Geological Survey, 41 p., 11 map sheets on CD, scale 41:6000., scale 1:6,000. Youberg, A., Koestner, K.A., Neary, D.G., and Koestner, P.E., 2011, Rainfall and Geomorphic Aspects of Post-Fire Soil Erosion – Schultz Fire 2010, Arizona Hydrological Society Annual Symposium: Flagstaff, AZ, Arizona Hydrological Society, p. 5.

234

Youberg, A., and Pearthree, P.A., 2011, Post-Monument Fire Floods and Debris Flows in the Huachuca Mountains, southern Arizona, Arizona Geology Tucson, Arizona Geological Survey, v. 41, no. 2. Youberg, A., and Pearthree, P.A., 2012, Geologic Map of Parts of the Lukeville, Diaz Peak, South of Lukeville and Blankenship Wells 7 1/2' Quadrangles, Organ Pipe Cactus National Monument, Pima County, Arizona: Arizona Geological Survey DGM-74, one map sheet at 1:24,000 map scale. Youberg, A., Pearthree, P.A., and Baker, V.R., 2006, Comparison of Debris Flows Generated in Adjacent Unburned and Recently-Burned Areas, Coronado National Memorial, Arizona, Eos Transactions AGU: San Francisco, v. 87(52), no. Fall Meeting Supplement, p. Abstact 12661. Young, J.J., Cook, J.P., and Pearthree, P.A., 2012, Surficial Geologic Map of the Western Piedmont of the Maricopa Mountains and the Southern Piedmont of the Buckeye Hills, Maricopa County, Arizona: Arizona Geological Survey DGM-75, v 1.0, 1 plate, map scale 1:24,000

235