Extreme events as drivers of landscape evolution in active mountain belts

PDF

Download

Open document

Flip pages

Summarize

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

EXTREME EVENTS AS DRIVERS OF LANDSCAPE EVOLUTION IN ACTIVE
MOUNTAIN BELTS

by

Maxwell Philip Boulet Dahlquist

A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
GEOLOGICAL SCIENCES

May 2019

i

Acknowledgments:
I wouldn’t have made it to the end of this road without an advisor as supportive and encouraging
as Josh West. He nourished my good ideas while steering me away from my bad ones and gave
me both the space and support to forge my own identity as a researcher.

Over the last few years, I have been helped and uplifted by faculty, staff, and students at USC
and beyond more times than I could count. Among those who have directly supported my
development in one way or another are Doug Hammond, John Platt, David Okaya, Frank
Corsetti, Meghan Miller, Thorsten Becker, John Yu, Cindy Waite, Karen Young, Barbara Grubb,
Vardui Ter-Simonian, John McRaney, Deborah Gormley, the West and Hammond lab groups,
Marin Clark, Dimitrios Zekkos, the Clark group at UM, Kristen Cook, Nova Roosmawati,
Cooper Harris, Paul Quackenbush, Gen Li, Abra Atwood, Julie Martinez, the USC Young
Researchers Program, and many, many others. I particularly thank James Dolan and Steve Nutt
for serving on my thesis committee.

I also thank the faculty, staff and students and the University of Southern Indiana, who helped
me establish my foundation as a scientist.

The community of USC Earth Sciences students has been essential to maintaining happiness and
mental health through tough times. I owe my sanity to my friends and colleagues with whom
I’ve shared laughter, commiseration, study sessions, beers, gripes, and field trips.

Above all, I thank my wife, Meryn, and son, Rookh, whose love is all the motivation I need, my
parents, Nan and Tom, without whose influence I would be less in every way, and my siblings,
Annika, Ben, Lillia, Lucia, Jude, and Greta, who helped me to learn the lessons I needed to grow
into a decent adult.

Financial support in my efforts was provided by NSF awards EAR-1546630 and EAR-1250214
and a USC Provost’s Fellowship.

ii

Dedicated to Meryn, who keeps me going.
AMDG

iii

Table of Contents

CHAPTER 1: Introduction – Extreme events and landscapes.........................................................1
1.1 J. Harlen Bretz and the Channeled Scablands…………..…………………………..….1
1.2 Tectonic geomorphology and the basis for this work……………………………..……3
1.3 Chapter summaries…………………………………………………………………….4
1.3.1 Chapter 2 Abstract: Initiation and runout of post-seismic debris flows…......7
1.3.2 Chapter 3 Abstract: Glacial lake outburst floods set the pace ……………...8
of erosion in the Himalaya
1.3.2 Chapter 4 Abstract: Landslide-driven drainage divide migration………….10
1.4 References……………………………………………………………………………11
CHAPTER 2: Initiation and runout of post-seismic debris flows ………………………….…....15
2.1 Introduction…………………………………………………….………………...…..15
2.1.2 The Gorkha Earthquake and associated landslides………………………...17
2.1.3 Multiple mechanisms for post-seismic debris flows: Type-1 vs. Type-2….17
2.2 Results and discussion…………………………………………………………….….19
2.2.1 Debris flows generated by the Gorkha Earthquake…………………….…..19
2.2.2 Channel geometry and runout distances……………………………….…...23
2.2.3 Hazard analysis: A framework for predicting the locations of Type-1.……27
debris flows following large earthquakes
2.2.4 Comparison of DEM resolutions…………………………………………..32
2.2.5 Confluence angles and runout…………………………………….………..34
2.3 Conclusions…………………………………………………………………………..37
2.4 Methods………………………………………………………………………………38
2.4.1 Debris flow mapping……………………………………………………….38
2.4.2 Topographic analysis…………………………………………………….…39
2.5 References……………………………………………………………………………46

CHAPTER 3: Glacial lake outburst floods set the pace of erosion in the Himalaya…………….52
3.1 Introduction……...…………………………………………………………………...52
3.1.1 Channel Network Response to Tectonic and Climatic Forcing……………52
3.1.2 Physical Relationships in Fluvial and Debris Flow Channel Networks…...59
3.1.2.1 The stream power law and normalized channel steepness……….59
3.1.2.2 a1 and debris flow dominated channels……………………….…..61
3.1.2.3 The integral method of channel profile analysis…………….……64
3.1.2.4 Channel width and normalized channel wideness…………..…….64
3.1.3 Toward a Threshold Channel Model…………………………………….…64
3.1.4 Regional Geology and Important Geomorphic Features in the Nepal…..…..68
Himalaya
3.2 Results and discussion…………………………………………………………….….71
3.2.1 Debris Flows and the Gorkha Earthquake………………………………….71
3.2.2 Evidence from a Debris Flow Inventory…………………………………...73
3.2.3 Glacial Lake Outburst Floods Set Base Level for Debris …...………….…76
Flow-Dominated Basins

iv

Table of Contents (cont.)

3.2.4 Debris Flows and Outburst Floods Cooperate to Produce High…………...84
Topography
3.3 Conclusions…………………………………………………………………………..89
3.4 Methods…………..…………………………………………………………………..89
3.5 References………..…………………………………………………………………..91

CHAPTER 4: Landslide-driven drainage divide migration………….…………………………100
4.1 Introduction…………………………………………………………………………100
4.1.1 Drainage divides and the threshold hillslope model……………………...100
4.1.2 Study areas………………………………………………………………………..102
4.2 Methods…….……………………………………………………………………….103
4.2.1 Landslide mapping………………………………………………………..103
4.2.2 Verifying geolocation of ridges……………………………………….…..106
4.2.3 Error introduced by DEM resolution……………………………….……..107
4.2.4 Topographic analysis………………………………………………….…..108
4.3 Results………………………………………………………………………………110
4.3.1 Area exchanged in each event……………………………………….…….110
4.3.2 Comparison of divide stability metrics………………………….………...112
4.4 Discussion…………………………………………………………………………..112
4.4.1 Landslides drive divides toward steady state…………………….………..112
4.4.2 A metric for divide migration…………………………………….……….115
4.5 Conclusions…………………………………………………………………………117
4.6 References…………………………………………………………………………..127

CHAPTER 5: Conclusions……………………………………………………………………..130
5.1 Extreme events drive landscape evolution from channel to ridge………….……….130

1

CHAPTER 1: Introduction – Extreme events and landscapes 1
1.1 J. Harlen Bretz and the Channeled Scablands 2
In 1909, high school biology teacher Harley Bretz found himself fascinated and baffled 3
by the landforms of the Channeled Scablands in Eastern Washington. In particular, he noticed a 4
precipice that appeared to be a pour point from Quincy Basin into the Columbia River, implying 5
the existence of a waterfall larger than any currently existing, despite the absence of a 6
watercourse feeding it. When experts at the University of Washington were unable to give him a 7
satisfying explanation for the formation of this feature, Bretz changed his name to the more 8
professionally respectable “J. Harlen” and headed east to the University of Chicago to earn a 9
Ph.D. (Hodges, 2017). He spent several decades studying the geomorphology of the Scablands, 10
and developed an explanation for the labyrinthine dry channel network and the stripped-bare 11
land surface: A Pleistocene megaflood emanating from somewhere near the margin of the 12
Cordilleran Ice Sheet inundated the scablands with discharges more than 30 times greater than 13
any occurring in outburst floods in recorded history (O’Connor and Costa, 2004), scouring the 14
land surface and carrying away dozens of cubic kilometers of rock. 15
Bretz’s reinterpretation of the Scablands (Bretz, 1923; 1925; 1928, many others) was met 16
with a generally icy reception by the early 20
th
century geological intelligentsia. The proposed 17
occurrence of a megaflood as a major force of landscape evolution in the recent geologic past 18
proved distasteful in the 1920s, when resurgent religious fundamentalism favored catastrophism 19
of a more supernatural variety. Orthodox scientific thinking was dominated by a reading of 20
Hutton and Lyell’s Uniformitarianism that itself was somewhat fundamentalist (Hutton, 1795; 21
Lyell, 1833). According to the standard interpretation, the anastomosing patterns of canyons, 22
massive gravel bars, and scoured earth, must have been the result of millions of years of fluvial 23

2

erosion, aggradation, and avulsion. This was held as a fact that was practically self-evident: 24
Large scale erosional features could only be formed by long-lasting, gradual processes. To 25
suggest otherwise, even with plenty of supporting data, was unscientific. There were some early 26
adopters of Bretz’s ideas, particularly among researchers who had seen the features in person 27
(e.g. Jenkins, 1925), but the reaction of the establishment as a whole was at best, skeptical. The 28
U.S. Geological Survey, at the time, was a very conservative organization and dominated the 29
discourse of American geologists. In 1927, the Geological Society of Washington D.C. invited 30
him to present his findings at the Cosmos Club, with the unstated goal of beating his ideas about 31
the “Spokane Flood” into the ground. After Bretz’s presentation, his opponents lit into him with 32
invective, with one participant calling him “incompetent” and the flood hypothesis 33
“preposterous”. Having accepted the invitation with the goal of changing minds, Bretz departed 34
Washington in a depression (Soennichsen, 2008). Fortunately for the progress of 35
geomorphology, he was a determined and self-confident man, and returned to his research with a 36
vengeance. He sent a series of letters to his detractors from the Cosmos Club, aiming to address 37
their criticisms and alternative hypotheses individually, and finally began to gain traction. He 38
was aided in this by Joseph Pardee’s work describing Glacial Lake Missoula, as catastrophic 39
draining of an ice-dammed lake was a more plausible explanation than Bretz’s initial proposed 40
mechanisms of sub-glacial volcanism or rapid climate fluctuations (Pardee, 1942). However, 41
despite the 1952 discovery of a smoking gun in the form of giant current ripples along the 42
outflow path, some gradualist partisans kept the fight alive until around 1970 or so (Bretz, 1956; 43
1969). 44
Controversy was renewed by Richard Waitt in 1980, when he proposed that the scablands 45
were actually carved by at least 40 separate megafloods, by repeated filling and draining of Lake 46

3

Missoula (Waitt, 1980; 1985). This built upon Bretz’s final work on the topic in 1969, where he 47
raised the possibility of multiple floods based on interpretation of the sedimentary record in 48
backwater basins (Bretz, 1969). Ironically, this explanation of a series of floods rather than a 49
single one might have been more palatable in the atmosphere of the community of the 1920s’ 50
distrust of singular events. However, the multiple flood/single flood debate carried on for a 51
further twenty years (e.g., Shaw et al., 1999; Benito and O’Connor, 2003) before the multiple 52
flood camp finally won out. 53
54
1.2 Tectonic geomorphology and the basis for this work 55
J. Harlen Bretz, his flood hypothesis, and the debate surrounding it, alongside the 56
simultaneous debate on continental drift and plate tectonics, led to a sea change in the 57
interpretation of geomorphic features. The discovery of a mechanism for long-lived growth and 58
evolution of orogens, coupled with the acceptance that rare, extreme events can exert enormous 59
influence on the land surface ushered in the new discipline of tectonic geomorphology, 60
expanding upon the geographical cycle of William Morris Davis and recent advances in 61
understanding of the process of fluvial erosion (e.g. Davis; 1899; Hack, 1957; Strahler, 1952; 62
Flint, 1974; Whipple and Tucker, 1999, alongside many others). Over the past few decades, the 63
role of outburst floods (Dunning et al., 2013; Cook et al., 2018), wildfires (Gabet and Bookter, 64
2008; DeLong et al., 2018), large earthquakes (Korup et al., 2013; Li et al., 2014), and extreme 65
rainfall (Caine et al., 1980; Lin et al., 2011) in the evolution of landscapes has begun to be 66
explored and quantified, particularly in actively uplifting regions where erosion is rapid. While 67
landscapes carved by outburst floods at the margins of continental ice sheets represent a near end 68
member for rarity and magnitude, many familiar landscapes may be generally controlled by 69

4

extreme events of varying frequency. For example, landslides and debris flows following 70
wildfires are an important control on erosion in mountain landscapes in the southwestern United 71
States (e.g. DeLong et al., 2018). Jökulhlaups (glacial lake outburst floods, in the strict sense 72
those generated by geothermal melting or subglacial eruptions) dominate incision in rivers 73
draining glaciated terrain in Iceland (e.g. Dunning et al., 2018). In the Himalaya, and other 74
mountain ranges bounding the India-Eurasia plate boundary, major earthquakes simultaneously 75
drive uplift along thrust faults and erosion by triggering tens of thousands of landslides which 76
may mobilize many cubic kilometers of debris (e.g. Li et al., 2014). This mobilized landslide 77
debris is the source for debris flows and fluvially transported sediment, which can drive river 78
incision downstream. 79
In this work, the term “extreme event” comes up repeatedly. There is no commonly 80
agreed upon definition for what sort of event qualifies as extreme, although 99
th
percentile 81
magnitudes or 100-year recurrence intervals appear commonly. We will not set any specific 82
definition for “extreme”, but note that the earthquakes and storms whose effects will be 83
examined in greater detail have recurrence intervals in the range of 100s to 1000s of years, and 84
the 2016 glacial lake outburst floods discussed in chapter 3 had a peak discharge an order of 85
magnitude greater than a 75-year monsoon flood 20 kilometers from its source (Cook et al., 86
2018; McPhillips et al., 2018). Therefore, by any commonly used definition of “extreme”, these 87
events should qualify. 88
89
1.3 Chapter Summaries 90
In recent years, the availability of high-quality digital topography and satellite imagery 91
have made it possible to examine the impact of rare, extreme events far more extensively than 92

5

ever before. The purpose of this study is to use these resources to collect novel datasets 93
illustrative of landscape evolution and apply topographic analysis techniques to explore 94
relationships between mass movements generated by earthquakes and storms, and the present 95
geometry of the landscape, in the context of regional tectonics. We mapped debris flows 96
associated with the 2015 Gorkha Earthquake in Nepal and examined their characteristics and 97
distribution in the context of the geometry of the Nepal Himalaya. Using this dataset, we 98
examined the necessary conditions to generate debris flows: available debris in a channel, 99
sufficient rainfall, and sufficient channel slope, and attempted to contextualize our data on 100
extreme event-triggered debris flows for better understanding of the cascading hazard chain that 101
follows large earthquakes. 102
We also examined our debris flow dataset in the context of glacial lake outburst floods 103
(GLOFs), which we propose are an important control on the pace of incision in the region. The 104
erosive potential of these floods has been recognized in recent work, but while they are common 105
from a geologic standpoint, the lack of long-term records and rates of their occurrence over large 106
areas makes interpreting their role in shaping steep mountain belts difficult. We attempted to use 107
the drainage area above a given altitude that a river captures as a proxy for the frequency of 108
GLOFs in that valley. Beginning with that assumption, we examined the relationships between 109
the geometry of channels dominated by GLOF incision and their tributaries whose base level is 110
set by the pace of incision in the main channels, to arrive at better understanding of how these 111
extreme floods might shape the geometry of the Nepal Himalaya specifically, and steep, 112
glaciated mountain belts in general 113
Finally, we collect a dataset of landslides triggered by two large earthquakes and a 114
tropical cyclone which have caused drainage divides to change their position, and examine them 115

6

in the context of channel morphology metrics thought to predict divide stability (Willett et al., 116
2014; Whipple et al., 2017; Forte and Whipple, 2018). The exchange of drainage area and, by 117
proxy, discharge between basins is an important means to redistribute erosive power toward 118
landscape mass balance in the presence of tectonic, climatic, or lithologic gradients. Divide 119
migration is thus an important factor in the long-term evolution of mountain belts. Using our 120
dataset of extreme event-triggered landslides that caused redistribution of drainage area, we 121
attempted to quantify the impact of these events toward topographic steady state, as well as 122
determine how predictable divide migration driven by extreme events can be. 123
A common thread running through these analyses is the threshold hillslope model, which 124
predicts that river incision will drive hillslope steepening up to a certain threshold slope angle 125
determined by material strength, above which additional steepening will be accommodated by 126
increased landslide rate rather than increased steepening. This model links river incision with 127
hillslope evolution in steep landscapes, providing an important framework for understanding the 128
way fluvial incision, controlled by climate, tectonics, and bedrock erodibility, is coupled to 129
hillslopes (Burbank et al., 1996; Montgomery, 2001; Larsen and Montgomery; 2012). 130
Although the point where hillslope processes give way to channel processes is fairly 131
obvious in landscapes, there are other process transitions that are critical in controlling the 132
geometry of mountain ranges. It is less apparent, for example, where the reaches of rivers 133
dominated by GLOFs or debris flows give way to those dominated by runoff-driven monsoon 134
flooding. We aim to extend the logic of the threshold hillslope model, in the sense of the simple 135
idea that a process occurring downstream will set the pace of the processes occurring above it, to 136
major channel networks in the Nepal Himalaya. To this end, we must consider that erosion in the 137
Himalaya and other mountain belts have components that are both bottom-up, in the case of 138

7

base-level lowering by tectonic uplift, and top-down, in the sense of climate-driven changes in 139
glacial processes and sediment supply to steep headwater channels from earthquake and storm- 140
triggered landslides. Additionally, we want to determine whether the signal of river incision- 141
conditioned landsliding that the threshold hillslope model predicts is identifiable even at the 142
ridges, through the signal of drainage divide migration. The separate parts of this thesis work as 143
pieces in enormous puzzle of tectonics, erosion, climate, and process regime transitions. 144
145
1.3.1 Chapter 2 Abstract: Initiation and runout of post-seismic debris flows 146
Post-seismic debris flows are one of the pernicious hazards following large earthquakes, 147
propagating destruction and loss of life many kilometers downstream from the steep hillslopes 148
where co-seismic landslides occur, and extending damage for months to years after shaking has 149
stopped. Debris flow hazard assessment is time-sensitive, as landslide material in steep, colluvial 150
channels may be readily mobilized during significant rainfall events. Improving understanding of 151
when and where post-seismic debris flows are likely to occur, and developing a rapidly 152
applicable model for hazard analysis, could help to protect vulnerable lives and optimally 153
allocate disaster response resources following major earthquakes. Large datasets of post-seismic 154
debris flows are necessary to predict their initiation and runout characteristics, yet such datasets 155
have remained scarce. In this study, we used satellite imagery supplemented by field 156
observations to compile an inventory of >1000 debris flows associated with the 2015 Gorkha 157
earthquake in the Nepal Himalaya. We find that post-seismic debris flows following this event 158
initiated in channels with a similar threshold slope to that observed for debris flow initiation in 159
flume experiments. Furthermore, we can identify distinct initiation and runout characteristics of 160
debris flows sourced from co-seismic landslide debris, versus those sourced from post-seismic, 161
8
monsoon-triggered landslides. Our results suggest that identifying channel segments which 162
exceed a threshold slope for debris flow formation and are within a certain distance to a co- 163
seismic landslide could offer a swiftly applicable model for assessing hazards posed by post- 164
seismic debris flows. 165
This work was supported by National Science Foundation grants EAR-1546630 and 166
EAR-1250214. Dahlquist received support from a University of Southern California Provost’s 167
Fellowship. We thank the USC Young Researchers Program and Julie Martinez for contributions 168
to debris flow mapping, Abra Atwood for constructing the SETSM DEM, and Marin Clark, 169
Dimitrios Zekkos, and Julien Emile-Geay for discussion that helped to focus the manuscript. I 170
thank Paul Morin from the PGC (Polar Geospatial Center) for providing imagery access and 171
support for acquiring Digital Globe satellite data through a NGA (National Geospatial- 172
Intelligence Agency) cooperative agreement with NSF (NextView License), which is supported 173
by NSF grants OPP-1043681, OPP-1559691 and OPP-1542736. Geospatial support for this work 174
was provided by the Polar Geospatial Center under NSF awards EAR-1719496 and EAR- 175
1719524. All images are the copyright of Digital Globe, 2017. This chapter is in revision for 176
publication under the title: “Initiation and runout of post-seismic debris flows: Insights from the 177
2015 Gorkha earthquake”. 178
179
1.3.2 Chapter 3 Abstract: Glacial lake outburst floods set the pace of erosion 180
in the Himalaya 181
Debris flows are an important process facilitating erosion in steep, tectonically active 182
landscapes, mobilizing material from steep channels, driving incision, and delivering coarse 183
sediment downstream. Similarly, glacial lake outburst floods (GLOFs) are powerful erosional 184
9
agents, though they operate on different spatial and temporal scales than debris flows. While 185
both processes have been the subject of individual study, regional-scale analysis of their roles in 186
landscape evolution has been limited. We use a new inventory of debris flows associated with 187
the 2015 Gorkha Earthquake to identify signatures in the morphology of basins dominated by 188
debris flow activity. We consider these basins in the context of wider topographic analysis of the 189
Nepal Himalaya, with the aim of better understanding how relationships between different 190
process domains (debris flow vs. GLOF vs. fluvial incision) contribute to shaping the landscape. 191
Patterns of normalized steepness index (ksn) and a similar metric more tailored to debris flow 192
channels (a1) reflect transitions between the dominant geomorphic process regimes. We note a 193
distinct difference in the geometry of basins which capture glaciated valleys (where GLOFs can 194
originate) versus those that do not, as well as increasing steepness in the tributaries of rivers 195
seeing more GLOFs. We suggest that local base level of debris flow-dominated basins is set by 196
incision of main trunk channels during GLOFs. Debris flows, specifically those associated with 197
the earthquake cycle, drive incision in tributary basins to keep pace. We also observe that the 198
distribution of debris flows following the Gorkha Earthquake reflects the main physiographic 199
transition between the Lesser and High Himalayas, interpreted as resulting from a gradient in 200
uplift and exhumation controlled by a mid-crustal ramp. We argue that, in addition to the 201
exhumation gradient, the distinct physiographic transition represents a transition between 202
dominant process regimes. Above the transition, debris flows (facilitated by sediment production 203
during co-seismic landsliding) keep pace with rapidly incising GLOF-dominated rivers. Below 204
the transition, more gently-sloping channels experience relatively slower incision from storm- 205
triggered debris flows not associated with seismicity. In this lower-relief region, monsoon-flood 206
processes predominate in large rivers. 207
10
This work was supported by National Science Foundation grants EAR-1546630 and 208
EAR-1250214. Dahlquist received support from a University of Southern California Provost’s 209
Fellowship. We thank the USC Young Researchers Program and Julie Martinez for contributions 210
to debris flow mapping, and Marin Clark, Dimitrios Zekkos, and Kristen Cook for discussion 211
that helped to focus the manuscript. This manuscript is in preparation for publication under the 212
title “Glacial lake outburst floods set the pace of erosion in the Himalaya”. 213
214
1.3.4 Chapter 4 Abstract: Landslide-driven drainage divide migration 215
Drainage divide migration reorganizes river basins, redistributing erosive energy and 216
contributing to feedbacks between tectonics, erosion, and climate. However, the conditions 217
governing divide migration and the timescales on which it occurs are poorly understood. By 218
connecting channels to hillslopes in steep landscapes, landslides are expected to play a central 219
role in divide migration and landscape evolution. In this study, we examine landslides triggered 220
by three events (two earthquakes and a tropical cyclone), seeking insight into controls on divide 221
migration. Of the ~100,000 landslides triggered, we mapped 365 which caused a divide to 222
migrate, resulting in a total exchange of about 2 km
2
between basins from ~82,000 km
2
affected 223
by landsliding. By applying several proposed metrics for divide stability based on river channel 224
morphology, we use our database of divide migrations to test for the role of landslides in 225
coupling between channels and divides. We find that, at the timescale of a single landslide- 226
generating event, patterns of area gain and loss between basins are consistent with landscapes 227
progressing toward steady state as inferred from channel metrics. We also propose a metric to 228
quantify divide migration, area exchange, and the contribution of an event toward topographic 229
steady state. Restricting our analysis to the main drainage divide and using estimates of 230
11
recurrence interval and the rate of topographic evolution in Taiwan, we calculate that landslides 231
triggered by large typhoons account for a minimum of 12-15% of southern Taiwan’s progress 232
toward steady state. 233
This work was supported by National Science Foundation grants EAR-1546630 and 234
EAR-1250214. Dahlquist received support from a University of Southern California Provost’s 235
Fellowship. We thank three reviewers for constructive comments that considerably 236
improved the manuscript, and Mark Quigley for editorial handling during the preparation of this 237
manuscript for publication. The manuscript was published in Geology under the title “Landslide- 238
driven drainage divide migration” 239
240
1.4 References 241
Benito, G. and J. E. O’Connor, 2003, Number and size of last-glacial Missoula floods in the 242
Columbia River valley between the Pasco Basin, Washington, and Portland, Oregon: 243
Geological Society of America Bulletin, v. 115, no. 5, p. 624-638. 244
245
Bretz, J. H., 1923, The Channeled Scablands of the Columbia Plateau: The Journal of Geology, 246
v. 31, no. 8, p. 617–649, doi:10.1086/623053. 247
248
Bretz, J. H., 1925, The Spokane Flood beyond the Channeled Scablands: The Journal of 249
Geology, v. 33, no. 2, p. 97–115, doi:10.1086/623179. 250
251
Bretz, J. H., 1928, The Channeled Scabland of Eastern Washington: Geographical Review, v. 18, 252
no. 3, p. 446-477. 253
254
Bretz, J. H., 1956, Channeled Scabland of Washington: New data and interpretations: Geological 255
Society of America Bulletin, v. 18, no. 3, p. 446-477. 256
257
Bretz, J. H., 1969, The Lake Missoula floods and the Channeled Scablands: The Journal of 258
Geology, v. 18, no. 3, p. 505-543. 259
260
Burbank, D. W., J. Leland, E. Fielding, R. S. Anderson, N. Brozovic, M. R. Reid, and C. 261
Duncan, 1996, Bedrock incision, rock uplift and threshold hillslopes in the northwestern 262
Himalayas: Nature, v. 379, no. 6565, p. 505–510, doi:10.1038/379505a0. 263
264
12
Caine N., 1980, The rainfall intensity-duration control of shallow landslides and debris flows: 265
Geografiska Annaler v. 62, no. 1, p. 23-27. 266
267
Cook, K. L., C. Andermann, F. Gimbert, B. R. Adhikari, and N. Hovius, 2018, Glacial lake 268
outburst floods as drivers of fluvial erosion in the Himalaya: Science, v. 362, no. 6410, p. 269
53-57. doi:10.1126/science.aat4981. 270
271
Davis, W.M., 1899, The geographical cycle: The Geographical Journal, v. 14, 272
p. 481–504, doi:10.2307/1774538 273
274
DeLong, S. B., A. M. Youberg, W. M. DeLong, and B. P. Murphy, 2018, Post-wildfire 275
landscape change and erosional processes from repeat terrestrial lidar in a steep 276
headwater catchment, Chiricahua Mountains, Arizona, USA: Geomorphology, v. 300, p. 277
13–30, doi:10.1016/j.geomorph.2017.09.028. 278
Dunning, S. A., A. R. G. Large, A. J. Russell, M. J. Roberts, R. Duller, J. Woodward, A.-S. 279
Meriaux, F. S. Tweed, and M. Lim, 2013, The role of multiple glacier outburst floods in 280
proglacial landscape evolution: The 2010 Eyjafjallajokull eruption, Iceland: Geology, v. 281
41, no. 10, p. 1123–1126, doi:10.1130/G34665. 282
Flint, J. J., 1974, Stream gradient as a function of order, magnitude, and discharge: Water 283
Resources Research, v. 10, no. 5, p. 969–973, doi:10.1029/WR010i005p00969. 284
285
Forte, A. M., and K. X. Whipple, 2018, Criteria and tools for determining drainage divide 286
stability: Earth and Planetary Science Letters, v. 493, p. 102–117, 287
doi:10.1016/j.epsl.2018.04.026. 288
289
Gabet, E. J., and A. Bookter, 2008, A morphometric analysis of gullies scoured by post-fire 290
progressively bulked debris flows in southwest Montana, USA: Geomorphology, v. 96, 291
no. 3–4, p. 298–309, doi:10.1016/j.geomorph.2007.03.016. 292
293
Hack, J. T., 1957, Studies of longitudinal stream profiles in Virginia and Maryland: US 294
Geological Survey professional paper, v. 294-B, p. 45-97 295
296
Hodges, G., 2017, Formed by megafloods, this place fooled scientists for decades: National 297
Geographic, accessed February 19, 2019, 298
https://news.nationalgeographic.com/2017/03/channeled-scablands/ 299
300
Hutton, J., 1795, Theory of the Earth, with Proofs and Illustrations, Volume II: Edinburgh, UK, 301
Geological Society, 248 p. 302
303
Jenkins, O. P., 1925, The Spokane Flood: A Discussion: The Journal of Geology, v. 33, no. 7 p. 304
747-748. 305
306
Korup, O., A. L. Densmore, and F. Schlunegger, 2010, The role of landslides in mountain range 307
evolution: Geomorphology, v. 120, no. 1–2, p. 77–90, 308
doi:10.1016/j.geomorph.2009.09.017. 309
13
Larsen, I. J., and D. R. Montgomery, 2012, Landslide erosion coupled to tectonics and river 310
incision: Nature Geoscience, v. 5, no. 7, p. 468–473, doi:10.1038/ngeo1479. 311
312
Li, G., A. J. West, A. L. Densmore, Z. Jin, R. N. Parker, and R. G. Hilton, 2014, Seismic 313
mountain building: Landslides associated with the 2008 Wenchuan earthquake in the 314
context of a generalized model for earthquake volume balance: Geochemistry, Geophysics, 315
Geosystems, v. 15, no. 4, p. 833–844, doi:10.1002/2013GC005067. 316
317
Lin, C.-W., W.-S. Chang, S.-H. Liu, T.-T. Tsai, S.-P. Lee, Y.-C. Tsang, C.-L. Shieh, and C.-M. 318
Tseng, 2011, Landslides triggered by the 7 August 2009 Typhoon Morakot in southern 319
Taiwan: Engineering Geology, v. 123, no. 1-2, p. 3–12, 320
doi:10.1016/j.enggeo.2011.06.007. 321
322
Lyell, C., 1833, Principles of Geology: London, Murray, 834 p. 323
324
McPhillips, L. E. et al., 2018, Defining Extreme Events: A Cross-Disciplinary Review: Earth’s 325
Future, v. 6, no. 3, p. 441–455, doi:10.1002/2017EF000686. 326
327
Montgomery, D. R., 2001, Slope Distributions, Threshold Hillslopes, and Steady-state 328
Topography: American Journal of Science, v. 301, no. 4–5, p. 432–454, 329
doi:10.2475/ajs.301.4-5.432. 330
331
O’Connor, J. E., and J. E. Costa, 2004, The World’s Largest Floods, Past and Present: Their 332
causes and magnitudes: U.S. Geological Survey Circular 1254, 13 p. 333
334
Pardee, J.T., 1942. Unusual currents in Glacial Lake Missoula, Montana: Geological Society of 335
America Bulletin, v. 53, p. 1569-1600. 336
337
Soennichsen, J., 2008, Bretz’s Flood: The Remarkable Story of a Rebel Geologist and the 338
World's Greatest Flood, Seattle, Sasquatch Books, 289 p. 339
340
Shaw, J., M. Munro-Stasiuk, B. Sawyer, C. Beaney, J. Lesemann, A. Musacchio, B. Rains, R. R. 341
Young; The Channeled Scabland: Back to Bretz?. Geology ; v. 27, no.7, p. 605–608. doi: 342
https://doi.org/10.1130/0091-7613(1999)027<0605:TCSBTB>2.3.CO;2 343
344
Strahler, A. N., 1952, Dynamic basis of geomorphology: Geological Society of America 345
Bulletin, v. 63, no. 9, p. 923, doi:10.1130/0016-7606(1952)63[923:dbog]2.0.co;2. 346
347
Waitt, R. B., 1980, About Forty Last-Glacial Lake Missoula Jökulhlaups through Southern 348
Washington: The Journal of Geology, v. 88, no. 6, p. 653–679, doi:10.1086/628553. 349
350
Waitt, R. B., 1985, Case for periodic, colossal jökulhlaups from Pleistocene glacial Lake 351
Missoula: Geologic Societaly of America Bulletin, v. 96, no. 10, p. 1271-1286. 352
353
354

14

Whipple, K. X., and G. E. Tucker, 1999, Dynamics of the stream-power river incision model: 355
Implications for height limits of mountain ranges, landscape response timescales, and 356
research needs: Journal of Geophysical Research: Solid Earth, v. 104, no. B8, p. 17661– 357
17674, doi:10.1029/1999JB900120. 358
359
Whipple, K. X., A. M. Forte, R. A. Dibiase, N. M. Gasparini, and W. B. Ouimet, 2017, 360
Timescales of landscape response to divide migration and drainage capture: Implications 361
for the role of divide mobility in landscape evolution: J. Geophys. Res.: Earth Surface, v. 362
122, no. 1, p. 248–273, doi:10.1002/2016jf003973. 363
364
Willett, S. D., S. W. Mccoy, J. T. Perron, L. Goren, and C.-Y. Chen, 2014, Dynamic 365
reorganization of river basins: Science, v. 343, no. 6175, p. 1248765-1–1248765-9, 366
doi:10.1126/science.1248765 367
368
369
370
371
372
373
374
375
376
377
378
379

15

CHAPTER 2: Initiation and runout of post-seismic debris flows: 380
Insights from the 2015 Gorkha earthquake 381
2.1.1 INTRODUCTION 382
The mobilization of saturated sediments as debris flows is a critical link in the chain of 383
“cascading hazards” that follow in the wake of large earthquakes, fires, or floods (Cannon et al., 384
2001a; Cannon et al., 2001b; Costa, 1984; Huang and Li, 2014; Jakob and Hungr, 2005). The 385
destructive power of debris flows can be borne out more than 100 kilometers downstream 386
(Iverson, 1997). Consequences include economic and infrastructure damage as well as loss of 387
life, with 10s to 100s of thousands of fatalities globally attributed to debris flows over the last 388
several decades (Dowling and Santi, 2014; Petley, 2012). Debris flows are also a primary driver 389
of erosion and sediment transport in steep landscapes (Stock and Dietrich, 2003; Stock and 390
Dietrich, 2006). Understanding debris flows is thus of deep human and scientific importance. 391
Debris flows after large earthquakes can have particularly damaging consequences. In 392
steep terrain, co-seismic landslides deliver large quantities of debris to colluvial channels 393
(Keefer, 1984; Jibson and Keefer, 1993; Dadson et al., 2004; Xu et al., 2012; Roback et al., 394
2018). In the months and years following, heavy rainfall mobilizes this debris (Iverson et al., 395
1997; Zhang and Zhang, 2017), while new post-seismic landslides form on weakened hillslopes 396
(Marc et al., 2015, Fan et al., 2017) and can generate additional debris flows. With 397
communications, transportation, and supply chains already disrupted and structures weakened, 398
post-seismic debris flows may be deadlier and more destructive than their inter-seismic 399
counterparts (Zhang et al., 2013). Understanding where and when post-seismic debris flows are 400
likely to occur is critical for developing models of prolonged hazards following earthquakes and 401
informing reconstruction strategies. 402

16

Debris flows have been a topic of much laboratory, field, and theoretical study, and the 403
physics governing them is complex (Iverson, 1997), making it important to identify whether 404
there are simple parameters that can help predict their occurrence. Many studies have explored 405
the interaction between precipitation and topography in controlling debris flow initiation and 406
runout (Fan et al., 2017; Iverson, 2014; Wei et al., 2018; Wilford et al., 2004; amongst others), 407
and rainfall thresholds provide a widely-used tool for debris flow prediction following wildfires 408
in steep landscapes (Cannon et al., 2001a; Cannon et al., 2001b, Staley et al., 2013). However, to 409
date, no similar framework has been developed for the aftermath of earthquakes. A major gap in 410
understanding post-seismic debris flows is the lack of large debris flow inventories associated 411
with a single earthquake. Smaller inventories have provided insights into local topography and 412
precipitation controls (Xu et al., 2012; Bertrand et al., 2013; Guo et al., 2016; Jackson et al., 413
1987; Ni et al., 2014; Shieh et al., 2009; Tang et al., 2009; Zhuang et al., 2010), setting the basis 414
for hazard prediction (Tang et al., 2012, Zhou et al., 2014). Yet the scientific understanding 415
needed for landscape-scale hazard assessment after a major earthquake remains lacking. 416
The 2015 Gorkha earthquake in Nepal led to many debris flows, offering a rare 417
opportunity to fill this knowledge gap. In this study, we have comprehensively mapped debris 418
flows occurring in 2015, 2016, and 2017 in the region affected by this earthquake. By 419
distinguishing debris flows generated by different mechanisms, we can identify characteristic 420
channel geometry (specifically initial slope angle of the channel bed) associated with debris flow 421
formation, as well as changes in precipitation thresholds over time. In the context of 422
complementary results from laboratory experiments, these observations bridge scales between 423
local flow processes and landscape-scale response, providing an empirical foundation for hazard 424
prediction and better understanding of the impact of earthquakes on erosion and sediment 425

17

transport. 426
427
2.1.2 The Gorkha Earthquake and associated landslides 428
The MW 7.8 Gorkha earthquake ruptured the Main Himalayan Thrust on April 25, 2015. 429
The epicenter was ~77 km northwest of Kathmandu and the focal depth was ~15 km. Despite the 430
shallow depth and large magnitude, there was no surface rupture (Avouac et al., 2015). Strong 431
ground shaking caused nearly 25,000 landslides (Roback et al., 2018). On May 12, 2015, the 432
strongest aftershock of the Gorkha earthquake sequence occurred with MW 7.3, triggering at least 433
221 new landslides (Martha et al., 2017). 434
The South Asian Monsoon typically takes place in Nepal between June and September. 435
Monsoon rains in 2015 and beyond mobilized loose co-seismic landslide debris and triggered 436
new landslides, some of which produced debris flows. We used high-resolution satellite imagery 437
collected at several time points before and after the Gorkha earthquake to map debris flows 438
(Figure 1). We mapped debris flows from initiation point to the end of the runout, extracted 439
profiles of channels hosting flows, and classified flows according to their apparent initiation 440
mechanisms. 441
442
2.1.3 Multiple mechanisms for post-seismic debris flows: Type-1 vs. Type 2 443
Multiple mechanisms for debris flow generation have been identified in past work 444
(Coussot and Meunier; 1996). First, a debris flow can initiate when loose sediment in the bed of 445
a sufficiently steep channel is subject to stress from fluid flow that exceeds the threshold for 446
motion. The threshold for motion is known as the Shields number and is defined as 447

18

𝜏𝜏 ∗
=
𝜏𝜏 ( 𝜌𝜌 𝑠𝑠 − 𝜌𝜌 ) 𝑔𝑔𝑔𝑔
(1) 448
where 𝜏𝜏 ∗
is a non-dimensional Shields parameter, 𝜏𝜏 is a dimensional shear stress imposed on 449
sediment by fluid flow, 𝜌𝜌 𝑠𝑠 is sediment density, 𝜌𝜌 is fluid density, 𝑔𝑔 is gravitational acceleration, 450
and 𝐷𝐷 is the characteristic dimension of sediment (Iverson et al., 1997; Cao et al., 2006; 451
Prancevic et al., 2014; Shields, 1936). Eq. 1 governs the initiation conditions for sediment 452
motion by fluvial and debris flow transport. These are distinct physical processes and the 453
transition from one to the other depends on channel slope; in flume experiments, debris flows 454
form above a threshold channel bed angle 𝜃𝜃 = 22º (Prancevic et al., 2014). We distinguished 455
debris flows that formed via this mechanism by evaluating whether the source of a given flow 456
was from co-seismic landslide debris. We refer to these debris flows as “Type-1” (Figure 1) and 457
use the Gorkha dataset to assess whether there is a landscape scale topographic threshold for this 458
debris flow mechanism, analogous to that observed in experiments. 459
460
Figure 1. A & B: Example of a debris flow sourced from a co-seismic landslide (a Type-1 461
debris flow) occurring in the monsoon season of 2015. Landslide is bordered in red, debris 462
flow path is indicated with red dashed line. Location is 27.826º N, 85.979º E. C-F: Debris 463
flow mechanisms observed in this study: Type-1 events occur when existing debris from co- 464
seismic landslides is mobilized in monsoon rains; Type-2 occur when new rainfall-triggered 465
hillslope failures are fluidized and continue to travel down channels. 466
467

19

Post-seismic debris flows can also form when a landslide is triggered by monsoon rains, 468
if landslide debris becomes fluidized and continues to flow downstream through colluvial 469
channels (Iverson et al., 1997; Montgomery and Dietrich, 1994). Increased rainfall-driven 470
landslide activity in the years following large earthquakes, likely due to slope weakening during 471
shaking (Marc et al., 2015; Fan et al., 2017; Yang et al., 2017), can generate many debris flows 472
by this mechanism. In this study, we refer to debris flows generated by post-seismic landsliding 473
as “Type-2” (Takahashi, 1978). We identify these debris flows based on their source from storm- 474
triggered landslides that occurred after the earthquake. Time windows between image collection 475
(typically several months) mean that we cannot be sure that every debris flow formed 476
simultaneously with a landslide, but this is a reasonable assumption. We compare these post- 477
seismic, Type-2 debris flows to those sourced from storm-triggered landslides occurring before 478
the earthquake. 479
Finally, we distinguished debris flows sourced from a combination of co-seismic and 480
storm-triggered landslide material, including those formed when new landslides entered channels 481
with preexisting landslide debris present and scenarios in which a co-seismic landslide was 482
reactivated and expanded during monsoon rains. 483
484
2.2 RESULTS AND DISCUSSION 485
2.2.1 Debris flows generated by the Gorkha Earthquake 486
From imagery spanning 2009-2017, we identified a total of 1164 debris flows in the 487
region of central and eastern Nepal where the Gorkha earthquake caused strong shaking (Figure 488
2). Of these, 501 flows were remobilized debris from co-seismic landslides (Type-1) while 371 489
were sourced from post-seismic landslides triggered during ensuing monsoon seasons (Type-2). 490

20

A further 188 debris flows were a combination of the first two types. Many of the Type-2 debris 491
flows occurred in the east of the study region (Figure 2), outside of the zone of intense ground 492
motion during the Gorkha main shock (>10% g), while no Type-1 debris flows were identified in 493
this region, consistent with the lack of coseismic landslides there (Roback et al., 2018).- 494
495
Figure 2. Map of debris flows in the Nepal Himalaya. Debris flows identified as having a 496
co-seismic source (Type-1) include those from landslides triggered both by the main shock 497
and the largest aftershock, though most by far were from the main shock (15). The fault 498
rupture propagated from west to east, hence the lack of debris flows to the west of the 499
epicenter. Contours show peak ground acceleration (PGA) during the main shock. 500
501
945 debris flows occurred during the 2015 monsoon, 33 in 2016, and 20 in 2017 (Figure 502
3B). For a further 63 flows, image coverage was insufficient to identify the year of occurrence. 503
We also mapped 104 Type-2 debris flows occurring in the same region between 2009 and 2015, 504
representing debris flows generated from storm-triggered landslides at the end of the interseismic 505
period. Of these, 64 were from a single large storm in 2009. The total number of pre-Gorkha 506
debris flows may be somewhat underestimated because quality imagery was sparse early in the 507
study period. 508
Assuming the pre-earthquake mapping is representative, the annual rate of Type-2 debris 509
flows increased from an average of ~17 before the Gorkha earthquake, peaking in 2015 (n=325) 510

21

and declining in 2016 (n=23) and 2017 (n=15). Rainfall is expected to strongly influence debris 511
flow activity (e.g. Fan et al., 2017; Staley et al., 2013; Caine, 1980; Dahal and Hasegawa, 2008; 512
Gabet and Mudd, 2006; Tang and Liang, 2008; Zhou and Tang, 2014), but precipitation was 513
similar between these monsoon seasons (Figures 3A, 4), suggesting increased activity in 2015 514
and after was a “hangover” resulting specifically from the earthquake. Changes in debris flow 515
activity in Nepal are consistent with observations from other settings where rainfall-induced 516
landslides became more frequent after large earthquakes but tapered off in following years (Marc 517
et al., 2015, Fan et al., 2017; Yang et al., 2017). 518
Figure 3. A: Intensity-duration plots of rainfall at debris-flow affected areas during the 519
2015-2017 monsoons. B: Number of debris flows of each type identified during the three 520
monsoons following the Gorkha earthquake. 521
522
We can evaluate the increase in Type-2 debris flow activity after the Gorkha 523
earthquake in terms of a lowering of the threshold precipitation required for hillslope failure. For 524
each mapped debris flow, we determined the maximum intensity-duration (I-D) of rainfall during 525
the monsoon season when the debris flow occurred (either 2015, 2016, or 2017), using Global 526
Precipitation Measurement (GPM) rainfall data (Hou et al., 2014). Compared to pre-seismic I-D 527
thresholds for initiation of shallow landslides in Nepal (Dahal and Hasegawa, 2008), Type-2 528

22

debris flows were associated with lower rainfall I-D in 2015 (Figure 3A). For debris flows in 529
2016 and 2017, I-D thresholds increased modestly relative to 2015 but remained below the pre- 530
earthquake threshold even as debris flow rates returned approximately to pre-Gorkha levels. 531
Although absolute quantification of rainfall from GPM data may contain biases (Prakash et al., 532
2018), and all debris flows in each year did not necessarily occur with the maximum I-D rainfall 533
event, we assert that comparison of the relative shift between large populations should yield 534
robust general trends. Similar changes to those in Nepal were documented for debris flows after 535
the 2008 Wenchuan earthquake in Sichuan, China, where the critical rainfall intensity for debris 536
flow triggering decreased by 23.6-69.6% and critical rainfall accumulation by 25.7-49.4% after 537
the earthquake (Tang and Liang, 2008; Zhou and Tang, 2014), with precipitation thresholds 538
increasing again in the following years (Dahal and Hasegawa, 2008). This evolution in landslide 539
activity with time could be due to rock “healing” after seismic weakening (Marc et al., 2015), or 540
to failure of the most vulnerable hillslopes immediately after the earthquake, leaving fewer 541
opportunities for failure in later years. 542
Nearly all Type-1 debris flows occurred during 2015 (Figure 3B). Rates of these debris 543
flows dropped off dramatically in the years following the earthquake, much more than for their 544
Type-2 counterparts. This indicates that co-seismic landslide material in the vast majority of 545
channels was effectively flushed out by storms during the 2015 monsoon season, at least for 546
channels prone to debris flows. In other words, discharge thresholds required to mobilize loose 547
landslide debris were met in most channels in first monsoon after the earthquake (Baer et al., 548
2017). Estimates of Shields stress values suggest this is reasonable: considering a channel bed 549
angle of 25-30°, the 𝜏𝜏 ∗
for debris flow initiation is approximately 0.2 (Prancevic et al., 2014). For 550
a median grain size of landslide debris in Nepal of D = 75 cm (the upper range of measured D50 551

23

values; Attal and Lavé, 2006), sediment density 𝜌𝜌 𝑠𝑠 of 2 g/cm
3
, and fluid density 𝜌𝜌 𝑠𝑠 of 1 g/cm
3
, 552
and assuming shear stress exerted by flow at the bed is 𝜏𝜏 = 𝜌𝜌 𝑔𝑔 𝜌𝜌𝜌𝜌 𝜌𝜌 𝜌𝜌 𝜃𝜃 , flow depth H would need 553
to be ~30 cm to trigger a debris flow in a sediment-laden channel. This value is easily within the 554
range of stage heights observed in first-order channels in Nepal during typical monsoon storms 555
(Brasington and Richards, 2000). Though simplified, this calculation suggests that an average 556
monsoon season has the capacity to trigger Type-1 debris flows wherever material is present in 557
sufficiently steep channels – much as we observe in the post-Gorkha dataset. As a result, we can 558
infer that spatial precipitation patterns would have had little influence on where Type-1 debris 559
flows occurred during the 2015 monsoon season, an important conclusion that may simplify 560
hazard prediction, though timing of precipitation would influence when these debris flows 561
occurred. 562
563
2.2.2 Channel geometry and runout distances 564
To understand what controlled where debris flows occurred and evaluate their potential 565
effects, we identified the channel slope angle at the point of initiation as well as the runout 566
distance for each debris flow. Channel positions extracted from a digital elevation model (DEM) 567
are characterized by nodes centered on DEM cells. We define the initiation zone of a mapped 568
debris flow to be the channel segment between the first two nodes, so depending on their 569
orientation it is either the first 30 or first ~42.2 meters of the channel, due to the 30m DEM 570
resolution. For Type-1 debris flows, the initial slope refers to the slope at the most upstream 571
position where the debris flow was visible in satellite imagery. For Type-2 and mixed-source 572

24

debris flows, initial slope refers to slope of the first channel segment where flow is obviously 573
confined based on visual analysis. 574
Figure 4. GPM precipitation during the 2015 and 2016 monsoons with debris flows from 575
each season. 945 debris flows mapped from 2015, 33 from 2016. 576
577
Channel gradients associated with debris flow initiation are shown in Figures 5 and 6. We 578
identify a threshold gradient for Type-1 debris flow initiation of ~20-23º, noting that only 3 579
debris flows initiated in channels with gradients below 20º and recognizing that some angles may 580
be underestimated due to the DEM resolution. Our observed threshold for Type-1 debris flows 581
agrees closely with the 22º value identified in flume experiments (Prancevic et al., 2014), 582

25

suggesting that this channel slope threshold may be widely relevant across scales, although 583
further testing in other field settings is warranted. 584
Figure 5. Initial slope and runout distance for debris flows of different mechanisms. Box 585
edges are 25th and 75th percentile values and whisker edges are 5th and 95th percentiles. 586
Identical outlier values are horizontally staggered. Orange cross is the mean value and 587
orange line the median. 588
589
Type-2 debris flows develop from rainfall-triggered landslides on hillslopes, and these 590
slopes will inevitably be steeper than the channel heads where we have quantified slope angles. 591
We therefore do not expect to observe a meaningful slope threshold for Type-2 debris flows in 592
our dataset. Nonetheless, channel gradient and angle of entry into a channel are important in 593
determining whether mass failures develop into debris flows (Brayshaw and Hassan, 2009). 594
Interestingly, we find that initial channel slopes associated with Type-2 debris flows are 595
generally less steep than Type-1 (Figures 5, 6). We suggest that Type-2 debris flow generation 596

26

may be possible at lower channel gradients because material is already in motion when it enters 597
channels. 598
Figure 6. Cumulative distribution of channel slope at the initiation zone of debris flows. 599
“All channels” refers to every channel in the area affected by co-seismic landsliding, with 600
slope measured at 0.01 km
2
contributing drainage area for consistency with debris flow 601
data. Table shows initial slope measurements for debris flow channels. 602
603
We determined runout distance for each debris flow based on the extent of visible 604
channel disturbance, including identifying the farthest point downstream where individual 605
boulders were observed to have moved or new gravel bars deposited (Figure 5B). Type-1 debris 606
flows ran out from 30-4850 m, with 95% between 200-3450 m and median distance of 700 m. 607
Post-seismic Type-2 flows ran out from 50-23500 m, with 95% between 200-15650 m and 608
median distance of 1100 m. Pre-seismic Type-2 flows ran out from 100-21100 m with 95% 609
between 250-8500 m and median distance of 3050 m. The longer runouts we observe for Type-2 610
debris flows than Type-1 may indicate greater volumes of material involved in these flows and 611
consequently greater momentum (Iverson, 1997). Type-2 flows may also occur in more saturated 612
conditions, producing more fluidized flows that are able to run out over longer distances. 613
Many of the shortest runouts occurred in tributaries that fed into larger channels after 614
only a few 10s or 100s of meters, with flows frequently ending where the tributary channel 615

27

entered a trunk stream with a gentler gradient. Thus, channel network topology is an important 616
control on runout in this case. We find no obvious relationships between channel geometry (such 617
as slope angles) and runout distance for debris flows in the Gorkha dataset, in contrast to prior 618
field data (Hungr, 1995; Prochaska et al., 2008). Flume experiments also show that initial slope 619
controls the speed and intensity of debris flow initiation for Type-1 debris flows, and ultimately 620
the scale and runout distance of the flow (Hu et al., 2014; Hungr et al., 2008). It is not clear why 621
similar topographic controls are not evident in our Gorkha dataset. 622
623
2.2.3 Hazard analysis: A framework for predicting the locations of Type-1 624
debris flows following large earthquakes 625
Based on our observations of Type-1 debris flows following the Gorkha earthquake, we 626
predict that channels are likely to experience a debris flow at some stage during the monsoon 627
season following a large earthquake if they are 1) steeper than the threshold slope angle for 628
debris flow formation and 2) contain material from co-seismic landslides. These criteria can be 629
evaluated following an earthquake using only elevation models and remote imagery for landslide 630
mapping, so they could provide an effective rapid-response tool for identifying locations that are 631
likely to experience a debris flow during ensuing rains. 632
We used the Gorkha dataset to develop and test a hazard prediction model based on these 633
principles. We used 3 km × 3 km cells to aggregate data, with the rationale that the total length 634
of potentially vulnerable channels within a certain area represents the relative hazard in that area. 635
The 3-km resolution was selected to provide a tractable scale of analysis given the size of the 636
study area; we tested other scales and found similar outcomes. We assumed that a channel could 637
generate a debris flow if it exceeded a threshold slope (S*) and was within a certain distance d* 638

28

of a co-seismic landslide which could contribute sediment (Blahut et al., 2010). For S*, we 639
initially adopted the 22º threshold observed in experimental data (Prancevic et al., 2014) and in 640
our mapping in Nepal (Figure 5). We also tested a range of values from 10 to 49°. For d*, we 641
tested values of 0, 30, 60, 90, 120, and 150 m. We defined channels from the DEM using a 642
minimum drainage area of 0.01 km
2
and identified channel segments that met the criteria for 643
debris flow formation for a given pair of S* and d* values. Since we identified some debris flows 644
initiating in gullies too small to be accurately distinguished by a 30m DEM, as well as debris 645
flows initiating at points other than lowest elevation outlet point on a landslide, we defined d* as 646
a simple 2-D ring buffer around landslide polygons (ArcGIS function “buffer”). Finally, the 647
lengths of these debris flow-prone channels were summed and binned within each 3 km cell, 648
with the total length representing the relative hazard posed by post-seismic debris flows in that 649
cell (Figure 7). 650
Overall, we found that this relatively simple model captures the spatial pattern of 651
observed debris flows well, predicting a high hazard probability where debris flows did occur, 652
and a relatively low probability where they did not. Some Type-1 debris flows occurred in areas 653
with no identified hazard, primarily because they were derived from landslides triggered by the 654
main aftershock (which the landslide inventory used in this study does not intentionally include) 655
or were in areas where early post-earthquake satellite images were obscured by clouds, so 656
landslides were not identified (Roback et al., 2018). 657

29

658
Figure 7. Map of predicted Type-1 debris flow hazard. Mapped debris flows in areas with 659
no recognized hazard (shown in faded color, e.g., in the northeast of the displayed region) 660
were likely sourced from landslides triggered by the major MW 7.3 aftershock on May 12, 661
2015, which were not included in the landslide inventory used in this study (15). Relative 662
hazard is assessed by identifying the total length of channel segments with minimum 0.01 663
km2 contributing drainage area having a slope of at least 22º that are within 30 meters of a 664
co-seismic landslide. Color map was applied using an equal interval scheme. 665
666
For a quantitative evaluation, we compared the total length of channels in which debris 667
flows are predicted to initiate (exceeding gradient S* and within d* of a co-seismic landslide) in 668
each cell to the observed number of Type-1 debris flows in that cell. We optimized the 669
correlation coefficient between observed and predicted debris flow distributions and found best 670
fit values of d* = 30 m and S* = 27° (Figure 9), near the 90% slope threshold from the inventory 671
(Figure 6). However, the optimization was not sensitive to S*, as a wide range of values 672
performed similarly (Figures 9, 10). To test our model predictions of the most vulnerable channel 673
segments, we compared our results to equivalent hazard values estimated in each cell based only 674
on co-seismic landslide density, not considering channel location or slope. Our model yielded a 675
measurable improvement in fit compared to an approach that considers co-seismic landslide 676
density only (r = 0.6781 versus r = 0.5812; Figures 8, 9). 677

30

Figure 8. Plot of modeled hazard
(measured by distance of threshold
channels within each 3 km × 3 km cell)
versus number of mapped debris flows per
cell. Parameter values are 30-meter buffer
distance from coseismic landslides and 27º
threshold slope.

Figure 9. Contour plot of correlation
coefficient between modeled hazard
(length of threshold channels) and
mapped debris flow locations (A) and the
test parameter described in Equation 2
(B) for different parameter
combinations. Modeled hazard is equal
to the total length of channel segments
within a 3 km × 3 km cell within a certain
buffer radius of a coseismic landslide and
exceeding a threshold angle.

To aid in identifying optimal values
for S* and d* and ensure that a small
number of cells do not bias the fit, we
performed cross-validations. 10% of spatial
cells were left out, selected randomly. On
the remaining 90% (training dataset), we
identified best-fit values for S* and d* that
maximized correlation between predicted hazard and mapped debris flow locations and
minimized a test statistic T (see Methods). Best-fit parameters identified in the training dataset

31

were tested on the remaining 10% (validation dataset). Results from 10,000 trials are shown in
Figure 10. Cross-validation narrowed the ranges of best-fit parameters compared to maximizing
the correlation coefficient between observed and predicted hazard. In cross-validation, the model
fit remained relatively insensitive to chosen values of S*.
678
Figure 10. Results of 10,000 cross-validation iterations for the test statistic described in 679
Equation 2 and correlation coefficient. Each cell represents a pair of parameter values 680
(buffer distance and threshold slope), and the trial count indicates the number of trials in 681
which the given value pair was optimal. Histograms at bottom show fit of the optimal 682
parameters identified in each trial on the training dataset applied to the 10% left out. 683
684
This model could be applied rapidly using only a DEM (readily available globally) and a 685
map of co-seismic landslides. For this study, we used a manually mapped landslide inventory 686
(Roback et al., 2018), but in a rapid-response context a landslide map generated by automated 687
image analysis or even a modeled landslide distribution could be used (Gallen et al., 2017, 688
Robinson et al., 2017), since some of the main drawbacks of automatically generated inventories 689

32

(such as agglomerating adjacent landslides; Li et al., 2014) should not affect results. Thus, debris 690
flow forecasts could be provided in a matter of days after an earthquake. Assessing debris flow 691
susceptibility by this method could allow early warning to residents of high-risk areas and better 692
allocation of disaster response resources. 693
Our approach identifies channel segments where Type-1 debris flows are likely to 694
initiate, but conditions causing slope failures that lead to Type-2 debris flows are not accounted 695
for. Consequently, our model will not capture all locations where debris flows might occur after 696
an earthquake and may overlook high hazard in some areas. For example, Type-2 debris flows 697
dominated in the eastern part of the mapped area following the Gorkha earthquake, far from the 698
epicenter (Figure 2). This may because shaking in this area (either due to the main shock or the 699
Mw 7.3 aftershock) was insufficient to trigger many co-seismic landslides but nonetheless 700
weakened rocks and reduced the precipitation threshold for new landslides (Figure 3). In any 701
case, of the 1060 debris flows mapped for this study that occurred after the Gorkha earthquake, 702
689 are at least partially derived from co-seismic landslides. By identifying locations where these 703
debris flows are likely to occur, this model on its own could provide useful information in the 704
aftermath of an earthquake, and future work could target predictions of Type-2 debris flows. 705
706
2.2.4 Comparison between DEM resolutions 707
Topographic metrics are highly dependent on DEM resolution, so it is important to 708
identify any potential biases introduced by the 30-meter resolution of the ASTER-patched SRTM 709
DEM (Deng et al., 2007). To accomplish this, we use a 2-meter resolution DEM derived from in- 710
track stereo imagery using the Surface Extraction with Triangulated Irregular Network-based 711
Search-space Minimization (SETSM) algorithm (Noh and Howat, 2015). The SETSM DEM 712

33

covers the Melamchi Valley northeast of Kathmandu, near the region of highest density of co- 713
seismic landslides and post-seismic debris flows (Figure 11). 714
715
Figure 11. Location of SETSM 2-meter DEM coverage with respect to debris flow 716
inventory. 717
718
We compared debris flow channels extracted from both the 30 and 2-meter DEMs to 719
identify any systemic differences in channel geometry as determined by different topographic 720
data. Included in the area with SETSM coverage were 47 Type-1 debris flows, 8 post-seismic 721
Type-2 debris flows, 27 mixed-source debris flows, and 1 pre-Gorkha Earthquake Type-2 debris 722
flow, for a total of 83 debris flows. Figure 12 shows comparisons of debris flow channel metrics 723
derived from the different DEMs. We find the slope distributions of initiation zones of debris 724
flows (Figure 12B) are quite similar as derived from both DEMs. Initiation zones derived from 725
both DEMs cover the same distance along the channel. Despite the 2-meter DEM sampling the 726
channel slope at more points, commensurate with the difference in cellsize, the middle 90% of 727
the slope distributions are nearly identical. However, looking at the entire runout distance of the 728
debris flow channels shows that channels derived from the 2-meter DEM are systematically less 729

34

steep than those derived from the 30-meter DEM. This holds true looking at slope distributions 730
alone as well as channels viewed in slope-area space (Figures 12A, 12C). 731
This difference could be important when we examine features such as the point of 732
divergence between debris flow dominated and non-debris flow dominated channels (See 733
Chapter 3 for further discussion of this matter). The uncertainties introduced by DEM resolution 734
may affect the exact values of geometric thresholds associated with debris flows, although data 735
on the initiation zones appear to be robust across DEM resolutions. 736
737
2.2.5 Confluence angles and debris flow runout 738
It has been established that the angle at which a tributary enters a trunk stream is an 739
important control on whether a debris flow initiating in that tributary will remain mobile through 740
the confluence (Benda and Cundy, 1990; Brayshaw and Hassan, 2009). The large dataset of 741
debris flows assembled for this study allows for an extensive look at this trend. We measured 742
junction angles for debris flows that either ended at or passed through a confluence. Confluences 743
were included only where the debris flow occupied the channel with a lower order and/or smaller 744
contributing drainage area. Flows occurring in channels that were tributaries to a channel with 745
another Gorkha debris flow were not included in the total regarding the endpoints of flows, as it 746
was impossible to tell whether they passed through the confluence in question, although where 747
these flows unambiguously passed through a confluence upstream of their endpoints the angle 748
measurement was included. 749

35

750
751
Figure 12A. Slope-area plots of debris flows in Melamchi Valley derived from 2-meter 752
SETSM and 30-meter SRTM DEMs. Bars represent upper and lower quartiles for each 753
area bin and are drawn on bin centers. 12B. Initial slopes of debris flows in Melamchi 754
Valley. Box edges are 25th and 75th percentile values and whisker edges are 5th and 95th 755
percentiles. Identical outlier values are horizontally staggered. Red cross is the mean value 756

36

and red line the median. 12C. Cumulative distributions of channel slopes in debris flow 757
channels in Melamchi Valley. 758

Figure 13. Confluence angles
along debris flow paths. Box
edges are 25th and 75th
percentile values and
whisker edges are 5th and
95th percentiles. Identical
outlier values are
horizontally staggered. Red
cross is the mean value and
red line the median.
Confluence angles were
measured as shown at right.
Arrow indicates direction of
flow.

759
760
We found that 81.6% of debris flows ended at a confluence. Of these, 91.6% ended at the 761
first confluence they encountered. So, the vast majority of debris flows’ runouts are limited by 762
the length of the first-order tributary in which they initiate. We chose to measure the angles of 763
tributaries with trunk streams in the downstream direction, in contrast to previous work, as the 764
downstream angle more realistically reflects controls on flow runout. The average confluence 765
angle where a debris flow passed through is 141°, the average angle where a debris flow 766
terminated was 105° (Figure 13). Debris flows with long runouts extending into higher order 767
streams typically initiated in the headwaters of the higher order stream rather than entering from 768
a lateral tributary. Other factors such as channel confinement and slope are likely to influence 769
debris flow runout as well (Benda and Cundy, 1990; Brayshaw and Hassan, 2009). Accounting 770

37

for more aspects of the geometry of confluences, particularly using high-resolution topography, 771
could yield important advances in our understanding of hazards associated with debris flow 772
runout. 773
774
2.3 CONCLUSIONS 775
The inventory of post-seismic debris flows compiled in this study yielded new insights 776
into landscape-scale post-seismic debris flow processes, expanding upon prior field and 777
laboratory studies. We identified 1,060 debris flows occurring in the monsoon seasons following 778
the Gorkha earthquake, compared to 104 occurring in the same area during the six years 779
immediately preceding. Debris flow rate rapidly declined after the first monsoon following the 780
earthquake. We identified distinct initial slope angles for debris flows operating by different 781
processes: Type-1 debris flows sourced from co-seismic landslides tended to occur in the 782
steepest channels, with a threshold for initiation similar to the 22° angle observed in flume 783
experiments, while Type-2 flows sourced from storm-triggered post-seismic and interseismic 784
landslides may occur in generally shallower sloping channels. Type-2 debris flows had generally 785
longer runouts than Type-1 flows; however no obvious relationships between channel geometry 786
and runout distance were apparent, and runout distance was partially controlled by the length of 787
the tributary in which a debris flow initiated. Type-1 debris flows occurred almost exclusively 788
during the 2015 monsoon season, indicating that co-seismic landslide debris is largely flushed 789
out of steep channels during the first major rainfalls after an earthquake. 790
We developed a simple model for assessing post-seismic Type-1 debris flow hazards that 791
can be applied rapidly in the wake of a major earthquake using easily obtainable data. Identifying 792
channels which both exceed a threshold slope and have access to co-seismic landslide material is 793

38

an effective method by which to assess areas where Type-1 debris flows are likely to occur 794
during heavy rainfall. We expect that further refinement of this method could yield improved 795
predictive capacity, and conclusions from this study should be tested in other regions where 796
controls on post-seismic debris flows may differ. 797
798
2.4 METHODS 799
2.4.1 Debris flow mapping 800
We mapped debris flows in the region affected by the Gorkha earthquake using Google 801
Earth Pro and DigitalGlobe satellite imagery (Table S1). Debris flow were identified based on 802
spectral contrast between disturbed and undisturbed parts of the landscape and characteristic 803
visible features. To distinguish debris flows from non-flow landslides, we identified features like 804
scour, bottlenecking, obvious confinement, and long runouts. Correspondence of these features 805
in satellite imagery to debris flow channels was confirmed through field observations of several 806
post-Gorkha debris flows. We defined the initiation zone of debris flows to be the most upstream 807
point where these features appear in a channel. Near-constant cloud cover during the monsoon 808
meant the timing of most debris flows could only be defined within a few months. Therefore, we 809
distinguished only whether a debris flow occurred before or after the Gorkha earthquake, then 810
subdivided by year of occurrence. While there are many channels in the Nepal Himalaya with 811
obvious debris flow deposits lining their beds, we included only those where before and after 812
imagery were both available. Precipitation data in Figures 3A and S1 are from GPM 813
3IMERGHH with 0.1º -by-0.1º spatial and 30-minute temporal resolution (Hou et al., 2014). 814
815
816

39

2.4.2 Topographic analysis 817
We extracted channel data using the Shuttle Radar Topography Mission (SRTM) 30m 818
DEM. Voids in SRTM data were patched with the Advanced Spaceborne Thermal Emission and 819
Reflection Radiometer (ASTER) 30m DEM. We used a minimum of 0.01 km
2
drainage area as a 820
threshold to define a channel. At smaller contributing drainage areas, channels defined by a 30m 821
DEM do not correspond accurately with channels in imagery. Some debris flows appear to 822
initiate at smaller contributing drainage areas, but the 0.01 km
2
threshold was selected as a 823
compromise value that captured the initiation zone for most of the mapped debris flows while 824
maintaining the integrity of DEM-derived channels with respect to those visible in imagery. For 825
the 2m SETSM-derived DEM we use the same method to identify channels, although the higher 826
resolution allows us to distinguish debris flow channel heads at smaller drainage areas than 0.01 827
km
2
, so we set no minimum threshold. Topographic metrics were calculated using TopoToolbox 828
2 (Schwanghart and Scherler, 2014). Slope is a scale-dependent quantity, so results should be 829
interpreted considering the resolution of the elevation data. Measurements of slope in this study 830
represent an average angle over a stream segment with a 10-meter minimum drop (to generate 831
the cumulative distribution function plots and identify slope thresholds; Figures 4, 5). Slope 832
angles for the hazard analysis were calculated using a steepest descent algorithm to avoid 833
underestimating the slope of stream segments where debris flows could potentially initiate. 834
In cross-validations for the hazard model, we designed a test statistic T that can be 835
adjusted to penalize cells with debris flows but no identified vulnerable channels and vice versa 836
to different degrees, defined as 837
𝑇𝑇 =
1
∑( 𝑟𝑟𝑟𝑟 − 𝑎𝑎 𝑧𝑧 𝑟𝑟 − 𝑏𝑏 𝑧𝑧 𝑑𝑑 )
(2) 838

40

where r is the normalized length of threshold channels within a specified buffer distance of a co- 839
seismic landslide in a cell, d is the number of debris flows in a cell, zr is the normalized length of 840
threshold channels in cells with zero mapped debris flows, zd is the number of debris flows in 841
channels with zero identified threshold channels, and a and b are penalty coefficients. We used 842
values of 3.5 and 4 for a and b, respectively, which produced a narrow range of potential best-fit 843
parameter values but can be altered to suit different needs (i.e., to penalize cells with low 844
predicted hazard but many debris flows). In the cross-validations, the data excluded in each trial 845
were selected by the Matlab “randi” function. Confluence angles were measured from Google 846
Earth imagery using the angle measurement tool in QGIS. 847
Table S1. Metadata for DigitalGlobe imagery used for debris flow mapping. These images 848
were supplemented with Google Earth imagery. 849
850
Source Collected Image ID
Area
Clouds
Image
Clouds
Area
Off
Nadir
Image
Off
Nadir
Bands
Max
GSD
Sun
Elevation
Max
Target
Azimuth
WV02 5/16/2018 103001007E8B3A00 0.10% 46.00% 22.0° 23.0° 8-BANDS 0.54m 73.3° 2.8°
WV03 5/15/2018 104001003E9E5700 0.00% 48.00% 6.9° 14.3° 8-BANDS 0.33m 75.8° 204.4°
WV03 5/15/2018 104001003D97DF00 0.00% 48.00% 18.7° 18.7° 8-BANDS 0.34m 75.7° 199.9°
WV02 5/8/2018 103001007DB52B00 10.70% 4.00% 22.3° 24.5° 8-BANDS 0.55m 71.5° 351.0°
WV02 5/8/2018 103001007C90B700 1.60% 16.00% 15.7° 18.5° 8-BANDS 0.51m 71.3° 336.3°
WV02 5/8/2018 103001007D52E400 4.80% 20.00% 13.1° 13.7° 8-BANDS 0.49m 71.2° 311.2°
WV02 5/8/2018 103001007D735000 1.00% 7.00% 9.9° 11.2° 8-BANDS 0.48m 71.2° 282.2°
WV02 5/8/2018 103001007DA0DC00 4.60% 5.00% 20.1° 20.3° 8-BANDS 0.52m 71.2° 223.8°
WV02 5/8/2018 103001007D3F3B00 6.90% 20.00% 29.0° 30.2° 8-BANDS 0.61m 71.1° 217.0°
WV03 4/14/2018 104001003A8B1D00 0.00% 9.00% 15.1° 16.2° 8-BANDS 0.33m 68.7° 54.2°
WV03 4/14/2018 104001003B089000 11.90% 19.00% 27.0° 27.3° 8-BANDS 0.38m 67.8° 188.2°
WV02 4/13/2018 103001007B56D600 9.80% 13.00% 24.9° 24.9° 8-BANDS 0.55m 67.7° 65.4°
WV02 4/5/2018 103001007A373D00 0.00% 73.00% 12.7° 12.7° 8-BANDS 0.48m 64.0° 61.8°
WV02 3/31/2018 103001007C657500 7.30% 49.00% 11.8° 12.6° 8-BANDS 0.48m 61.6° 206.7°
WV02 3/31/2018 103001007AB2FC00 0.00% 35.00% 15.8° 20.9° 8-BANDS 0.52m 61.6° 200.1°
WV03 3/26/2018 104001003AA0A100 0.00% 0.00% 11.5° 11.6° 8-BANDS 0.32m 60.4° 223.0°
WV02 3/23/2018 103001007C2EE200 7.00% 4.00% 28.9° 29.2° 8-BANDS 0.59m 56.5° 253.8°
WV03 3/19/2018 10400100396E7500 0.00% 0.00% 27.2° 27.2° 8-BANDS 0.38m 57.6° 227.8°
WV03 3/19/2018 104001003873F900 0.00% 0.00% 28.8° 28.8° 8-BANDS 0.39m 57.1° 227.8°
WV02 3/12/2018 103001007AA4F400 0.00% 16.00% 27.1° 27.1° 8-BANDS 0.58m 52.8° 321.5°
WV02 3/6/2018 103001007B72C600 0.00% 0.00% 27.6° 27.8° 8-BANDS 0.58m 53.9° 67.1°
WV02 1/30/2018 10300100770E9C00 0.00% 0.00% 22.5° 23.9° 8-BANDS 0.54m 41.1° 187.9°
WV02 1/27/2018 103001007720C100 0.00% 70.00% 22.9° 22.9° 8-BANDS 0.54m 40.6° 15.6°
WV02 1/27/2018 1030010076A2D700 0.00% 74.00% 20.5° 23.1° 8-BANDS 0.54m 40.5° 195.8°
WV02 1/5/2018 1030010077CF6800 0.00% 0.00% 23.6° 23.6° 8-BANDS 0.54m 37.5° 81.5°
WV02 1/5/2018 1030010077B9FE00 0.00% 0.00% 23.3° 24.0° 8-BANDS 0.54m 38.1° 110.4°
WV02 1/5/2018 1030010078B70800 0.00% 0.00% 22.8° 23.6° 8-BANDS 0.54m 38.0° 118.9°
WV03 1/3/2018 104001003617EE00 0.00% 0.00% 13.6° 13.6° 8-BANDS 0.32m 37.5° 290.7°
WV03 1/3/2018 1040010037690C00 0.00% 0.00% 27.2° 27.2° 8-BANDS 0.38m 36.6° 259.1°
WV02 12/23/2017 10300100758A9F00 0.00% 0.00% 24.0° 24.1° 8-BANDS 0.55m 36.1° 220.4°

41

WV03 12/16/2017 1040010035251F00 0.00% 0.00% 8.9° 9.3° 8-BANDS 0.32m 37.4° 359.8°
WV03 12/9/2017 1040010035CF7200 0.00% 0.00% 26.3° 26.3° 8-BANDS 0.37m 38.1° 210.7°
WV02 12/9/2017 103001007574 0.00% 0.00% 10.0° 17.7° 8-BANDS 0.50m 37.3° 20.4°
WV02 12/9/2017 1030010075A3EF00 0.30% 2.00% 2.3° 3.3° 8-BANDS 0.46m 37.5° 64.7°
WV02 12/9/2017 1030010076CF6C00 0.00% 0.00% 24.0° 24.3° 8-BANDS 0.55m 37.6° 171.0°
WV02 12/6/2017 1030010076B62300 0.00% 0.00% 22.2° 22.4° 8-BANDS 0.53m 39.0° 71.6°
WV02 12/6/2017 1030010076360100 0.00% 2.00% 28.0° 28.0° 8-BANDS 0.58m 39.1° 95.5°
WV02 12/6/2017 103001007482CC00 0.00% 0.00% 26.0° 26.6° 8-BANDS 0.57m 39.1° 102.1°
WV02 12/6/2017 1030010075C63500 0.00% 0.00% 22.1° 24.7° 8-BANDS 0.55m 39.0° 143.6°
WV02 11/20/2017 1030010073261700 7.10% 3.00% 17.6° 18.3° 8-BANDS 0.51m 40.5° 200.9°
WV02 11/20/2017 10300100739E7C00 0.00% 0.00% 26.2° 26.4° 8-BANDS 0.56m 40.9° 179.8°
WV02 11/12/2017 103001007553EA00 0.00% 0.00% 28.2° 28.4° 8-BANDS 0.58m 43.6° 182.0°
WV02 11/9/2017 103001007591B400 3.90% 11.00% 3.6° 5.4° 8-BANDS 0.47m 43.7° 140.2°
WV03 10/27/2017 1040010034573900 0.00% 7.00% 12.6° 12.8° 8-BANDS 0.32m 48.4° 68.3°
WV03 10/20/2017 104001003475F900 0.00% 1.00% 22.8° 24.1° 8-BANDS 0.36m 49.7° 256.6°
WV02 9/16/2017 103001007140CA00 8.40% 37.00% 24.8° 27.0° 8-BANDS 0.57m 59.8° 344.8°
WV02 9/16/2017 1030010070CF6E00 0.00% 17.00% 13.4° 14.6° 8-BANDS 0.49m 59.7° 286.5°
WV02 9/16/2017 103001007169D000 7.80% 17.00% 21.7° 22.2° 8-BANDS 0.53m 59.6° 232.2°
WV02 9/5/2017 1030010070354500 0.00% 34.00% 4.0° 4.0° 8-BANDS 0.46m 63.8° 270.4°
WV02 8/20/2017 103001006F142F00 4.80% 63.00% 7.2° 7.2° 8-BANDS 0.47m 67.4° 301.3°
WV03 5/5/2017 104001002C5E7000 7.60% 28.00% 28.7° 28.8° 8-BANDS 0.39m 73.6° 196.0°
WV02 3/21/2017 10300100669B9800 0.00% 0.00% 25.7° 26.6° 8-BANDS 0.57m 57.7° 28.1°
WV02 3/8/2017 103001006516C000 7.10% 27.00% 28.0° 28.2° 8-BANDS 0.58m 50.2° 304.0°
WV03 3/4/2017 104001002863FB00 12.70% 26.00% 12.7° 14.6° 8-BANDS 0.33m 51.5° 239.9°
WV02 2/25/2017 1030010066104700 0.00% 0.00% 22.5° 22.7° 8-BANDS 0.53m 46.6° 304.3°
WV02 2/8/2017 103001006327B300 0.00% 0.00% 24.9° 25.0° 8-BANDS 0.55m 43.5° 151.0°
WV03 2/7/2017 1040010027B22A00 0.00% 0.00% 26.4° 26.4° 8-BANDS 0.37m 43.2° 214.5°
WV02 12/30/2016 1030010063CAF400 1.30% 9.00% 27.9° 28.1° 8-BANDS 0.58m 35.8° 206.9°
WV02 12/11/2016 103001006258C900 0.00% 0.00% 23.7° 23.7° 8-BANDS 0.54m 35.9° 238.8°
WV03 12/7/2016 1040010025407600 0.00% 7.00% 26.1° 27.2° 8-BANDS 0.38m 37.4° 324.4°
WV03 12/7/2016 10400100255B6800 0.00% 8.00% 24.8° 25.7° 8-BANDS 0.37m 37.3° 246.7°
WV02 11/30/2016 1030010061076200 0.30% 42.00% 15.5° 15.7° 8-BANDS 0.49m 38.3° 182.1°
WV02 11/22/2016 10300100615B4800 0.00% 2.00% 27.3° 28.2° 8-BANDS 0.58m 40.1° 200.9°
WV03 11/6/2016 1040010024227100 0.00% 0.00% 23.7° 24.3° 8-BANDS 0.36m 44.0° 328.7°
WV03 11/6/2016 10400100246F4300 0.00% 0.00% 20.0° 20.8° 8-BANDS 0.35m 44.0° 249.4°
WV02 11/6/2016 103001005E57BA00 0.00% 0.00% 29.7° 30.2° 8-BANDS 0.61m 42.8° 330.1°
WV02 11/6/2016 103001005F4DF300 0.00% 0.00% 13.8° 14.6° 8-BANDS 0.49m 43.9° 346.8°
WV02 11/6/2016 103001005FAB7A00 0.00% 0.00% 19.4° 19.5° 8-BANDS 0.51m 43.0° 270.5°
WV02 11/6/2016 1030010060138100 0.00% 0.00% 25.2° 25.2° 8-BANDS 0.55m 42.7° 251.3°
WV02 11/6/2016 103001005E8DD500 0.00% 0.00% 25.5° 25.9° 8-BANDS 0.56m 43.8° 208.1°
WV02 11/3/2016 103001005F693200 0.00% 0.00% 6.8° 7.5° 8-BANDS 0.47m 45.3° 193.0°
WV03 10/31/2016 104001002503F400 0.00% 0.00% 23.2° 24.6° 8-BANDS 0.37m 48.4° 46.7°
WV03 10/31/2016 1040010024985C00 2.40% 1.00% 19.5° 21.7° 8-BANDS 0.35m 48.3° 148.2°
WV03 10/31/2016 10400100247E5500 1.80% 1.00% 22.9° 24.9° 8-BANDS 0.37m 48.4° 159.5°
WV02 10/23/2016 103001005C3E2000 0.00% 0.00% 6.5° 6.5° 8-BANDS 0.47m 48.1° 127.6°
WV02 10/23/2016 103001005FD14300 0.90% 9.00% 26.4° 27.4° 8-BANDS 0.58m 48.4° 174.4°
WV02 10/18/2016 103001005F4AF100 0.00% 0.00% 13.8° 14.9° 8-BANDS 0.49m 49.0° 281.1°
WV02 10/15/2016 103001005D04A100 1.10% 72.00% 20.1° 20.2° 8-BANDS 0.52m 52.0° 47.6°
WV02 10/12/2016 103001005D4D5400 0.00% 0.00% 24.1° 25.3° 8-BANDS 0.56m 52.0° 39.1°
WV03 10/12/2016 10400100234CAB00 9.50% 36.00% 20.4° 20.5° 8-BANDS 0.35m 52.2° 276.5°
WV02 10/2/2016 103001005D1C6500 14.30% 42.00% 23.7° 28.1° 8-BANDS 0.59m 54.3° 330.6°
WV02 10/2/2016 103001005EAA0A00 10.80% 39.00% 24.0° 27.6° 8-BANDS 0.58m 54.2° 259.2°
WV02 9/21/2016 103001005D88C100 0.00% 6.00% 22.9° 24.7° 8-BANDS 0.55m 58.3° 350.0°
WV02 9/21/2016 103001005CA21200 0.00% 6.00% 26.5° 27.3° 8-BANDS 0.57m 58.1° 216.7°
WV03 8/24/2016 10400100211A3000 0.00% 36.00% 19.0° 20.3° 8-BANDS 0.34m 66.7° 253.8°
WV02 8/6/2016 1030010059653B00 12.80% 50.00% 21.4° 21.9° 8-BANDS 0.53m 66.3° 312.6°
WV03 5/17/2016 104001001C682B00 0.00% 30.00% 26.0° 26.4° 8-BANDS 0.37m 76.7° 57.3°
WV02 5/17/2016 10300100562D5400 6.20% 58.00% 26.2° 28.0° 8-BANDS 0.58m 69.0° 323.4°
WV02 5/17/2016 103001005615CF00 0.80% 66.00% 24.8° 24.9° 8-BANDS 0.55m 68.8° 254.6°
WV02 4/30/2016 1030010055830600 0.00% 0.00% 26.7° 26.7° 8-BANDS 0.57m 69.8° 34.2°
WV03 4/28/2016 104001001C72BD00 0.00% 0.00% 19.9° 22.0° 8-BANDS 0.35m 72.8° 71.3°
WV03 4/28/2016 104001001BA1A600 0.00% 0.00% 23.9° 25.5° 8-BANDS 0.37m 72.7° 149.3°

42

WV02 4/25/2016 10300100542C1200 0.00% 0.00% 10.7° 17.5° 8-BANDS 0.51m 67.1° 353.9°
WV02 4/25/2016 103001005530D300 0.00% 0.00% 5.8° 6.9° 8-BANDS 0.47m 66.9° 292.8°
WV02 4/22/2016 103001005401B300 7.90% 11.00% 24.3° 25.2° 8-BANDS 0.55m 68.0° 41.2°
WV02 4/22/2016 103001005444C600 10.30% 13.00% 16.9° 18.4° 8-BANDS 0.51m 67.8° 154.4°
WV03 4/21/2016 104001001BB3A500 6.80% 12.00% 25.2° 25.5° 8-BANDS 0.37m 70.4° 39.8°
WV03 4/21/2016 104001001C63D900 10.50% 12.00% 21.4° 22.0° 8-BANDS 0.35m 70.2° 164.2°
WV02 4/6/2016 1030010053367600 2.40% 34.00% 26.6° 28.3° 8-BANDS 0.58m 61.8° 7.6°
WV02 4/6/2016 1030010053822100 1.90% 35.00% 19.8° 26.0° 8-BANDS 0.56m 61.8° 6.4°
WV02 4/6/2016 10300100557E6500 5.20% 23.00% 7.0° 8.1° 8-BANDS 0.47m 61.9° 199.8°
WV02 3/29/2016 10300100540E9500 0.00% 0.00% 25.0° 26.6° 8-BANDS 0.57m 57.8° 336.0°
WV02 3/29/2016 10300100537A1100 0.00% 0.00% 19.5° 19.6° 8-BANDS 0.51m 57.7° 251.4°
WV02 3/18/2016 1030010053627700 10.40% 30.00% 4.1° 11.3° 8-BANDS 0.48m 55.0° 343.0°
WV02 3/10/2016 10300100515B1B00 0.00% 3.00% 23.4° 25.1° 8-BANDS 0.55m 51.8° 358.2°
WV02 3/10/2016 1030010052D6A100 0.00% 0.00% 15.3° 22.7° 8-BANDS 0.54m 51.6° 351.5°
WV02 3/10/2016 10300100529EFF00 0.70% 6.00% 10.5° 11.7° 8-BANDS 0.48m 51.6° 318.1°
WV02 3/10/2016 10300100511B4C00 0.00% 2.00% 10.3° 14.2° 8-BANDS 0.49m 51.4° 301.2°
WV02 3/10/2016 1030010053124200 5.20% 5.00% 16.9° 17.4° 8-BANDS 0.50m 51.3° 236.5°
WV02 2/25/2016 103001005306F500 0.00% 0.00% 14.6° 23.0° 8-BANDS 0.54m 47.2° 68.3°
WV02 2/25/2016 1030010053903D00 0.00% 0.00% 11.2° 11.3° 8-BANDS 0.48m 47.2° 79.5°
WV03 2/18/2016 10400100180D2A00 0.00% 0.00% 17.3° 17.4° 8-BANDS 0.33m 46.6° 71.5°
WV02 2/1/2016 1030010050A78F00 1.50% 4.00% 27.6° 28.1° 8-BANDS 0.58m 39.1° 331.7°
WV02 2/1/2016 103001004F9CCA00 0.50% 2.00% 20.9° 21.0° 8-BANDS 0.52m 39.0° 259.9°
WV02 1/18/2016 103001004F18F100 0.00% 0.00% 25.6° 27.7° 8-BANDS 0.58m 37.0° 22.3°
WV02 1/18/2016 10300100509CBE00 0.00% 0.00% 17.7° 18.9° 8-BANDS 0.51m 37.2° 46.3°
WV02 1/18/2016 10300100509CB700 0.00% 0.00% 6.7° 7.2° 8-BANDS 0.47m 36.9° 158.8°
WV02 1/18/2016 103001004F9EC300 0.00% 0.00% 24.6° 24.7° 8-BANDS 0.55m 37.1° 172.7°
WV03 1/11/2016 104001001625BD00 0.20% 11.00% 25.4° 26.6° 8-BANDS 0.38m 38.6° 58.3°
WV03 1/11/2016 10400100161F5700 0.00% 0.00% 22.4° 22.5° 8-BANDS 0.35m 38.2° 67.8°
WV03 1/11/2016 1040010017A28800 0.00% 6.00% 22.3° 23.7° 8-BANDS 0.36m 38.6° 140.5°
WV03 1/11/2016 10400100175D6B00 0.00% 0.00% 27.4° 27.5° 8-BANDS 0.38m 38.1° 151.9°
WV02 1/7/2016 103001004F89CB00 3.70% 1.00% 24.7° 25.3° 8-BANDS 0.56m 36.4° 58.1°
WV02 1/7/2016 103001004E783B00 8.70% 1.00% 25.4° 25.6° 8-BANDS 0.56m 36.3° 150.4°
WV02 12/30/2015 103001004E97D900 0.00% 0.00% 8.0° 8.2° 8-BANDS 0.47m 35.9° 98.1°
WV02 12/30/2015 103001004F8D6600 0.00% 0.00% 5.1° 5.2° 8-BANDS 0.47m 35.1° 205.1°
WV03 12/28/2015 1040010015779000 1.70% 3.00% 27.0° 27.5° 8-BANDS 0.38m 36.3° 321.5°
WV03 12/28/2015 104001001601E900 0.80% 1.00% 24.9° 25.3° 8-BANDS 0.37m 36.3° 252.8°
WV02 12/27/2015 103001004D970700 0.00% 0.00% 27.4° 27.4° 8-BANDS 0.58m 37.1° 70.7°
WV02 12/27/2015 103001004D6D6A00 0.00% 0.00% 26.2° 26.2° 8-BANDS 0.56m 36.1° 72.6°
WV02 12/27/2015 103001004F206A00 0.00% 0.00% 24.4° 26.8° 8-BANDS 0.57m 37.0° 132.4°
WV02 12/27/2015 1030010050A68F00 0.00% 0.00% 27.9° 28.1° 8-BANDS 0.58m 36.1° 140.2°
WV03 12/23/2015 1040010016006C00 0.00% 0.00% 27.8° 28.0° 8-BANDS 0.38m 37.3° 49.3°
WV03 12/23/2015 10400100163ADE00 0.00% 0.00% 25.2° 25.3° 8-BANDS 0.37m 37.4° 63.9°
WV03 12/23/2015 10400100162FEB00 0.00% 0.00% 23.5° 23.6° 8-BANDS 0.36m 37.3° 134.7°
WV02 12/14/2015 103001004FB31300 0.00% 0.00% 18.8° 19.0° 8-BANDS 0.51m 35.8° 217.6°
WV03 12/10/2015 1040010015192300 9.60% 5.00% 18.8° 20.2° 8-BANDS 0.35m 37.1° 348.8°
WV03 12/10/2015 1040010015813900 0.00% 5.00% 12.5° 12.5° 8-BANDS 0.32m 37.1° 337.9°
WV02 11/25/2015 103001004D88CD00 1.90% 17.00% 14.3° 14.6° 8-BANDS 0.49m 39.4° 209.7°
WV03 11/21/2015 1040010014A4F400 0.00% 0.00% 24.2° 26.0° 8-BANDS 0.37m 41.6° 148.1°
WV03 11/21/2015 10400100147C5900 0.00% 0.00% 28.7° 28.9° 8-BANDS 0.39m 41.6° 153.1°
WV02 11/17/2015 103001004D568700 0.00% 0.00% 21.9° 26.0° 8-BANDS 0.56m 40.3° 335.0°
WV02 11/17/2015 103001004DBAED00 0.00% 0.00% 19.1° 20.4° 8-BANDS 0.52m 40.3° 313.5°
WV02 11/17/2015 103001004C1AB400 0.00% 0.00% 20.5° 20.9° 8-BANDS 0.52m 40.2° 256.6°
WV02 11/17/2015 103001004D729C00 0.00% 0.00% 26.9° 28.0° 8-BANDS 0.58m 40.2° 237.0°
WV03 11/14/2015 1040010013CE5B00 0.00% 0.00% 12.7° 13.1° 8-BANDS 0.32m 42.2° 319.6°
WV03 11/14/2015 1040010013602400 6.60% 1.00% 27.5° 27.5° 8-BANDS 0.38m 42.1° 214.8°
WV02 11/14/2015 103001004BC42B00 0.00% 0.00% 17.7° 18.1° 8-BANDS 0.51m 41.4° 1.7°
WV02 11/14/2015 103001004B13AE00 0.00% 0.00% 19.9° 20.0° 8-BANDS 0.52m 41.4° 204.2°
WV02 11/11/2015 103001004CC33A00 0.00% 1.00% 25.0° 26.3° 8-BANDS 0.57m 42.4° 34.7°
WV02 11/11/2015 103001004BD24E00 0.10% 7.00% 13.7° 13.7° 8-BANDS 0.49m 43.0° 59.5°
WV02 11/11/2015 103001004B226F00 0.00% 1.00% 18.5° 18.5° 8-BANDS 0.51m 42.2° 162.4°
WV03 11/9/2015 10400100148BBD00 0.00% 0.00% 29.3° 29.3° 8-BANDS 0.39m 43.5° 54.0°

43

WV03 11/9/2015 1040010014997700 2.00% 3.00% 27.4° 28.2° 8-BANDS 0.38m 44.8° 125.6°
WV03 11/9/2015 10400100146B3300 0.00% 0.00% 27.4° 27.4° 8-BANDS 0.38m 43.4° 149.3°
WV02 11/3/2015 103001004B244B00 0.00% 13.00% 17.9° 18.2° 8-BANDS 0.51m 44.9° 29.6°
WV02 11/3/2015 103001004BC89000 0.00% 3.00% 2.4° 2.4° 8-BANDS 0.46m 44.2° 240.7°
WV02 11/3/2015 103001004B0DD500 0.00% 14.00% 24.8° 25.1° 8-BANDS 0.55m 44.8° 182.5°
WV03 11/2/2015 1040010013B94500 0.00% 0.00% 16.0° 16.0° 8-BANDS 0.33m 46.1° 54.8°
WV02 10/26/2015 103001004ACDC000 0.00% 0.00% 27.5° 28.4° 8-BANDS 0.58m 46.9° 2.5°
WV02 10/26/2015 103001004A55F900 0.00% 0.00% 11.3° 11.4° 8-BANDS 0.48m 46.8° 225.4°
WV02 10/23/2015 10300100496D0000 0.00% 0.00% 9.3° 9.3° 8-BANDS 0.47m 48.0° 120.8°
WV02 10/18/2015 103001004AAA6F00 4.80% 14.00% 7.4° 7.7° 8-BANDS 0.47m 50.8° 35.7°
WV02 10/7/2015 1030010049D05C00 0.00% 0.00% 29.9° 32.7° 8-BANDS 0.64m 52.8° 359.8°
WV02 10/7/2015 10300100495CC200 0.00% 0.00% 21.7° 23.9° 8-BANDS 0.54m 52.7° 350.3°
WV02 10/7/2015 103001004993D500 0.00% 0.00% 6.4° 6.6° 8-BANDS 0.47m 53.4° 329.5°
WV02 10/7/2015 103001004A096F00 0.00% 0.00% 7.6° 9.1° 8-BANDS 0.47m 52.6° 285.3°
WV02 10/7/2015 10300100499B0F00 0.00% 0.00% 20.4° 20.5° 8-BANDS 0.52m 52.5° 223.8°
WV03 8/31/2015 1040010010290A00 0.90% 57.00% 28.4° 28.5° 8-BANDS 0.39m 65.5° 35.6°
WV03 8/31/2015 104001001015A100 0.00% 52.00% 13.0° 13.3° 8-BANDS 0.32m 65.4° 135.0°
WV03 6/3/2015 104001000C326400 0.00% 0.00% 19.0° 19.1° 8-BANDS 0.34m 73.1° 359.5°
WV03 6/3/2015 104001000DBCE300 0.00% 0.00% 9.1° 11.0° 8-BANDS 0.32m 73.0° 217.5°
WV02 5/27/2015 10300100440A4B00 11.00% 19.00% 28.0° 28.4° 8-BANDS 0.58m 73.0° 206.0°
WV02 5/24/2015 1030010042897A00 6.40% 42.00% 19.6° 19.8° 8-BANDS 0.51m 75.2° 51.2°
WV02 5/24/2015 1030010043148A00 14.30% 29.00% 21.8° 22.8° 8-BANDS 0.53m 76.4° 101.6°
WV02 5/16/2015 10300100420DE900 0.00% 2.00% 25.8° 27.8° 8-BANDS 0.58m 72.0° 212.3°
WV02 5/13/2015 1030010040B8D900 6.30% 17.00% 24.5° 24.8° 8-BANDS 0.55m 74.1° 42.6°
WV02 5/8/2015 1030010041167600 1.00% 3.00% 28.6° 31.1° 8-BANDS 0.62m 71.6° 7.0°
WV02 5/8/2015 1030010041057500 10.90% 9.00% 14.0° 15.8° 8-BANDS 0.50m 70.2° 299.2°
WV02 5/8/2015 1030010042868200 12.10% 8.00% 17.0° 17.2° 8-BANDS 0.50m 70.3° 241.2°
WV03 5/8/2015 104001000B3B2300 3.20% 1.00% 25.8° 27.3° 8-BANDS 0.38m 68.7° 319.9°
WV03 5/8/2015 104001000BA62E00 4.50% 2.00% 26.1° 27.2° 8-BANDS 0.38m 68.5° 246.7°
WV02 5/5/2015 1030010041701000 1.80% 3.00% 24.2° 26.1° 8-BANDS 0.56m 71.7° 29.2°
WV02 5/5/2015 1030010042B72900 1.40% 2.00% 14.1° 23.2° 8-BANDS 0.54m 71.6° 46.5°
WV02 5/5/2015 1030010041766 3.60% 5.00% 8.0° 8.7° 8-BANDS 0.47m 71.5° 55.0°
WV02 5/5/2015 10300100420E0C00 14.20% 3.00% 5.7° 11.3° 8-BANDS 0.48m 71.3° 171.0°
WV03 5/3/2015 104001000A499E00 5.90% 6.00% 26.6° 28.2° 8-BANDS 0.38m 72.1° 71.1°
WV03 5/3/2015 104001000B55EA00 3.80% 4.00% 24.7° 27.3° 8-BANDS 0.38m 72.0° 124.7°
WV03 5/3/2015 104001000A702C00 3.70% 7.00% 27.6° 29.4° 8-BANDS 0.39m 72.1° 132.8°
WV03 5/3/2015 104001000B64B100 0.00% 42.00% 26.7° 26.7° 8-BANDS 0.37m 70.7° 157.7°
WV02 5/3/2015 10300100400CE500 0.80% 8.00% 28.4° 28.9° 8-BANDS 0.59m 67.8° 288.8°
WV02 5/2/2015 10300100405ACD00 6.40% 17.00% 25.7° 26.9° 8-BANDS 0.57m 73.5° 91.1°
WV02 5/2/2015 103001004164D900 8.40% 10.00% 24.1° 26.6° 8-BANDS 0.57m 73.3° 121.7°
WV02 5/2/2015 10300100428AFA00 4.70% 21.00% 26.2° 27.4° 8-BANDS 0.57m 72.1° 147.3°
WV03 5/2/2015 104001000B473D00 11.80% 9.00% 23.1° 24.2° 8-BANDS 0.36m 68.7° 353.7°
WV03 5/2/2015 104001000B14F100 1.60% 4.00% 7.7° 20.5° 8-BANDS 0.35m 68.7° 346.7°
WV03 5/2/2015 104001000A34AD00 0.00% 2.00% 9.8° 12.1° 8-BANDS 0.32m 68.7° 235.0°
WV02 4/27/2015 1030010041893300 0.00% 8.00% 8.9° 11.5° 8-BANDS 0.48m 68.6° 282.8°
WV03 4/19/2015 104001000A0D5B00 0.60% 14.00% 18.7° 18.7° 8-BANDS 0.34m 66.5° 11.2°
WV02 4/8/2015 103001003FAA5F00 0.00% 0.00% 13.9° 22.7° 8-BANDS 0.54m 63.7° 194.9°
WV02 4/8/2015 1030010040AF2000 2.00% 10.00% 23.0° 25.3° 8-BANDS 0.55m 63.8° 191.8°
WV03 3/25/2015 1040010009750600 0.00% 0.00% 27.1° 34.2° 8-BANDS 0.44m 58.3° 34.2°
WV03 3/25/2015 1040010009CDB800 0.00% 0.00% 19.4° 20.5° 8-BANDS 0.35m 58.1° 174.3°
WV02 3/12/2015 103001003F4BC200 0.00% 1.00% 20.6° 24.5° 8-BANDS 0.55m 55.3° 77.0°
WV02 3/12/2015 103001003F0E8A00 0.00% 2.00% 17.2° 17.4° 8-BANDS 0.50m 55.3° 85.1°
WV02 3/12/2015 103001003FC96600 13.30% 6.00% 15.7° 17.4° 8-BANDS 0.50m 55.1° 129.4°
WV02 3/12/2015 103001003FB2CA00 0.00% 0.00% 17.1° 18.3° 8-BANDS 0.51m 55.1° 140.2°
WV02 3/12/2015 1030010040088D00 0.00% 0.00% 23.2° 26.1° 8-BANDS 0.56m 55.0° 163.0°
WV02 3/12/2015 103001003F88F500 1.40% 1.00% 27.1° 28.0° 8-BANDS 0.58m 54.9° 168.3°
WV03 3/7/2015 1040010008873500 7.50% 6.00% 16.2° 16.4° 8-BANDS 0.33m 51.3° 46.5°
WV03 3/5/2015 1040010008276900 2.00% 8.00% 24.7° 26.1° 8-BANDS 0.37m 48.7° 293.2°
WV03 2/22/2015 1040010008793A00 0.00% 0.00% 28.2° 28.2° 8-BANDS 0.39m 46.4° 33.4°
WV03 2/22/2015 1040010008B10800 0.00% 0.00% 16.5° 16.5° 8-BANDS 0.33m 46.3° 158.2°
WV02 2/10/2015 103001003D381400 0.30% 5.00% 23.8° 24.0° 8-BANDS 0.54m 44.1° 34.8°

44

WV02 2/10/2015 103001003E9DBA00 0.30% 4.00% 11.0° 11.5° 8-BANDS 0.48m 44.1° 139.1°
WV03 2/3/2015 1040010007160000 0.00% 0.00% 23.7° 24.9° 8-BANDS 0.37m 41.9° 66.8°
WV03 2/3/2015 104001000742C800 0.00% 0.00% 18.6° 23.1° 8-BANDS 0.36m 41.7° 134.3°
WV03 2/3/2015 1040010007A71E00 0.00% 0.00% 23.8° 24.9° 8-BANDS 0.37m 41.7° 148.0°
WV02 1/22/2015 103001003C24BE00 0.00% 0.00% 23.2° 23.2° 8-BANDS 0.54m 39.8° 70.2°
WV02 1/22/2015 103001003C132600 0.00% 0.00% 18.6° 18.9° 8-BANDS 0.51m 39.7° 97.4°
WV02 1/22/2015 103001003D934D00 0.00% 0.00% 27.7° 27.9° 8-BANDS 0.58m 39.7° 150.4°
WV03 1/22/2015 1040010006841 0.00% 0.00% 23.0° 23.0° 8-BANDS 0.36m 38.2° 72.4°
WV03 1/21/2015 10400100062AFA00 0.00% 2.00% 19.9° 21.1° 8-BANDS 0.35m 36.4° 329.8°
WV03 1/21/2015 1040010006520200 0.00% 0.00% 29.1° 29.2° 8-BANDS 0.39m 36.2° 222.7°
WV03 1/9/2015 1040010006833 0.00% 0.00% 26.5° 26.5° 8-BANDS 0.37m 35.0° 199.8°
WV03 12/28/2014 1040010006B38F00 0.00% 0.00% 20.9° 21.0° 8-BANDS 0.35m 35.3° 68.3°
WV03 12/28/2014 1040010005CC8200 0.00% 0.00% 21.9° 23.8° 8-BANDS 0.36m 35.2° 150.2°
WV03 12/28/2014 1040010006906600 0.00% 0.00% 27.4° 27.4° 8-BANDS 0.38m 35.3° 158.0°
WV03 12/21/2014 1040010005A19700 0.00% 0.00% 14.4° 15.1° 8-BANDS 0.33m 34.3° 323.2°
WV02 12/7/2014 103001003B477100 0.00% 0.00% 15.8° 16.2° 8-BANDS 0.50m 37.5° 15.4°
WV02 12/7/2014 1030010039010100 0.00% 0.00% 12.2° 12.2° 8-BANDS 0.48m 37.5° 191.3°
WV02 12/2/2014 103001003A3B9B00 0.00% 16.00% 18.2° 18.6° 8-BANDS 0.51m 38.5° 333.5°
WV02 12/2/2014 103001003BC03700 0.00% 35.00% 16.6° 16.7° 8-BANDS 0.50m 38.4° 239.1°
WV03 11/19/2014 10400100046BFB00 0.00% 0.00% 22.3° 22.3° 8-BANDS 0.35m 39.5° 198.2°
WV03 11/13/2014 10400100044AC800 0.00% 1.00% 27.7° 27.9° 8-BANDS 0.38m 41.8° 181.7°
WV02 11/2/2014 103001003939CE00 0.00% 30.00% 25.0° 26.0° 8-BANDS 0.56m 47.0° 44.0°
WV03 10/25/2014 104001000396F800 13.50% 7.00% 19.9° 21.2° 8-BANDS 0.35m 47.8° 43.8°
WV03 10/25/2014 10400100037BF300 1.80% 8.00% 15.8° 17.0° 8-BANDS 0.33m 47.7° 136.8°
WV03 10/25/2014 104001000323BC00 5.90% 5.00% 17.1° 18.3° 8-BANDS 0.34m 47.8° 153.2°
WV02 10/22/2014 103001003882DD00 2.70% 9.00% 26.6° 27.9° 8-BANDS 0.58m 49.2° 4.9°
WV02 10/22/2014 103001003952FB00 1.90% 9.00% 13.1° 13.2° 8-BANDS 0.48m 49.1° 215.8°
WV02 10/19/2014 1030010036626800 5.00% 15.00% 26.5° 27.1° 8-BANDS 0.57m 50.9° 43.6°
WV02 10/19/2014 1030010039B4B200 3.00% 12.00% 14.7° 15.3° 8-BANDS 0.49m 50.8° 123.3°
WV03 10/18/2014 104001000341CA00 6.90% 18.00% 25.6° 25.6° 8-BANDS 0.37m 48.7° 228.7°
WV02 10/17/2014 1030010039D63900 5.40% 8.00% 19.3° 26.4° 8-BANDS 0.57m 51.6° 359.8°
WV03 10/12/2014 1040010002746600 1.30% 3.00% 17.8° 18.0° 8-BANDS 0.33m 51.4° 183.1°
WV02 10/8/2014 1030010037459900 2.50% 24.00% 27.9° 27.9° 8-BANDS 0.58m 55.0° 68.4°
WV02 10/8/2014 1030010038B10800 0.50% 21.00% 25.9° 26.3° 8-BANDS 0.56m 55.0° 130.5°
WV02 10/6/2014 1030010039B26E00 0.60% 13.00% 17.1° 24.2° 8-BANDS 0.55m 55.8° 48.1°
WV02 10/6/2014 1030010038AB9A00 2.90% 14.00% 9.9° 10.4° 8-BANDS 0.48m 55.8° 54.7°
WV03 10/6/2014 10400100022B5100 2.20% 23.00% 26.6° 26.6° 8-BANDS 0.38m 53.2° 33.3°
WV03 10/6/2014 1040010002389700 4.00% 20.00% 18.9° 19.0° 8-BANDS 0.34m 53.0° 162.2°
WV02 10/1/2014 10300100385C3800 0.40% 21.00% 26.7° 28.2° 8-BANDS 0.58m 56.1° 327.4°
WV02 10/1/2014 1030010038849000 2.00% 30.00% 23.1° 23.2° 8-BANDS 0.54m 56.0° 256.9°
WV02 5/29/2014 1030010032968300 4.40% 33.00% 14.5° 22.1° 8-BANDS 0.53m 75.5° 351.6°
WV02 5/29/2014 103001103155D500 0.00% 27.00% 8.2° 9.7° 8-BANDS 0.47m 75.6° 323.7°
WV02 3/11/2014 103001002E546F00 12.00% 10.00% 9.4° 10.7° 8-BANDS 0.48m 55.5° 149.3°
WV02 1/16/2014 103001002C973200 0.00% 0.00% 22.8° 23.3° 8-BANDS 0.54m 37.4° 288.2°
WV02 1/16/2014 103001002BA05E00 0.00% 0.00% 25.6° 25.6° 8-BANDS 0.56m 37.4° 254.8°
WV02 1/10/2014 103001002C1DBC00 0.00% 0.00% 10.7° 10.9° 8-BANDS 0.48m 38.2° 107.6°
WV02 1/2/2014 103001002A727F00 0.00% 0.00% 26.1° 26.1° 8-BANDS 0.56m 37.7° 13.5°
WV02 1/2/2014 103001002B2D6300 0.00% 0.00% 21.8° 22.2° 8-BANDS 0.53m 37.2° 5.8°
WV02 1/2/2014 103001002C67BC00 0.00% 0.00% 17.0° 17.3° 8-BANDS 0.50m 37.1° 204.1°
WV02 12/3/2013 10300100298CE300 0.00% 0.00% 8.9° 9.6° 8-BANDS 0.47m 39.4° 139.6°
WV02 11/25/2013 103001002A5FE600 0.00% 6.00% 26.4° 28.2° 8-BANDS 0.58m 39.7° 351.1°
WV02 11/25/2013 1030010029066600 0.00% 6.00% 15.8° 16.1° 8-BANDS 0.50m 39.6° 241.4°
WV02 11/22/2013 10300100297A1200 14.60% 6.00% 14.8° 15.0° 8-BANDS 0.49m 40.9° 62.9°
WV02 11/17/2013 1030010029470C00 0.00% 0.00% 25.5° 28.0° 8-BANDS 0.58m 41.3° 335.0°
WV02 11/17/2013 1030010029AF3000 0.00% 0.00% 18.6° 20.7° 8-BANDS 0.52m 41.3° 317.3°
WV02 11/17/2013 103001002A3F9200 0.00% 0.00% 20.6° 20.6° 8-BANDS 0.52m 41.2° 254.2°
WV02 11/17/2013 1030010029BAF700 0.00% 0.00% 27.2° 27.6° 8-BANDS 0.58m 41.3° 232.3°
WV02 11/3/2013 1030010029D63900 9.60% 1.00% 14.8° 15.5° 8-BANDS 0.49m 46.7° 127.9°
WV02 10/29/2013 103001002806F900 2.20% 3.00% 17.1° 17.6° 8-BANDS 0.50m 47.7° 311.5°
WV02 10/21/2013 1030010028273400 2.60% 22.00% 29.1° 29.5° 8-BANDS 0.60m 49.9° 318.9°
WV02 6/26/2013 10300100240D3C00 4.20% 70.00% 10.6° 10.7° 8-BANDS 0.48m 75.8° 308.9°

45

WV02 4/19/2013 1030010020324A00 0.00% 2.00% 16.0° 16.7° 8-BANDS 0.50m 70.7° 50.9°
WV02 3/15/2013 103001001F1DD400 2.50% 1.00% 26.6° 28.7° 8-BANDS 0.59m 56.2° 338.7°
WV02 3/15/2013 1030010021894300 6.10% 2.00% 18.4° 19.5° 8-BANDS 0.51m 56.3° 317.5°
WV02 3/15/2013 103001002051AF00 4.90% 1.00% 14.6° 17.4° 8-BANDS 0.50m 56.3° 312.0°
WV02 3/15/2013 103001002012DB00 0.80% 3.00% 15.5° 15.8° 8-BANDS 0.49m 56.4° 258.3°
WV02 3/15/2013 10300100207BC300 1.20% 2.00% 20.3° 25.3° 8-BANDS 0.56m 56.4° 247.8°
WV02 3/15/2013 1030010021D46400 5.50% 7.00% 26.2° 27.0° 8-BANDS 0.57m 56.4° 221.2°
WV02 3/9/2013 10300100209D7A00 5.80% 3.00% 10.7° 10.8° 8-BANDS 0.48m 54.7° 169.8°
WV02 2/13/2013 103001001F006500 0.00% 1.00% 28.1° 28.6° 8-BANDS 0.59m 46.1° 357.9°
WV02 2/13/2013 1030010020049700 0.90% 1.00% 25.9° 26.4° 8-BANDS 0.57m 46.0° 352.7°
WV02 2/13/2013 103001001FCB4200 11.40% 2.00% 13.6° 14.3° 8-BANDS 0.49m 45.8° 318.9°
WV02 2/13/2013 103001001F4DFD00 0.00% 0.00% 14.4° 14.4° 8-BANDS 0.49m 45.8° 253.6°
WV02 2/13/2013 103001001FBDE200 0.00% 0.00% 17.6° 25.9° 8-BANDS 0.56m 45.7° 245.7°
WV02 2/13/2013 1030010020712 0.00% 2.00% 26.9° 28.5° 8-BANDS 0.58m 45.6° 224.0°
WV02 2/10/2013 103001001EB75F00 1.60% 1.00% 21.5° 22.8° 8-BANDS 0.54m 46.0° 68.6°
WV02 2/10/2013 103001001FB97A00 0.00% 0.00% 6.1° 6.2° 8-BANDS 0.47m 45.4° 74.7°
WV02 1/25/2013 103001001E1E4800 0.00% 38.00% 14.3° 14.3° 8-BANDS 0.49m 40.7° 275.1°
WV02 1/25/2013 103001001E807B00 0.00% 0.00% 10.3° 10.4° 8-BANDS 0.47m 40.3° 245.7°
WV02 12/23/2012 103001001E374900 0.00% 44.00% 7.4° 9.3° 8-BANDS 0.47m 38.1° 111.4°
WV02 12/23/2012 103001001E932A00 0.00% 0.00% 7.0° 7.2° 8-BANDS 0.47m 37.0° 109.6°
WV02 11/23/2012 103001001ED05100 0.00% 0.00% 28.5° 29.4° 8-BANDS 0.60m 41.7° 134.1°
WV02 10/30/2012 103001001CC6FC00 0.00% 0.00% 0.6° 1.1° 8-BANDS 0.46m 47.8° 118.3°
WV02 8/18/2012 103001001A1E1700 9.20% 47.00% 13.3° 14.4° 8-BANDS 0.49m 69.0° 292.2°
WV02 8/7/2012 103001001B423400 0.00% 99.00% 28.1° 29.3° 8-BANDS 0.59m 71.8° 208.4°
WV02 6/11/2012 1030010019C08F00 0.00% 61.00% 2.8° 3.7° 8-BANDS 0.46m 77.2° 296.5°
WV02 6/11/2012 10300100189CF700 0.00% 25.00% 9.5° 9.5° 8-BANDS 0.47m 76.7° 241.2°
WV02 5/23/2012 103001001970D600 0.00% 17.00% 3.1° 4.0° 8-BANDS 0.46m 76.5° 295.9°
WV02 5/20/2012 1030010018655700 4.30% 13.00% 12.4° 12.5° 8-BANDS 0.48m 77.5° 80.8°
WV02 5/15/2012 1030010018537600 0.00% 0.00% 21.1° 21.1° 8-BANDS 0.52m 73.7° 284.8°
WV02 5/4/2012 10300100188ED700 0.50% 21.00% 13.6° 14.0° 8-BANDS 0.49m 73.3° 352.4°
WV02 5/4/2012 1030010018C46900 0.00% 17.00% 8.1° 9.2° 8-BANDS 0.47m 73.4° 353.8°
WV02 5/4/2012 1030010019A06E00 0.00% 38.00% 1.7° 3.0° 8-BANDS 0.46m 73.3° 301.9°
WV02 5/4/2012 10300100182D0700 9.60% 26.00% 13.6° 14.2° 8-BANDS 0.49m 73.7° 196.3°
WV02 4/29/2012 10300100182B7600 0.00% 39.00% 21.3° 23.4° 8-BANDS 0.54m 71.9° 340.6°
WV02 4/23/2012 1030010013248200 0.00% 3.00% 13.7° 13.7° 8-BANDS 0.49m 71.9° 103.8°
WV02 4/7/2012 1030010012347100 0.00% 1.00% 7.6° 9.3° 8-BANDS 0.47m 64.9° 294.3°
WV02 3/30/2012 1030010012AF8100 0.00% 1.00% 20.4° 20.4° 8-BANDS 0.52m 61.4° 285.0°
WV02 3/27/2012 1030010012738D00 14.70% 9.00% 24.1° 24.9° 8-BANDS 0.55m 61.5° 7.7°
WV02 3/11/2012 1030010012CAB700 0.00% 0.00% 29.2° 29.4° 8-BANDS 0.60m 53.5° 299.6°
WV02 12/28/2011 1030010010146A00 0.00% 0.00% 9.3° 10.1° 8-BANDS 0.48m 37.1° 110.9°
WV02 12/28/2011 103001000F8FE300 0.00% 0.00% 15.2° 16.7° 8-BANDS 0.50m 37.1° 163.2°
WV02 12/20/2011 1030010010497A00 1.00% 2.00% 7.9° 8.8° 8-BANDS 0.47m 37.9° 108.3°
WV02 12/9/2011 103001000F821A00 2.00% 14.00% 20.1° 21.5° 8-BANDS 0.53m 39.3° 108.8°
WV02 12/9/2011 103001000F1BA800 0.50% 20.00% 24.0° 26.2° 8-BANDS 0.56m 39.3° 135.4°
WV02 10/13/2011 103001000E930A00 10.90% 7.00% 5.4° 5.6° 8-BANDS 0.47m 52.8° 113.3°
WV02 10/13/2011 103001000E270700 0.90% 10.00% 14.6° 15.0° 8-BANDS 0.49m 53.0° 163.0°
WV02 10/10/2011 103001000DD4BE00 0.00% 3.00% 23.5° 23.5° 8-BANDS 0.54m 54.5° 109.2°
WV02 10/5/2011 103001000E2DE300 1.70% 10.00% 14.4° 14.7° 8-BANDS 0.49m 55.4° 207.0°
WV02 10/5/2011 103001000E3D5A00 3.00% 31.00% 23.3° 23.6° 8-BANDS 0.54m 55.3° 204.1°
WV02 7/10/2011 103001000CC3BB00 14.10% 86.00% 27.2° 29.4° 8-BANDS 0.60m 73.3° 226.7°
WV02 6/15/2011 103001000B3D5900 0.00% 47.00% 18.1° 19.9° 8-BANDS 0.52m 77.6° 169.7°
WV02 4/27/2011 103001000B8EF400 0.00% 7.00% 26.8° 29.6° 8-BANDS 0.60m 70.9° 198.2°
WV02 2/23/2011 1030010009100B00 0.00% 0.00% 15.4° 19.6° 8-BANDS 0.51m 48.7° 196.3°
WV02 2/9/2011 103001000992DF00 0.00% 4.00% 18.7° 19.5° 8-BANDS 0.51m 45.1° 124.7°
WV02 12/8/2010 1030010008914A00 0.00% 0.00% 8.4° 8.6° 8-BANDS 0.47m 38.6° 105.2°
WV02 6/7/2010 1030010005047 2.40% 4.00% 14.7° 15.5° 8-BANDS 0.50m 74.2° 348.0°
WV02 6/7/2010 1030010005302000 0.40% 7.00% 8.8° 9.4° 8-BANDS 0.48m 74.3° 337.2°
WV02 6/7/2010 10300100053BC200 14.40% 8.00% 9.3° 9.7° 8-BANDS 0.47m 74.2° 229.4°
WV02 6/7/2010 1030010005615F00 0.40% 4.00% 17.3° 18.0° 8-BANDS 0.51m 74.0° 215.1°
WV02 6/7/2010 1030010005962 0.00% 2.00% 27.6° 31.4° 8-BANDS 0.62m 73.8° 208.6°
WV02 6/4/2010 10300100054A6700 0.00% 21.00% 16.6° 17.9° 8-BANDS 0.50m 75.9° 160.7°

46

WV02 5/16/2010 1030010005859200 12.70% 15.00% 17.3° 19.8° 8-BANDS 0.52m 73.5° 198.1°
WV02 4/24/2010 1030010004939 0.00% 0.00% 17.2° 19.9° 8-BANDS 0.52m 69.1° 195.5°
WV02 1/29/2010 103001000361E600 0.00% 14.00% 22.0° 24.9° 8-BANDS 0.55m 40.5° 195.4°
851
2.5 REFERENCES 852
Attal, M., and J. Lavé, 2006, Changes of bedload characteristics along the Marsyandi River 853
(central Nepal): Implications for understanding hillslope sediment supply, sediment load 854
evolution along fluvial networks, and denudation in active orogenic belts, in Special 855
Paper 398: Tectonics, Climate, and Landscape Evolution: Geological Society of America, 856
p. 143–171, doi:10.1130/2006.2398(09). 857
858
Avouac, J.-P., L. Meng, S. Wei, T. Wang, and J.-P. Ampuero, 2015, Lower edge of locked Main 859
Himalayan Thrust unzipped by the 2015 Gorkha earthquake: Nature Geoscience, v. 8, no. 860
9, p. 708–711, doi:10.1038/ngeo2518. 861
862
Baer, P., C. Huggel, B. W. McArdell, and F. Frank, 2017, Changing debris flow activity after 863
sudden sediment input: a case study from the Swiss Alps: Geology Today, v. 33, no. 6, p. 864
216–223, doi:10.1111/gto.12211. 865
866
Benda, L. E., and T. W. Cundy, 1990, Predicting deposition of debris flows in mountain 867
channels: Canadian Geotechnical Journal, v. 27, no. 4, p. 409–417, doi:10.1139/t90-057. 868
869
Bertrand, M., F. Liébault, and H. Piégay, 2013, Debris-flow susceptibility of upland catchments: 870
Natural Hazards, v. 67, no. 2, p. 497–511, doi:10.1007/s11069-013-0575-4. 871
872
Blahut, J., C. J. van Westen, and S. Sterlacchini, 2010, Analysis of landslide inventories for 873
accurate prediction of debris-flow source areas: Geomorphology, v. 119, no. 1–2, p. 36– 874
51, doi:10.1016/j.geomorph.2010.02.017. 875
876
Brasington, J., and K. Richards, 2000, Turbidity and suspended sediment dynamics in small 877
catchments in the Nepal Middle Hills: Hydrological Processes, v. 14, no. 14, p. 2559– 878
2574, doi:10.1002/1099-1085(20001015)14:14<2559::AID-HYP114>3.0.CO;2-E. 879
880
Brayshaw, D., and M. A. Hassan, 2009, Debris flow initiation and sediment recharge in gullies: 881
Geomorphology, v. 109, no. 3–4, p. 122–131, doi:10.1016/j.geomorph.2009.02.021. 882
883
Caine N., 1980, The rainfall intensity-duration control of shallow landslides and debris flows: 884
Geografiska Annaler v. 62, no. 1, p. 23-27. 885
886
Cannon, S. H., E. R. Bigio, and E. Mine, 2001, A process for fire-related debris flow initiation, 887
Cerro Grande fire, New Mexico: Hydrological Processes, v. 15, no. 15, p. 3011–3023, 888
doi:10.1002/hyp.388. 889
890

47

Cannon, S. H., R. M. Kirkham, and M. Parise, 2001, Wildfire-related debris-flow initiation 891
processes, Storm King Mountain, Colorado: Geomorphology, v. 39, no. 3–4, p. 171–188, 892
doi:10.1016/S0169-555X(00)00108-2. 893
894
Cao, Z., G. Pender, and J. Meng, 2006, Explicit Formulation of the Shields Diagram for Incipient 895
Motion of Sediment: Journal of Hydraulic Engineering, v. 132, no. 10, p. 1097–1099, 896
doi:10.1061/(ASCE)0733-9429(2006)132:10(1097). 897
Costa, J. E., 1984, Physical Geomorphology of Debris Flows, in J. E. Costa, and P. J. Fleisher, 898
eds., Developments and Applications of Geomorphology: Berlin, Heidelberg, Springer 899
Berlin Heidelberg, p. 268–317, doi:10.1007/978-3-642-69759-3_9. 900
901
Coussot, P., and M. Meunier, 1996, Recognition, classification and mechanical description of 902
debris flows: Earth-Science Reviews, v. 40, no. 3–4, p. 209–227, doi:10.1016/0012- 903
8252(95)00065-8. 904
905
Dadson, S. J. et al., 2004, Earthquake-triggered increase in sediment delivery from an active 906
mountain belt: Geology, v. 32, no. 8, p. 733, doi:10.1130/G20639.1. 907
908
Dahal, R. K., and S. Hasegawa, 2008, Representative rainfall thresholds for landslides in the 909
Nepal Himalaya: Geomorphology, v. 100, no. 3–4, p. 429–443, 910
doi:10.1016/j.geomorph.2008.01.014. 911
912
Deng, Y., J. P. Wilson, and B. O. Bauer, 2007, DEM resolution dependencies of terrain attributes 913
across a landscape: International Journal of Geographical Information Science, v. 21, no. 914
2, p. 187–213, doi:10.1080/13658810600894364. 915
916
Dowling, C. A., and P. M. Santi, 2014, Debris flows and their toll on human life: a global 917
analysis of debris-flow fatalities from 1950 to 2011: Natural Hazards, v. 71, no. 1, p. 918
203–227, doi:10.1007/s11069-013-0907-4. 919
920
Fan, L., P. Lehmann, B. McArdell, and D. Or, 2017, Linking rainfall-induced landslides with 921
debris flows runout patterns towards catchment scale hazard assessment: 922
Geomorphology, v. 280, p. 1–15, doi:10.1016/j.geomorph.2016.10.007. 923
924
Gabet, E. J., and S. M. Mudd, 2006, The mobilization of debris flows from shallow landslides: 925
Geomorphology, v. 74, no. 1–4, p. 207–218, doi:10.1016/j.geomorph.2005.08.013. 926
927
Gallen, S. F., M. K. Clark, J. W. Godt, K. Roback, and N. A. Niemi, 2017, Application and 928
evaluation of a rapid response earthquake-triggered landslide model to the 25 April 2015 929
Mw 7.8 Gorkha earthquake, Nepal: Tectonophysics, v. 714–715, p. 173–187, 930
doi:10.1016/j.tecto.2016.10.031. 931
932
Guo, X., P. Cui, Y. Li, L. Ma, Y. Ge, and W. B. Mahoney, 2016, Intensity–duration threshold of 933
rainfall-triggered debris flows in the Wenchuan Earthquake affected area, China: 934
Geomorphology, v. 253, p. 208–216, doi:10.1016/j.geomorph.2015.10.009. 935
936

48

Guo, X., P. Cui, Y. Li, Q. Zou, and Y. Kong, 2016, The formation and development of debris 937
flows in large watersheds after the 2008 Wenchuan Earthquake: Landslides, v. 13, no. 1, 938
p. 25–37, doi:10.1007/s10346-014-0541-6. 939
940
Guo, X., P. Cui, L. Marchi, and Y. Ge, 2017, Characteristics of rainfall responsible for debris 941
flows in Wenchuan Earthquake area: Environmental Earth Sciences, v. 76, no. 17, 942
doi:10.1007/s12665-017-6940-y. 943
944
Hou, A. Y., R. K. Kakar, S. Neeck, A. A. Azarbarzin, C. D. Kummerow, M. Kojima, R. Oki, K. 945
Nakamura, and T. Iguchi, 2014, The Global Precipitation Measurement Mission: Bulletin 946
of the American Meteorological Society, v. 95, no. 5, p. 701–722, doi:10.1175/BAMS-D- 947
13-00164.1. 948
949
Hu, W., Q. Xu, T. W. J. van Asch, X. Zhu, and Q. Q. Xu, 2014, Flume tests to study the 950
initiation of huge debris flows after the Wenchuan earthquake in S-W China: Engineering 951
Geology, v. 182, p. 121–129, doi:10.1016/j.enggeo.2014.04.006. 952
953
Huang, R., and W. Li, 2014, Post-earthquake landsliding and long-term impacts in the 954
Wenchuan earthquake area, China: Engineering Geology, v. 182, p. 111–120, 955
doi:10.1016/j.enggeo.2014.07.008. 956
957
Hungr, O., 1995, A model for the runout analysis of rapid flow slides, debris flows, and 958
avalanches: Canadian Geotechnical Journal, v. 32, no. 4, p. 610–623, doi:10.1139/t95- 959
063. 960
961
Hungr, O., S. McDougall, M. Wise, and M. Cullen, 2008, Magnitude–frequency relationships of 962
debris flows and debris avalanches in relation to slope relief: Geomorphology, v. 96, no. 963
3–4, p. 355–365, doi:10.1016/j.geomorph.2007.03.020. 964
965
Iverson, R. M., 1997, The physics of debris flows: Reviews of Geophysics, v. 35, no. 3, p. 245– 966
296, doi:10.1029/97RG00426. 967
968
Iverson, R. M., M. E. Reid, and R. G. LaHusen, 1997, Debris-flow mobilization from landslides: 969
Annual Review of Earth and Planetary Sciences, v. 25, no. 1, p. 85–138, 970
doi:10.1146/annurev.earth.25.1.85. 971
972
Iverson, R. M., 2014, Debris flows: behaviour and hazard assessment: Geology Today, v. 30, no. 973
1, p. 15–20, doi:10.1111/gto.12037. 974
975
Jackson, L. E., R. A. Kostaschuk, and G. M. Macdonald, 1987, Identification of debris flow 976
hazard on alluvial fans in the Canadian Rocky Mountains: Debris Flows/Avalanches: 977
Process, Recognition, and Mitigation: Reviews in Engineering Geology, v.7, p. 115–124. 978
979
Jakob, M., and O. Hungr, 2005, Debris-flow hazards and related phenomena: Berlin; New York, 980
Springer, 739 p. 981

49

Jibson, R. W., and D. K. Keefer, 1993, Analysis of the seismic origin of landslides: Examples 982
from the New Madrid seismic zone: Geological Society of America Bulletin, v. 105, p. 983
521-536. 984
985
Keefer, D. K., 1984., Landslides caused by earthquakes: Geological Society of America Bulletin 986
v. 95, p. 406-421. 987
988
Li, G., A. J. West, A. L. Densmore, Z. Jin, R. N. Parker, and R. G. Hilton, 2014, Seismic 989
mountain building: Landslides associated with the 2008 Wenchuan earthquake in the 990
context of a generalized model for earthquake volume balance: Geochemistry, 991
Geophysics, Geosystems, v. 15, no. 4, p. 833–844, doi:10.1002/2013GC005067. 992
993
Marc, O., N. Hovius, P. Meunier, T. Uchida, and S. Hayashi, 2015, Transient changes of 994
landslide rates after earthquakes: Geology, v. 43, no. 10, p. 883–886, 995
doi:10.1130/G36961.1. 996
997
Martha, T. R., P. Roy, R. Mazumdar, K. B. Govindharaj, and K. V. Kumar, 2017, Spatial 998
characteristics of landslides triggered by the 2015 Mw 7.8 (Gorkha) and Mw 7.3 999
(Dolakha) earthquakes in Nepal: Landslides, v. 14, no. 2, p. 697–704, 1000
doi:10.1007/s10346-016-0763-x. 1001
1002
Montgomery, D. R., and W. E. Dietrich, 1994, A physically based model for the topographic 1003
control on shallow landsliding: Water Resources Research, v. 30, no. 4, p. 1153–1171, 1004
doi:10.1029/93WR02979. 1005
1006
Ni, H., C. Tang, W. Zheng, R. Xu, K. Tian, and W. Xu, 2014, An overview of formation 1007
mechanism and disaster characteristics of post-seismic debris flows triggered by 1008
subsequent rainstorms in Wenchuan Earthquake extremely stricken areas: Acta 1009
Geologica Sinica - English Edition, v. 88, no. 4, p. 1310–1328, doi:10.1111/1755- 1010
6724.12290. 1011
1012
Noh, M.-J., and I. M. Howat, 2015, Automated stereo-photogrammetric DEM generation at high 1013
latitudes: Surface Extraction with TIN-based Search-space Minimization (SETSM) 1014
validation and demonstration over glaciated regions: GIScience & Remote Sensing, v. 52, 1015
no. 2, p. 198–217, doi:10.1080/15481603.2015.1008621. 1016
1017
Petley, D., 2012, Global patterns of loss of life from landslides: Geology, v. 40, no. 10, p. 927– 1018
930, doi:10.1130/G33217.1. 1019
1020
Prancevic, J. P., M. P. Lamb, and B. M. Fuller, 2014, Incipient sediment motion across the river 1021
to debris-flow transition: Geology, v. 42, no. 3, p. 191–194, doi:10.1130/G34927.1. 1022
1023
Prakash, S., A. K. Mitra, A. AghaKouchak, Z. Liu, H. Norouzi, and D. S. Pai, 2018, A 1024
preliminary assessment of GPM-based multi-satellite precipitation estimates over a 1025
monsoon dominated region: Journal of Hydrology, v. 556, p. 865–876, 1026
doi:10.1016/j.jhydrol.2016.01.029. 1027

50

Prochaska, A. B., P. M. Santi, J. D. Higgins, and S. H. Cannon, 2008, Debris-flow runout 1028
predictions based on the average channel slope (ACS): Engineering Geology, v. 98, no. 1029
1–2, p. 29–40, doi:10.1016/j.enggeo.2008.01.011. 1030
1031
Roback, K., M. K. Clark, A. J. West, D. Zekkos, G. Li, S. F. Gallen, D. Chamlagain, and J. W. 1032
Godt, 2018, The size, distribution, and mobility of landslides caused by the 2015 M w 7.8 1033
Gorkha earthquake, Nepal: Geomorphology, v. 301, p. 121–138, 1034
doi:10.1016/j.geomorph.2017.01.030. 1035
1036
Robinson, T. R., N. J. Rosser, A. L. Densmore, J. G. Williams, M. E. Kincey, J. Benjamin, and 1037
H. J. A. Bell, 2017, Rapid post-earthquake modelling of coseismic landslide intensity and 1038
distribution for emergency response decision support: Natural Hazards and Earth System 1039
Sciences, v. 17, no. 9, p. 1521–1540, doi:10.5194/nhess-17-1521-2017. 1040
1041
Schwanghart, W., and D. Scherler, 2014, Short Communication: TopoToolbox 2 – MATLAB- 1042
based software for topographic analysis and modeling in Earth surface sciences: Earth 1043
Surface Dynamics, v. 2, no. 1, p. 1–7, doi:10.5194/esurf-2-1-2014. 1044
1045
Shields, A., 1936, Application of the theory of similarity and turbulence research to the bedload 1046
movement; English translation by W.P. Ott and J.C. van Uchelen, available as 1047
Hydrodynamics Laboratory Publication, Hydrodynamics Lab, California Institute of 1048
Technology, Pasadena, California v. 167, p. 1–36. 1049
1050
Shieh, C. L., Y. S. Chen, Y. J. Tsai, and J. H. Wu, 2009, Variability in rainfall threshold for 1051
debris flow after the Chi-Chi earthquake in central Taiwan, China: International Journal 1052
of Sediment Research, v. 24, no. 2, p. 177–188, doi:10.1016/S1001-6279(09)60025-1. 1053
1054
Staley, D. M., J. W. Kean, S. H. Cannon, K. M. Schmidt, and J. L. Laber, 2013, Objective 1055
definition of rainfall intensity–duration thresholds for the initiation of post-fire debris 1056
flows in southern California: Landslides, v. 10, no. 5, p. 547–562, doi:10.1007/s10346- 1057
012-0341-9. 1058
1059
Stock, J., and W. E. Dietrich, 2003, Valley incision by debris flows: Evidence of a topographic 1060
signature: Water Resources Research, v. 39, no. 4, doi:10.1029/2001WR001057. 1061
1062
Stock, J. D., and W. E. Dietrich, 2006, Erosion of steepland valleys by debris flows: Geological 1063
Society of America Bulletin, v. 118, no. 9–10, p. 1125–1148. 1064
1065
Takahashi, T., 1978, Mechanical characteristics of debris flow: Journal of the Hydraulics 1066
Division, v. 104, no. H48, p. 1153-1169. 1067
1068
Tang C., and J. T. Liang, 2008, Characteristics of debris flows in Beichuan epicenter of the 1069
Wenchuan earthquake triggered by rainstorm on September 24, 2008: Journal of 1070
Engineering Geology, v. 16 no. 6, p. 751–757. 1071
1072

51

Tang, C., J. Zhu, W. L. Li, and J. T. Liang, 2009, Rainfall-triggered debris flows following the 1073
Wenchuan earthquake: Bulletin of Engineering Geology and the Environment, v. 68, no. 1074
2, p. 187–194, doi:10.1007/s10064-009-0201-6. 1075
1076
Tang, C., J. Zhu, M. Chang, J. Ding, and X. Qi, 2012, An empirical–statistical model for 1077
predicting debris-flow runout zones in the Wenchuan earthquake area: Quaternary 1078
International, v. 250, p. 63–73, doi:10.1016/j.quaint.2010.11.020. 1079
1080
Wei, Z., Y.-P. Xu, H. Sun, W. Xie, and G. Wu, 2018, Predicting the occurrence of channelized 1081
debris flow by an integrated cascading model: A case study of a small debris flow-prone 1082
catchment in Zhejiang Province, China: Geomorphology, v. 308, p. 78–90, 1083
doi:10.1016/j.geomorph.2018.01.027. 1084
1085
Wilford, D. J., M. E. Sakals, J. L. Innes, R. C. Sidle, and W. A. Bergerud, 2004, Recognition of 1086
debris flow, debris flood and flood hazard through watershed morphometrics: Landslides, 1087
v. 1, no. 1, p. 61–66, doi:10.1007/s10346-003-0002-0. 1088
1089
Xu, Q., S. Zhang, W. L. Li, and T. W. J. van Asch, 2012, The 13 August 2010 catastrophic 1090
debris flows after the 2008 Wenchuan earthquake, China: Natural Hazards and Earth 1091
System Science, v. 12, no. 1, p. 201–216, doi:10.5194/nhess-12-201-2012. 1092
1093
Yang, W., W. Qi, M. Wang, J. Zhang, and Y. Zhang, 2017, Spatial and temporal analyses of 1094
post-seismic landslide changes near the epicentre of the Wenchuan earthquake: 1095
Geomorphology, v. 276, p. 8–15, doi:10.1016/j.geomorph.2016.10.010. 1096
1097
Zhang, S., and L. M. Zhang, 2017, Impact of the 2008 Wenchuan earthquake in China on 1098
subsequent long-term debris flow activities in the epicentral area: Geomorphology, v. 1099
276, p. 86–103, doi:10.1016/j.geomorph.2016.10.009. 1100
1101
Zhang, S., L.-M. Zhang, H.-X. Chen, Q. Yuan, and H. Pan, 2013, Changes in runout distances of 1102
debris flows over time in the Wenchuan earthquake zone: Journal of Mountain Science, 1103
v. 10, no. 2, p. 281–292, doi:10.1007/s11629-012-2506-y. 1104
1105
Zhou, W., C. Tang, T. W. J. Van Asch, and C. Zhou, 2014, Rainfall-triggering response patterns 1106
of post-seismic debris flows in the Wenchuan earthquake area: Natural Hazards, v. 70, 1107
no. 2, p. 1417–1435, doi:10.1007/s11069-013-0883-8. 1108
1109
Zhuang, J., P. Cui, K. Hu, X. Chen, and Y. Ge, 2010, Characteristics of earthquake-triggered 1110
landslides and post-earthquake debris flows in Beichuan County: Journal of Mountain 1111
Science, v. 7, no. 3, p. 246–254, doi:10.1007/s11629-010-2016-0. 1112
1113
1114

52

CHAPTER 3: Glacial lake outburst floods set the pace of 1115
erosion in the Himalaya 1116
3.1 INTRODUCTION 1117
3.1.1 Channel Network Response to Tectonic and Climatic Forcing 1118
In mountainous regions, the rhythms of erosion and exhumation are governed by a 1119
variety of different surface processes operating in regions of the landscape with distinct slope, 1120
elevation, and drainage characteristics. As tectonic and climatic forcings change, transitions 1121
between process regimes migrate accordingly to maintain a configuration where accretionary 1122
flux into the orogen and erosional flux out of it are relatively balanced, both by tectonic 1123
adjustment of the geometry of the orogenic wedge and transient changes in erosion rate at the 1124
surface. In the context of an evolving collisional orogen, sustained, rapid uplift pushes terrain 1125
above the equilibrium line altitude (ELA) where snow persists from one year to the next and 1126
glaciers may form and constitute an important control on the height and relief of a range 1127
(Mitchell and Montgomery, 2006; Egholm, et al., 2009). Increased seasonal discharge from 1128
snowmelt, glacial lake outburst floods (GLOFs), especially during periods of glacial retreat, and 1129
orographic precipitation generated by rising topography enhance discharge and therefore incision 1130
downstream (Bathurst and Ashiq, 1998; Bookhagen and Burbank, 2010; Chen et al., 2010; Cook 1131
et al., 2018; Roe et al., 2013; Xu, 1988). Tributary channels with steep, unglaciated headwaters 1132
which drain into major rivers that transect the range steepen in response to drops in their relative 1133
base level driven by accelerated incision in trunk streams (Ouimet et al., 2009; Schumm, 1993). 1134
Landsliding rates begin to increase once hillslopes reach a threshold steepness, providing ample 1135
material to source debris flows, which drive incision in headwater catchments (Baer et al., 2017; 1136

53

Bovis and Jakob, 1999; Burbank et al., 1996; Harvey, 2001; Larsen and Montgomery, 2012; 1137
Marc et al., 2019; Montgomery, 2001; Quackenbush et al., in review). As channels steepen to 1138
facilitate more effective fluvial erosion, debris flows can effectively incise further downstream 1139
(Stock and Dietrich 2003; 2006; Penserini et al., 2017) (Figure 1). Decoupled erosion and 1140
precipitation in the Himalaya provide evidence for mass failures and outburst floods, among 1141
other extreme events, playing a central role in erosion (Burbank et al, 2003). For detailed 1142
examination of these systems of feedbacks surrounding landscape-scale hillslope/channel 1143
network systems, the Nepal Himalaya is an ideal study location. Several major rivers have their 1144
headwaters in Tibet and flow across the High Himalaya to the Lesser Himalaya, ultimately 1145
draining onto the Gangetic Plain. Along their courses, these rivers capture tributaries with basins 1146
widely varying in topographic characteristics, and, therefore, dominant geomorphic processes 1147
(Montgomery and Foufoula-Georgiou, 1993; Whipple and Tucker, 1999). 1148
1149
1150
Figure 1. Schematic showing channel response to increased uplift and, ultimately, 1151
glaciation in a region where hillslopes have achieved threshold angles. 1152
1153
A prominent physiographic transition defined by active, out-of-sequence thrusts separates 1154
the precipitous High Himalaya from the gentler Middle Hills (Dhital et al., 2015; Hodges et al., 1155
2004; Wobus et al., 2005; 2006). The abruptness of the transition may indicate that a change in 1156

54

dominant process domains is facilitating the topographic response to the steep gradient in uplift 1157
rate. Central Nepal lies along a relatively structurally straightforward segment of the Himalaya, 1158
simplifying the relationship between tectonics and erosion. A major structural nappe near 1159
Kathmandu juxtaposes high-grade rocks from the High Himalayan sequence into the Lesser 1160
Himalayan physiographic region without significantly affecting the topography (Dhital, 2015), 1161
allowing lithologic changes to be ruled out as a first-order control on geomorphology in the 1162
region. The 2015 MW 7.8 Gorkha earthquake generated ~25,000 landslides, and >1,000 debris 1163
flows indirectly during the monsoon seasons following the quake (Dahlquist and West, in 1164
revision; Roback et al., 2018) (Figure 2A). The distribution of landslides reflects patterns of 1165
long-term exhumation, as predicted by the threshold hillslope model, demonstrating that 1166
coseismic landslides are a mechanism by which hillslope erosion keeps pace with tectonic uplift 1167
(Quackenbush et al., in review). Careful analysis of the reginal topography, including channel 1168
and basin forms, and the distribution and geometry of debris flows, may reveal the transmission 1169
of the pulse of erosion imposed by the earthquake downstream by debris flow action, and the 1170
ways that landscape form reflects transitions between geomorphic process regimes can be better 1171
understood. 1172

55

FIGURE 2A
FIGURE 2B

56

FIGURE 2C
FIGURE 2D

57

FIGURE 2E

FIGURE 2F

58

FIGURE 2G

Figure 2A. Map of basins used in our analysis showing their position relative to the LGM 1173
ELA. Dashed line represents the approximate location of the physiographic transition, as 1174
mapped visually based on an abrupt increase in landscape relief and steepness. Mapped 1175
debris flows are from Dahlquist and West (in revision). 2B. Basin-averaged ksn for study 1176
basins. Basins entirely above 4200 meters are not included, as their channel geometry is 1177
likely controlled more by glacial erosion than channel processes. 2C. ksn differences 1178
between tributary and trunk channels. Tributaries are defined as having Strahler order 1 1179
or 2, trunk channels as order 4 or above. We include only channels below 4200 meters, and 1180
ksn differences are normalized to the values in the tributary streams. 2D. Basin-averaged 1181
ksn differences between tributary and trunk streams. 2E. a1 values for basins. Note that the 1182
color bar is inverted (lower values in warmer colors), as lower values of a1 indicate more 1183
effective debris flow erosion. 2F. Chi map. We exclude channels above 4200 meters for this 1184
calculation to avoid including channels whose geometry is controlled by glacial erosion. 2G. 1185
Density of channels in each basin exceeding the 11º threshold above which debris flow may 1186
drive incision. 1187
1188
Aiming toward a holistic understanding of incision and sediment transport in rapidly 1189
uplifting and eroding basins, we recognize several important questions. What controls how far 1190
downstream debris flows can effectively incise (Stock and Dietrich 2003; 2006, Penserini et al., 1191
2017)? How do channels incise to keep up with uplift when delivery of coarse sediment from 1192
landslides and debris flows should inhibit incision in many places (Benda, 1990; Brayshaw and 1193
Hassan 2009; Pratt et al., 2002; Pratt-Sitaula et al., 2004; Seidl et al., 1994; Johnson et al., 2007; 1194

59

Shobe et al., 2016, Yanites et al., 2011)? How might the abrupt physiographic transition between 1195
the Lesser and High Himalaya reflect a shift in process domain, and how do transitions between 1196
process domains transmit a signal of regional uplift, manifesting as a change in base level, from 1197
river mouths up to ridges, if incision is largely controlled by debris flows and GLOFs (Anderson 1198
et al., 2015; Livers and Wohl, 2016; Montgomery, 1999)? In this chapter, we attempt to bridge 1199
process domains using an inventory of debris flows associated with the Gorkha earthquake and 1200
topographic analysis using multiple proxies at multiple scales to answer these questions. Based 1201
on the apparent primacy of landslides and debris flows in steep but unglaciated headwater low- 1202
order basins and of GLOFs in large rivers draining the High Himalaya, we approach these 1203
problems in the framework of a hypothesis that erosion is responding to tectonic forcing from the 1204
top down (Figure 1), as opposed to transient signals migrating upstream from regional base level. 1205
1206
3.1.2 Physical Relationships in Fluvial and Debris Flow Channel Networks 1207
3.1.2.1 The stream power law and normalized channel steepness 1208
In actively uplifting landscapes, the geometry of the land surface is largely governed by 1209
competition between uplift and gravity, mediated by a series of processes with a variety of 1210
controlling factors. In time, this competition tends to result in stalemate, a time-invariant 1211
condition of topographic steady state (Whipple and Tucker, 1999; Willett and Brandon, 2002). 1212
For most of the planet, local boundary conditions for erosion are set by the pace of incision or 1213
aggradation by channel processes. In the channel networks that are the most prominent features 1214
in these landscapes, the relationship between channel slope and contributing drainage area is 1215
illustrative of the processes that define a segment of the channel network. Downstream reaches 1216
of the channel network, controlled by fluvial processes are described by the power law function 1217

60

𝐸𝐸 = 𝐾𝐾 𝐴𝐴 𝑚𝑚 𝑆𝑆 𝑛𝑛 (1) 1218
where E is erosion rate, K is the erosion coefficient, which is governed by local lithology, 1219
climate, and the process that control incision in the area, A is drainage area, S is local slope, m 1220
and n are empirical constants which have a range of possible vales depending on local conditions 1221
(Howard and Kerby, 1983). Under steady-state conditions, where uplift and erosion can be 1222
assumed to be equal, this results in 1223
𝑆𝑆 = �
𝑈𝑈 𝐾𝐾 �
1
𝑛𝑛 𝐴𝐴 𝑚𝑚 𝑛𝑛 (2) 1224
where U is uplift (Whipple and Tucker, 1999). This equation can be recast as 𝑆𝑆 = 𝑘𝑘 𝑠𝑠 𝐴𝐴 − 𝜃𝜃 , known 1225
as Flint’s Law, where ks defines a channel steepness �
𝑈𝑈 𝐾𝐾 �
1
𝑛𝑛 , and 𝜃𝜃 = m/n, termed the concavity, 1226
represents the slope of a slope-area trend in log-log space and is generally accepted to be 1227
insensitive to uplift rate (Flint, 1974). ks varies with uplift rate, but contains units that are 1228
dependent on 𝜃𝜃 . In order to make a reasonable comparison of ks among channels with different 1229
𝜃𝜃 , we must fix the value of 𝜃𝜃 to a 𝜃𝜃 𝑟𝑟𝑟𝑟 𝑓𝑓 that represents an average value for the channels in the 1230
area of interest, typically between 0.35-0.65, but that may vary widely depending on local factors 1231
(Wobus et al., 2006). Fixing 𝜃𝜃 to 𝜃𝜃 𝑟𝑟𝑟𝑟 𝑓𝑓 results in the normalized channel steepness index ksn 1232
which is calculated as a best fit value for a given channel reach and is frequently and effectively 1233
used as a proxy in broad comparisons of uplift and incision rates across landscapes (Dibiase et 1234
al., 2010; Kirby and Whipple, 2001; 2012). One limitation of ksn that is sometimes ignored in the 1235
way it is applied is that by definition, it only directly applies to channels where the dominant 1236
mode of incision is fluvial processes. For example, incision by lake outburst floods is a vastly 1237
more efficient process than incision by ordinary discharge (Cook et al., 2018), in that it can do 1238
more erosive work on lower gradient channels with less contributing drainage area, meaning ksn 1239

61

could systematically underestimate incision in channels where outburst flooding is an important 1240
geomorphic agent. Or, in very steep catchments such as those examined in this study, debris 1241
flows can be the process that controls channel geometry at drainage areas of up to ~5 km
2
. Since 1242
channels incising due to debris flow action do not follow a power law relationship between slope 1243
and drainage area, the use of ksn simply as an uplift-incision proxy in these catchments is 1244
problematic (Stock and Dietrich, 2006). In regions where other processes besides fluvial incision 1245
are controlling the geometry of channels, patterns of ksn may instead be most useful for 1246
elucidating transitions between process regimes. We calculated ksn excluding areas where 1247
channel geometry reflects glacial erosion (above 4200 meters). We examine relationships 1248
involving basin-averaged ksn (Figure 2B), and differences between ksn in tributaries and trunk 1249
streams at confluences (Figures 2C and 2D). 1250
1251
3.1.2.2 a1 and debris flow dominated channels 1252
Steep, low order channels in active landscapes often have geometry controlled by debris 1253
flow action. These channels plot as a curve in slope-area space, representing a continuum 1254
between hillslopes and the fluvial channel network. Stock and Dietrich (2003) proposed an 1255
equation 1256
S = S
0
/(1 + 𝑎𝑎 1
A
𝑎𝑎 2
) (3) 1257
to describe the curved slope-area trend in debris flow channels where S0 is the slope at a valley 1258
head, which is dependent on hillslope processes, a1 is a constant that is inversely proportional to 1259
the curvature of the debris flow network, and a2 describes the slope of the fluvial network in log- 1260
log space at the transition from the debris flow regime. Similarly to 𝜃𝜃 in the ksn calculation, a2 1261
can be fixed to a reference value to compare a1 among drainage basins. a1 has been shown to 1262

62

scale with coupled uplift-incision, describing the increase in efficacy of debris flow erosion in 1263
more rapidly uplifting basins (Penserini et al., 2017; Stock and Dietrich, 2003). The shape of this 1264
curved region in slope-area space describes how far downstream debris flows can effectively 1265
drive incision. The transition between debris flow and fluvial channels will shift to the right as 1266
uplift and erosion increase, though the means by which this occurs is unclear. A classic model of 1267
fluvial incision involves the formation of a knickpoint, or localized steepening in response to 1268
uplift which manifests as a drop in a river’s base level. The increased steepness at a knickpoint 1269
causes localized increased erosion, and the knickpoint propagates upstream. Knickpoint retreat 1270
driven by GLOF action has been measured in basalt lithologies in Iceland, where erosion driven 1271
by the toppling of columns results in a landscape ideally suited to erode by knickpoint retreat 1272
(Baynes et al., 2015). 1273
However, in regions where debris flow or GLOF incision is facilitated by abrasion, this 1274
model may not be accurate. Channels in which erosion is dominated by bedload abrasion may 1275
have knickpoint retreat rates that are decoupled from overall incision rates (Jansen et al., 2013; 1276
Wilson et al., 2013) and knickpoints may be rapidly (years to decades) smoothed out in the 1277
presence of copious bedload and sufficient discharge (Cook et al., 2013). This is likely consistent 1278
with erosion in the Nepal Himalaya, as lithologies are not ideally suited to a plucking-dominated 1279
erosion model which could optimize knickpoint propagation, and sediment flux from hillslopes 1280
is plentiful. Instead, we propose an interplay of top-down and bottom-up processes, where the 1281
initiation of debris flows and incision they drive in headwater reaches is largely independent of 1282
incision in downstream channels, but steepening in those channels in ultimately enables debris 1283
flows to drive incision further downstream. We calculated a1 for basins throughout the study area 1284

63

to identify relationships between debris flow incision and large-scale landscape evolution in the 1285
Nepal Himalaya (Figures 2E, 3). 1286
1287
Figure 3. Slope-area plots for basins shown in figure 6, showing fit with equation 3 and 1288
parameter values. With a2 fixed to a reference value (0.62, best fit for all basins), a wider 1289
region of curvature between the hillslope-to-colluvial valley transition and the colluvial 1290
valley-to-fluvial network transition represents increased debris flow incision, specifically 1291
propogation of debris flow incision further downstream. 1292
1293
1294

64

3.1.2.3 The integral method of channel profile analysis 1295
While ksn has proven a useful metric to compare incision across rivers, slope 1296
measurements can be quite noisy, particularly if available topographic data are fairly coarse, and 1297
potentially interesting features in channel and basin geometry may be obscured. To facilitate the 1298
identification of areas undergoing transient erosion, Perron and Royden (2013) proposed the 𝜒𝜒 1299
metric, defined as 1300
𝜒𝜒 = ∫ �
𝐴𝐴 0
𝐴𝐴 �
𝑚𝑚 / 𝑛𝑛 𝑑𝑑𝑑𝑑 𝑥𝑥 𝑥𝑥 𝑏𝑏 (4) 1301
where where xb is a point on the channel at base level, 1302
and A0 is an arbitrary reference drainage area that gives 𝜒𝜒 dimensions of length. Performing the 1303
𝜒𝜒 transform on a river which has a concave-up, steady-state profile and plotting 𝜒𝜒 versus 1304
elevation generates a linear plot where any perturbation represents a transient adjustment. The 𝜒𝜒 1305
calculation is sensitive to U and K, as well as choice of base level and m/n, and carries the same 1306
assumption of detachment-limited fluvial incision as ksn, but has proven useful in a variety of 1307
applications. Most notably, 𝜒𝜒 differences in channels opposite a drainage divide are expected to 1308
reflect divide instability, with a lower-χ stream expected to capture area from a higher-χ stream 1309
(Willett et al., 2014). We apply χ transforms for this purpose, to examine whether increased 1310
incision rates due to increased GLOF frequency may translate to drainage area capture by the 1311
tributary basins to the GLOF channels (Figure 2F). 1312
1313
3.1.2.4 Channel width and normalized channel wideness 1314
Another common observation in fluvial networks is the power-law increase in the width 1315
of the active channel with contributing drainage area. This relationship is governed by a number 1316
of factors, including erosion rate, lithology, and climate, among others. Particularly in regions 1317

65

where extreme events can generate massive sediment inputs, channel width increases due to 1318
aggradation (i.e. Yanites et al., 2019) or decreases to concentrate unit stream power to flush out 1319
sediment (i.e. Croissant et al., 2017). Dynamic channel width may thus be illustrative of channel 1320
response to tectonic or process-driven forcing. We can approach a width-area trend using an 1321
equation with the same form as slope-area, although the relationship between area and width is 1322
positive, where 1323
𝑊𝑊 = 𝑘𝑘 𝑤𝑤 𝐴𝐴 𝑏𝑏 (5), 1324
where W is the channel width, kw is a channel wideness index analogous to ks, and b is the power 1325
law exponent. By fixing a best-fit reference value for b, we can examine local variation in 1326
channel wideness in response to enhanced erosion by increased GLOF activity, calculating a 1327
normalized channel wideness index as 1328
𝑊𝑊 = 𝑘𝑘 𝑤𝑤 𝑛𝑛 𝐴𝐴 𝑏𝑏 𝑟𝑟𝑟𝑟𝑟𝑟
(6) 1329
(Allen et al., 2013; Yanites, 2018; Yanites et al., 2018). Since our channel width measurements 1330
are most interested in the response of large rivers with headwaters in Tibet, and a dramatic 1331
gradient in precipitation exists from the Gangetic Plain to the Tibetan Plateau, we adjust drainage 1332
area to mean annual precipitation to correct for a width-discharge trend (Figure 4). 1333
1334
3.1.3 Toward a Threshold Channel Model 1335
In regions experiencing rapid coupled uplift and erosion, the threshold hillslope model 1336
describes how steep hillslopes respond to uplift-driven river incision. Under conditions of 1337
increasing uplift and incision, slopes steepen to a threshold angle governed by material strength. 1338
Further increases in incision rate are expressed by an increase in rate of landsliding rather than 1339
steepening, as river incision into the base of a threshold slope conditions it to fail when perturbed 1340

66

by an earthquake or a large store. Additionally, in tectonically active regions, or regions where 1341
extreme weather may destabilize hillslopes, landslides may mobilize sediment and drive hillslope 1342

Figure 4 (top). Channel
width vs. discharge in the
Nepal Himalaya. Discharge
data are derived from
average TRMM
precipitation between
1998-2009 (Bookhagen, in
review). We fit equation 6
to the width-discharge
curve, with coefficients
shown in the inset
equation. 5B. Channel
width normalized to
discharge vs. glacial
drainage area. We use the
modern ELA value in the
central Nepal Himalaya of
5500 meters to define
glaciated drainage areas
(Asahi et al., 2010). Bars
represent quantile bin
centers with median and
top and bottom quartiles.

lowering irrespective of river incision (Korup et al., 2010).
The threshold hillslope/river incision/landslide cycle contains its own system of 1343
feedbacks: Landslides deliver debris to steep, colluvial channels. Landslide debris, mobilized and 1344
channelized as debris flows, drives incision in headwater channels. Further downstream, debris 1345
flows often transport huge boulders into reaches of the channel network considered to be 1346

67

dominated by fluvial processes controlled by typical monsoon floods, despite that, under most 1347
circumstances, fluvial processes are inadequate to transport boulders as large as debris flows can 1348
deliver (Cook et al., 2018) (Figure 5). In a study of gravel bars in the Marsyandi river in central 1349
Nepal alongside local landslides which constitute potential sediment sources, Attal and Lavé 1350
(2006) found that local bedload may be largely reflective of landslides on local hillslopes, 1351
highlighting the importance of hillslope-channel coupling. Since bedload composition, 1352
availability, and grain size are major controls on channel incision, particularly in abrasion 1353
regimes, this evidences the importance of local landslides in controlling local channel incision 1354
(Benda and Dunne, 1997; Cook et al., 2013; Lamb et al., 2008; Sklar and Dietrich, 2001; 2004). 1355
In glaciated mountain ranges, glacial lake outburst floods (GLOFs) generate erosive forces that 1356
can extend farther downstream than storm-triggered debris flows to evacuate debris flow 1357
deposits and drive incision in low-gradient rivers with thick, coarse bed cover. 1358
Because of the complexity of the transitions and feedbacks between process domains, 1359
study of any of these processes in isolation, or broadly generalizing a mathematical relationship 1360
that describes a specific physical process (such as detachment-limited river incision) will 1361
necessarily produce an incomplete picture. Hillslope-channel relationships are defined as an 1362
abrupt transition between process domains. Because hillslopes and channels are geometrically 1363
distinct, it is simple to delineate which parts of a landscape are dominated by hillslope versus 1364
channel processes. Transitions between process domains in channels dominated by different 1365
processes are more complicated to distinguish, and the boundaries between them are less distinct. 1366
For example, debris flows dominate incision in most steep headwater catchments, but debris 1367
flow runouts can vary with channel geometry and the conditions leading to their initiation 1368
(Brayshaw and Hassan, 2009; Fan et al., 2017; Hungr, 2005), blurring the transition between the 1369

68

domain they dominate and the regions of the channel network dominated by fluvial incision 1370
farther downstream This leads to a complicated system with numerous feedbacks among 1371
processes, the geometry of hillslopes and channels, the frequency and magnitude of extreme 1372
events, and other factors. However, the basic relationship described by the threshold hillslope 1373
model, where one process sets base level for the process occurring directly above it, and 1374
increasing erosion rates are expressed as continued steepening of the landscape only up to a 1375
point, can be applied broadly to a channel network where a diversity of processes are active in 1376
channels of different orders with different geometries. In this chapter, we will present evidence 1377
for a threshold channel model to accompany the threshold hillslope model, particularly focusing 1378
on the cooperation and competition between debris flows and glacial lake outburst floods 1379
(GLOFs) in the Nepal Himalaya, and discuss the implications of the system of feedbacks 1380
between channel processes on shaping the Himalaya in particular and glaciated mountain belts in 1381
general. 1382
1383
3.1.4 Regional Geology and Important Geomorphic Features in the Nepal 1384
Himalaya 1385
The Nepal Himalaya is relatively structurally straightforward compared to other regions 1386
of the orogen. The Main Central Thrust, which generated the 2015 Gorkha Earthquake, 1387
juxtaposes the low-grade to unmetamorphosed sediments of the Lesser Himalayan Sequence 1388
rocks against the gneisses and migmatites of the Greater Himalayan Crystalline Complex 1389
(GHCC) (Figure 6). One of the most prominent topographic features in this region is the 1390
physiographic transition between the gentle Lesser and more rugged High Himalaya. There is a 1391
sharp contrast in slope and relief between the two regions, tectonically controlled by an abrupt 1392

69

increase in uplift rate north of the transition, generally accepted as being driven by the presence 1393
of the Mid-Crustal Ramp, a steepening in the décollement at depth (Lemonnier et al., 1999), but 1394
at least partially accommodated by surface-breaking thrust faults around the physiographic 1395
transition (Hodges et al., 2004; Wobus et al., 2005). Coincidence of the loci of deformation and 1396
peak precipitation suggest climate-erosion-tectonics feedbacks (Hodges et al., 2004). A nappe 1397
juxtaposes high-grade GHCC rocks into the Lesser Himalayan physiographic region around 1398
Kathmandu and to the east, with little to no effect on relief. 1399
1400
Figure 5. Field photo from Melamchi River, Nepal. Location is 28.0105º N, 85.5355º N. 1401
Centered at the photo location, the average channel gradient over a 1 km reach of the river 1402
is ~0.096, or ~5.5º, well below the slope where we expect significant debris flow incision. 1403
Large boulders in the channel protect the bed and are unlikely to be mobile except during 1404
GLOFs or extreme debris flows. For reference, the truck with the open door is 1405
approximately equidistant to the photographer with the large boulder at the center of the 1406
frame. Little to no bedrock is exposed in the channel, although there is extensive exposure 1407
in the adjacent canyon walls. 1408
1409

70

1410
Figure 6. Overview map showing points of interest to this study and simplified 1411
lithostratigraphy from Amatya and Jnawali (1994). Channel networks outlined in blue and 1412
orange are shown in chi plots in Figure 11. 1413
1414
In the Himalaya, steep terrain and the summer monsoon result in frequent debris flows, 1415
and moraine and landslide-dammed lakes catastrophically drain often enough to make them an 1416
important agent of geomorphic change. Lake outburst floods contribute to driving downward 1417
incision on the channel bed and undercutting hillslopes to trigger landslides. In the long term, 1418
uplift rates are generally balanced by incision, though episodes of aggradation and valley fills or 1419
episodes of transient incision can persist for long durations (Lavé and Avouac, 2001; Hodges et 1420
al., 2001; Pratt-Sitaula et al., 2004; Ruhl and Hodges, 2005; Schwanghart et al., 2016). For 1421
example, the Kathmandu Valley paleolake persisted through a wide variety of tectonic and 1422
climatic changes throughout the Pleistocene (Sakai et al., 2002). Recurring large earthquakes 1423
contribute to mountain building and uplift and landslide-driven erosion and sediment export (Li 1424
et al., 2014), while monsoon-triggered landslides may contribute a roughly equal amount to the 1425
overall erosion budget. The largest landslides (>1 km
2
) which have the greatest overall impact on 1426
erosion have recurrence intervals of ~1000 years in catchments (Marc et al., 2019). These factors 1427

71

result in great variability between short and long-term erosion rates, as is consistent with other 1428
steep landscapes where erosion is dominated by extreme events (Anderson et al., 2015; Kirchner 1429
et al., 2001; Thiede and Ehlers, 2013). However, Li et al., (2017) examined the co-seismic 1430
landslide distribution from the 2008 Wenchuan earthquake alongside erosion rates at yr-kyr 1431
timescales and Myr exhumation rates from low-temperature thermochronology in the Longmen 1432
Shan. Considering these rates at various scales and estimated recurrence intervals for 1433
earthquakes in this region, they propose that coseismic landslide-driven denudation mirrors long- 1434
term exhumation over many earthquake cycles. We expect that this observation can be 1435
reasonably extrapolated to other steep, active mountain belts, including the Himalaya. 1436
1437
3.2 RESULTS AND DISCUSSION 1438
3.2.1 Debris Flows and the Gorkha Earthquake 1439
Co- and post-seismic landslides associated with the Gorkha Earthquake and its 1440
aftershocks sourced more than 1000 debris flows in the ensuing monsoon seasons (Dahlquist and 1441
West, in revision). Two primary mechanisms were at work. Debris flows can be mobilized 1442
during heavy rains from stationary landslide material that has been dumped into steep, colluvial 1443
channels (Prancevic et al., 2014; Takahashi, 1978), or a landslide can initiate in saturated near- 1444
surface conditions during heavy rains and become channelized as a debris flow, retaining its 1445
momentum (Iverson et al., 1997; Montgomery and Dietrich, 1994). We refer to these 1446
mechanisms as Type-1 and Type-2, respectively. The channels hosting flows of each type have 1447
some important distinctions in their geometry, which we will discuss further. Debris flows 1448
occurring by a combination of these mechanisms, as well as Type-2 debris flows sourced from 1449
storm-triggered landslides in the years before the Gorkha Earthquake, were also identified and 1450

72

appear to have distinct geometry. Broadly, Type-1 debris flows occurred in steeper channels than 1451
post-seismic Type-2, which occurred in steeper channels than Type-2 debris flows occurring 1452
before the Gorkha Earthquake. Different types of debris flow channels have similar average 1453
concavity, but differ in steepness (Figure 7). Pre-Gorkha Type-2 debris flow channels have a 1454
slope distribution that is nearly identical to the slopes of all channels in the study area above a 1455
threshold of ~11° (Seidl and Dietrich, 1992). This is a critical observation to our understanding 1456
of debris flow-dominated channels and will be discussed in depth. 1457
Figure 7. (Top)
Cumulative
distribution of slope
over total length of
debris flow channels
initiated by different
mechanisms.
Minimum
contributing
drainage area is 0.01
km2. While some
debris flows
appeared to initiate
with smaller
contributing
drainage areas, the
location of colluvial
channels defined by
a 30-meter DEM
with smaller
drainage areas than
this do not appear to
correspond
accurately with the
locations of channels
in satellite imagery.
“All channels” refers
to every channel in
the area affected by
coseismic landsliding, including those where no debris flows were mapped. (Bottom) Slope-
area plots for debris flow channels. Bars represent upper and lower quartiles for each area
bin and are drawn on bin centers.

73

3.2.2 Evidence from a Debris Flow Inventory 1458
The slope distributions of debris flows triggered by different mechanisms raises some 1459
interesting questions about how erosion propagates throughout the channel network after a large 1460
earthquake. In regions where the detachment-limited stream power law accurately describes 1461
erosion, channel incision is expected to take place primarily by the upstream migration of 1462
knickpoints at a rate proportional to upstream drainage area. Erosion in response to tectonic 1463
uplift progresses as a lowering of base level, which steepens downstream reaches of rivers. The 1464
knickpoints formed by this process travel upstream, as steepened reaches achieve greater erosive 1465
power with the same discharge. This mechanism is often invoked to describe channel processes 1466
and landscape evolution in regions where it does not necessarily best describe the actual 1467
dominant processes. As an illustration, if the detachment-limited stream power law best 1468
describes channel incision in most of the Nepal Himalaya, we should expect to see the 1469
physiographic transition prominently manifest as a sharp increase in normalized channel 1470
steepness (ksn), which is derived from the stream power law, corresponding with the sharp 1471
increase in long-term exhumation rates. Instead, basin-averaged ksn, while generally increasing 1472
across the transition, does not sharply define a boundary. The a1 metric, which instead describes 1473
the magnitude of incision by debris flows in a basin, defines a much starker contrast between the 1474
slowly eroding Lesser Himalaya and the rapidly eroding High Himalaya (Figures 2B and 2E). 1475
This indicates that the dominant mechanism of incision in most of the channels in the Nepal 1476
Himalaya is debris flows, and operating on the assumption that the dynamics of debris flows best 1477
describes the way the majority of the landscape in this region evolves will lead to the most 1478
accurate predictions about the behavior of these landscapes. With this in mind, we identify some 1479

74

important trends in the geometry of channels that hosted debris flows associated with the 2015 1480
Gorkha Earthquake. 1481
The slope distributions of channels hosting different types of debris flows reveal some 1482
insights into the way a pulse of incision passes through the channel network in the wake of a 1483
major earthquake. Assuming that debris flows are the primary mechanism of incision in channels 1484
where they occur, this may have implications for the long-term evolution of landscapes subject 1485
to large earthquakes. Type-1 debris flows occur in the steepest channels, and are by definition 1486
controlled by the distribution of co-seismic landslides, while post-seismic Type-2 flows occur in 1487
gentler sloping channels, and Type-2 flows from the end of the inter-seismic period gentler still 1488
(Figure 7). Throughout an earthquake cycle, therefore, the steepest headwater channels receive a 1489
pulse of incision first, which eventually progresses to less steep channels throughout the post- 1490
seismic and inter-seismic periods. Additionally, Type-1 debris flows almost exclusively occurred 1491
during the first monsoon following the earthquake. Likely, the vast majority of loose co-seismic 1492
landslide material that is vulnerable to remobilization in debris flows is mobilized during the first 1493
few intense rainstorms (Dahlquist and West, in revision). This is an important observation in the 1494
context of a coupled threshold hillslope and channel model. In a landscape where hillslopes have 1495
achieved threshold slopes, the rate of landsliding increases to accommodate more rapid uplift and 1496
erosion instead of further hillslope steepening. This necessarily means more landslide debris will 1497
be delivered to steep channels in these landscapes. If landslide debris is readily flushed out 1498
during monsoon rains when it occupies steep colluvial channels, the debris flow rate, over the 1499
long term, may scale with the landslide rate (Baer et al., 2017; Bovis and Jakob, 1999). In the 1500
Nepal Himalaya, we can conclude that that the assumption that landslide debris is exported 1501

75

relatively quickly from steep channels is reasonable, as little alluviation is apparent in headwater 1502
channels, even in the less steep areas of the Lesser Himalaya. 1503
To fully understand transitions between process regimes, it is necessary to identify 1504
thresholds that govern the processes involved. Above ~11°, the slope distributions of channels 1505
hosting Type-2 debris flows occurring during the interseismic period, which affected the least 1506
steep channels, are nearly identical to the distributions of channel slopes in the entire study area, 1507
including those without debris flows (Figure 7). This indicates that above the ~11º bed angle, the 1508
geometry of channels in the study area is primarily controlled by debris flow incision. This ~11º, 1509
or 0.2 value can also be observed in the slope-area plots as the major break in slope that 1510
represents the transition to the fluvial regime. The prevalence of channels in a basin that exceed 1511
11º in slope is also an effective metric with which to identify the physiographic transition (Figure 1512
2G). This bed angle represents a threshold below which debris flows transition from a primarily 1513
erosive force to a force primarily driving sediment transport and deposition (Seidl and Dietrich, 1514
1992). It is also important to note the existence of a threshold slope for debris flow initiation, at 1515
~22º for Type-1 debris flows (Prancevic et al., 2014; Dahlquist and West, 2019). It is important 1516
to point out a broad observation for interpreting the relationship between patterns of ksn and 1517
debris flow incision, as well as a test of the effectiveness in examining the way ksn differences 1518
between tributary and trunk streams can elucidate process transitions. We find a strong positive 1519
trend between basin a1 and ksn differences between tributary and trunk streams (Figure 8). So, as 1520
debris flows are increasingly dominant in controlling incision in a basin, channel geometry 1521
becomes increasingly similar throughout that basin. 1522
1523
1524

76

Figure 8. Basin a1 vs basin
averaged, normalized ksn
differences. Only basins
at least partially below
4200 meters elevation are
included.

1525
3.2.3 Glacial Lake Outburst Floods Set Base Level for Debris Flow- 1526
Dominated Basins 1527
GLOFs are an important driver of channel incision in glaciated mountain belts, as well as 1528
a major hazard to life and infrastructure downstream. Outburst floods are capable of driving 1529
erosion more effectively than a flood of the same magnitude driven by runoff. Since in an 1530
outburst flood the pace of the water bore exceeds that of entrained bedload, the leading edge of 1531
the flood will remain below its transport capacity and capable of mobilizing additional material 1532
(Cook et al., 2018). A major flood occurs in the Himalaya about every two years (Korup and 1533
Tweed, 2007). Given the inaccessibility of many of the areas most affected, improved image 1534
analysis techniques may reveal they are broadly more common than current inventories have 1535
predicted. A study of GLOFs in the Hindu Kush Himalaya applying a new processing chain to 1536
Landsat images to detect flood events increased the size of the inventory from 1988-2016 by 1537

77

91% (Veh et al., 2018). In spite of such recent developments, their relative infrequency means 1538
that studies into their geomorphic impact are generally limited to case studies of individual 1539
events or small populations. To address this limitation, we will attempt to infer the long-term 1540
impacts of GLOFs using the geometry of channels we expect should be at least partially 1541
controlled by GLOF incision. Our analysis requires one important assumption: that the amount 1542
of contributing drainage area from glacial valleys in a given catchment serves as a proxy for 1543
GLOF frequency in that catchment. Rapid incision driven by increasing frequency of glacial lake 1544
outburst floods should manifest in the geometry of channels capturing varying amounts of 1545
glaciated terrain. 1546
We use the equilibrium line altitude (the altitude above which snow persists over the 1547
summer, allowing for the accumulation of glacial ice) during the last glacial maximum (4200 1548
meters) as the threshold elevation for defining regions which may have sourced GLOFs at 1549
timescales relevant to the analysis of channel long profiles. For analysis of width of the active 1550
channel, which should integrate processes over much shorter timescales, we use a modern ELA 1551
value of 5500 meters (Asahi, 2010) (Figure 4). 1552
The 2016 GLOF originating in the Zhangzangbo River traveled down the Bhote Khosi 1553
valley generating numerous landslides through undercutting of steep banks, while mobilizing 1554
coarse alluvium and driving vertical incision (Cook et al., 2018). The combination of vertical and 1555
lateral incision means the impact of GLOFs on channel width is not straightforward. Both 1556
bedrock and alluvial rivers follow a power law relationship between discharge and channel 1557
width. Following Typhoon Morakot in southern Taiwan, which produced an extremely high 1558
density of landslides in the steep Central Range, channels widened considerably due to the influx 1559
of sediment, even in locations where channels were relatively steep (Yanites et al., 2018). This 1560

78

indicates that sediment input from upstream drives aggradation and can control where bedrock 1561
incision takes place. Croissant et al. (2017) find model rivers narrow in the years and decades 1562
following an extreme sediment input to facilitate increased sediment transport. Normalizing the 1563
width of a channel to its discharge can reveal patterns of local widening or narrowing which in 1564
turn can say something about incision or deposition in these areas (Yanites, 2018). When we 1565
normalize channel width in large, north-south draining rivers capturing substantial amounts of 1566
glaciated drainage area, we see a weak trend toward lower normalized wideness in rivers likely 1567
experiencing more GLOFs (Figure 4). For this analysis, we use a modern value for the ELA of 1568
5,500 meters, since the width of the active channel can adjust more quickly to perturbations than 1569
can a river’s long profile. This indicates that, broadly, rivers experiencing more GLOFs are 1570
somewhat less likely to be heavily alluviated, which is the primary means by which an active 1571
channel would widen. Coarse alluvium deposited in the channel by landslides, debris flows, or 1572
previous GLOFs that could armor the bed and prevent incision does not remain for as long as it 1573
might in a channel with less frequent floods. 1574
A simple field observation from large, low-gradient rivers in the Nepal Himalaya – the 1575
presence of large (>1 meter long axis) boulders – highlights the need for a mechanism besides 1576
detachment-limited stream power to evacuate the coarse sediment and enable incision into 1577
bedrock. Glacial lake outburst floods (GLOFs) can drive vertical and lateral incision far 1578
downstream in these large rivers, mobilizing coarse material delivered by landslides and debris 1579
flows and pulverizing large sediment to facilitate downstream fining to a point where it can be 1580
mobilized by fluvial processes during runoff-driven monsoon flooding (Cook et al., 2018; Costa 1581
and Schuster, 1988; Dunning et al., 2013; Emmer et al., 2017; Lang et al., 2013; Mool, 1995; Xu, 1582
1988). The GLOFs allow for far greater erosional efficiency than during normal discharge-driven 1583

79

floods, which leads to effective channel incision occurring in channels with considerably lower 1584
ksn than would be expected if ordinary, precipitation or snowmelt driven floods were the 1585
dominant geomorphic agent. In the Langtang Valley since the Middle Holocene, glacial 1586
fluctuations, and ultimately, deglaciation, have driven accelerated incision and rapid reworking 1587
of earlier sediments, a pattern we expect to see elsewhere in the Himalaya (Barnard et al., 2006). 1588
Since GLOFs are relatively rare events compared to debris flows, which may occur in large 1589
numbers during a single intense storm, it is not possible to compile a large inventory of events, 1590
and studies are limited to a single event or small population. We must, then, infer the relative 1591
frequencies of GLOFs in different watersheds and the areas that should be considerably affected 1592
by these events. However, a critical observation from the 2016 Bhote Khosi GLOF, as well as 1593
another slightly larger GLOF originating in the same tributary in 1981, is revelatory as to the role 1594
of these events in controlling the pace of erosion in the High Himalaya. The GLOF flood peak 1595
attenuates as a flood travels downstream, and the peak discharge decreases, leading to a point 1596
where GLOF discharges are surpassed by a monsoon flood with the same return time. Recorded 1597
GLOFs in the Bhote Khosi occurred in 1935 and 1964 in addition to the 1981 and 2016 events, 1598
so a 30-year recurrence interval is reasonable (Mool, 1995, Cook et al., 2018). This point in the 1599
landscape is the point below which incision, and therefore channel geometry, is more controlled 1600
by non-GLOF fluvial incision. For the 1981 and 2016 GLOFs the Bhote Koshi valley, these 1601
crossover points was located 55 and 45 km downstream from the confluence with the 1602
Zhangzangbo River, respectively, for a 30-year monsoon flood (Cook et al., 2018; Xu, 1988). 1603
Crucially, these points are just below the physiographic transition, hinting that the sharp 1604
transition to the High Himalaya may be controlled to some extent by the downstream extent of 1605
GLOF incision. Above this point, GLOFs have been observed to mobilize boulders as large as 1606

80

13.4 meters in diameter (Xu, 1988). Meanwhile, Cook et al., (2018) observed boulders as small 1607
as 2-3 meters remained stable during all monsoon floods back to 2004. While bedrock incision 1608
during outburst floods cannot be conclusively verified, it is likely given that mobilization of 1609
armoring boulders is a prerequisite. 1610
Figure 10. ksn ratio
between tributary and
trunk stream versus
amount of glaciated
drainage area in trunk
stream headwaters for
all confluences in
Figure 2C where trunk
streams had glaciated
headwaters. We use the
LGM ELA (4200
meters) to define
glaciated regions in this
calculation, since we
assume that river long
profiles have not
adjusted to
deglaciation. Bars
represent quantile bin
centers with mean and
one standard deviation
error.
1611
While these observations form a foundation for the hypothesis of top-down, GLOF- 1612
controlled erosion in the High Himalaya, the relative infrequency of these events necessitates 1613
that we make inferences based on topographic analysis of basins affected by outburst floods. To 1614
compare the effect of GLOF incision among rivers, we hold a simple but critical assumption: that 1615
GLOF frequency down a river is proportional to the amount of glaciated drainage area the river 1616
captures. Additionally, we assume that channel long profiles are similar to their configuration at 1617
the last glacial maximum (LGM) and are still in a state of adjustment to deglaciation, consistent 1618

81

with Schlunegger and Norton (2013)’s findings in the Alps. With this in mind, we can look for 1619
relationships between glaciation in a basin and the geometry of downstream channels. We divide 1620
basins into three categories: basins that lie entirely above 4200 meters, the equilibrium line 1621
altitude (ELA) at the LGM in the central Nepal Himalaya (Asahi, 2010) basins that straddle the 1622
LGM ELA, and basins that lie entirely below it (Figure 2A). The location of the physiographic 1623
transition, identified visually, tracks fairly closely with the location of the transition between 1624
basins entirely below the ELA and basins which straddle it. This is indicative of the importance 1625
of GLOFs in controlling the geometry of the Himalayas, and glaciated mountain ranges in 1626
general. Selecting a larger upper area limit (we use 100 km
2
) for our basin analysis could 1627
produce an even better fit between the mapped PT and the southern boundaries of glaciated 1628
basins. 1629
Figure 8. Cumulative
distribution of ksn
differences between
trunk (Strahler order 4+)
and tributary (order 1
and 2) streams for
channels in glaciated
versus unglaciated
basins. Basins are
considered glaciated if
there exist channels
above 4200 meters (the
last glacial maximum
equilibrium line altitude).

1630

82

While we have discussed that comparing ksn between rivers is not an ideal method to 1631
infer relative incision rates where not all rivers have geometry controlled by the same process, 1632
ksn can still be a useful tool. In a region where the evolution of all rivers is accurately described 1633
by the detachment-limited stream power law, Comparing ksn in tributaries versus the trunk 1634
streams into which they drain may reveal where river incision is being controlled by different 1635
processes at different scales. We find several interesting relationships between the extent of 1636
glaciation in a basin and the relationships between downstream channels. We see a strong 1637
positive correlation between the amount of glaciated drainage area a river captures, and, thus, the 1638
frequency of GLOFs it experiences, and the ratio between the tributary and trunk ksn (Figure 9). 1639
Furthermore, if we examine glaciated versus nonglaciated basins as a whole, we see greater 1640
disparity in tributary and trunk ksn in glaciated basins (Figure 10). In these patterns, we can see 1641
steepening of tributary channels in response to increasingly rapid incision driven by frequent 1642
GLOFs in the rivers capturing the most glacial valleys. These large, north-south draining rivers, 1643
many with headwaters in Tibet, see more and larger outburst floods, and may be setting base 1644
level for the smaller basins that drain into them (Cook et al., 2018; Lang et al., 2013). Alongside 1645
the channel width trend illustrated in Figure 4, this suggests local base level in the High 1646
Himalaya is controlled by the frequency of GLOFs. If rates of incision in main-stem channels are 1647
increasing as a mountain belt uplifts and larger areas become glaciated, nonglaciated catchments 1648
will either steepen to keep pace, or will be unable to, and relict topography will develop as a 1649
consequence. We expect that in regions of threshold hillslopes and rapid incision driven by 1650
GLOFs, landsliding rates should be high, but GLOFs should be effective in sweeping it 1651
downstream. 1652

83

A chi map of the study area hints at drainage reorganization driven by rapid incision by 1653
rivers that capture large amounts of glaciated drainage area and thus experience a high frequency 1654
of GLOFs. Chi disparities across major north-south divides may reflect the relative incision rates 1655
of the mainstem channels, mediated by the steepening of tributary basins in response to rapid 1656
incision. In steep, tectonically active regions, earthquake and storm-triggered landslides may be 1657
effective drivers of drainage divide migration, transmitting the signal of incision all the way up 1658
to the divides (Dahlquist et al., 2018). Two major rivers with headwaters in Tibet appear to have 1659
subcatchments which are capturing drainage area from adjacent subcatchments of rivers 1660
capturing less glaciated drainage area, based on lower chi values for the basins draining to the 1661
rivers with Tibetan headwaters (Figures 2F, 11). This is consistent with our hypothesis that rapid 1662
incision driven by GLOFs is controlling the pace of erosion in subcatchments, and headwater 1663
channels are responding by steepening to keep pace, and furthermore, this process may drive 1664
drainage reorganization. 1665
Figure 11. Chi plots for
channels shown in map
in Figure 2F. If two
channels share a divide,
the channel with lower
chi is expected to be
gaining drainage area.
If uplift, erodibility,
climate, and dominant
incision mechanisms in
the two basins are
relatively equal, we
expect the blue network
to gain area at the
expense of the orange.
Reference concavity
was set at 0.525 for this
calculation, the best fit value for these two channel networks.
1666

84

3.2.4 Debris Flows and Outburst Floods Cooperate to Produce High 1667
Topography 1668
Figure 12. Basin ksn
(top) and a1 (bottom)
versus detrital (blue)
and bedrock (red)
apatite (U-Th)/He
ages. Older ages
represent slower
exhumation rates,
dependent on the
geothermal gradient.
Error bars represent
one standard deviation
among the ages.
Thermochronologic
data are from
Quackenbush et al.,
2019.

In the context of a coupled threshold hillslope-channel model, where a chain of processes 1669
whose behavior is governed by various topographic thresholds controls the response of a 1670
landscape to various forcings (Benda et al., 2005; Church, 2002; Montgomery, 1999), these 1671

85

observations hint at long-term trends in the surface evolution in an orogen. In regions where 1672
hillslopes are at or near threshold angles, and erosion in response to uplift is generally controlled 1673
by an increase in landsliding rate rather than further steepening (Larsen and Montgomery, 2012), 1674
channels will typically steepen to accommodate the influx of debris from hillslopes and the 1675
continual lowering of base level likely to be occurring in such an active landscape (Quackenbush 1676
et al., in review). In a case where a transient increase in uplift rate has occurred and the 1677
landscape is in adjustment, channel steepening, like hillslope steepening, will occur up to a point 1678
where debris flows can drive incision in all channels throughout the basin (Stock and Dietrich, 1679
2003; 2006, Penserini et al., 2017). Above the threshold angle for debris flow incision, 1680
increasing coupled uplift-erosion may be accommodated in channels by increasing debris flow 1681
frequency, accompanying the increase in landsliding rate in basins where hillslopes have largely 1682
reached threshold angles. This process of channels steepening to slopes susceptible to debris 1683
flow incision is evidenced by decrease in the a1 metric as well as the density of channels with 1684
slopes exceeding 11º across the physiographic transition (Figures 2D and 2G). a1 describes the 1685
curved region in a slope-area plot, where colluvial channels transition downstream to fluvial 1686
channels (Stock and Dietrich, 2003). A decrease in this metric reflects an increase in the range of 1687
contributing drainage areas where we expect to see debris flow incision as the dominant process. 1688
We observe that debris flows in the study area frequently initiate with very small contributing 1689
drainage areas, as low as 10,000 m
2
(Dahlquist and West, in revision). Therefore, channels in the 1690
region of slope-area space typically associated with colluvial processes may experience debris 1691
flow incision, and do not need to be excluded from the analysis, desensitizing calculation of 1692
these metrics to the minimum drainage area. We use a value of 0.48 km
2
(Roback et al., 2018). 1693
Basin-averaged ksn generally increases across the transition, but less sharply than when viewed 1694

86

through metrics tailored to debris flow erosion. However, while absolute ksn values are less 1695
enlightening, differences between ksn in tributaries and trunk streams tell an interesting story. As 1696
in the case of the a1 metric and the prevalence of threshold streams, the distribution of basin- 1697
averaged ksn differences reflects the transition between the lesser and High Himalaya. In the 1698
High Himalaya, we expect local base level to be controlled by the rate of incision from GLOFs 1699
in the main north-south rivers, and where this incision rate is faster, basins emptying into these 1700
large north-south rivers respond by steepening until debris flow incision is effective through 1701
larger portions of the basin’s area. If downstream portions of the basin are still below the 1702
threshold slope for debris flow incision, steeper channels may still help debris flows to transport 1703
more material out of the subcatchment, and the increase in ksn can help to evacuate sediment and 1704
expose the bed for incision by fluvial erosion. Differences between tributary and trunk stream 1705
ksn, not only in the Nepal High Himalaya but in the Lesser Himalaya as well, reveal this 1706
behavior. In the Lesser Himalaya, where channels landslides and debris flows occur less 1707
frequently than in the High Himalaya, differences between ksn in tributaries and trunk streams are 1708
quite high (Figures 2B, 2C), despite evidence that overall uplift and incision are quite rapid 1709
(Lavé and Avouac, 2001). This evidences a contrast where debris flow processes control the 1710
geometry in low-order tributaries, where detachment-limited fluvial processes still predominate 1711
in the trunk streams. Farther north, where GLOF incision is more effective in the major 1712
drainages and coupled uplift-exhumation is higher overall, ksn values in tributaries and trunk 1713
streams become more similar. Steepening of the subcatchment channels facilitates increasing 1714
landsliding rates, and the influx of landslide material is mobilized in increasing numbers of 1715
debris flows. Inferences made from the topography seem to be reflective of long-term 1716
exhumation as well. We examined detrital and bedrock apatite (U-Th)/He ages from 1717

87

Quackenbush et al., (in review) in the context of the geometry of the sampled basins and found 1718
although bedrock samples had poor correlation, detrital ages correlated well with a1 for the 1719
basins they averaged (Figure 12). The overall weaker correlation between thermochronometric 1720
ages and basin ksn is indicative of the dominance of debris flow processes versus detachment- 1721
limited in the sampled subcatchments. 1722
1723
3.3 CONCLUSIONS 1724
We find a variety of evidence to support the idea that the evolution of the channel 1725
network in the Nepal Himalaya is governed by the prevalence of GLOFs, which set the pace of 1726
incision in major north-south rivers. The response to GLOF incision, which manifests essentially 1727
as a decrease in base level of tributaries, is transmitted up sub-catchments by the steepening of 1728
tributary channels. This is steepening response is predicted in fluvial reaches by the stream 1729
power law, but it also allows debris flows initiating in the headwaters to propagate incision 1730
further downstream. The threshold hillslope model establishes that landsliding rate increases to 1731
accommodate accelerated rates of coupled uplift-erosion in regions where hillslope angle has 1732
steepened to a threshold value determined by material strength. If there is no evidence of 1733
extensive alluviation in threshold hillslope basins where landsliding rates can be assumed to be 1734
high, and fluvial processes cannot account for the evacuation of the large boulders and other 1735
coarse sediment landslides can deliver, debris flows must also increase in frequency and/or 1736
magnitude alongside landslides. From an inventory of debris flows associated with the 2015 1737
Gorkha Earthquake, we identify a threshold channel bed slope at ~11° above which channel 1738
geometry appears to be controlled primarily by debris flow incision, where at lower angles other 1739

88

processes, including fluvial erosion described by the detachment-limited stream power model, 1740
dominate. 1741
We identify several trends that indicate that the location of the physiographic transition in 1742
the Nepal Himalaya is controlled by transitions in dominant process regimes. In particular, the 1743
abrupt decrease in basin a1 north of the physiographic transition, coupled with the increase in the 1744
prevalence of channels which exceed the ~11° threshold we identify for effective debris flow 1745
incision indicates that a dramatic increase in the effectiveness of debris flow incision is an 1746
important factor in the transition between the Lesser and High Himalaya. a1 values also increase 1747
with detrital apatite (U-Th)/He ages, indicating that trends in long-term exhumation rates are at 1748
least partially accommodated by an increase in the frequency and/or effectiveness of debris flow 1749
incision. The distribution of basin-averaged ksn differences between tributary and trunk streams 1750
also reflects the transition. In the High Himalaya, outside of major north-south rivers with 1751
glaciated headwaters, debris flows are the predominant means of incision in both the tributaries 1752
and the mainstems, and ksn values are similar. South of the transition, trunk streams are 1753
increasingly controlled by fluvial processes, but headwater streams remain debris flow 1754
controlled. We also find a systematic increase in ksn differences with the prevalence of glaciated 1755
drainage area in the trunk stream. Since GLOFs can be powerful drivers of incision in rivers with 1756
low gradients, this is evidence of debris flow dominated tributary catchments steepening to keep 1757
up with increasing frequency of GLOFs in the mainstem channels. GLOFs also appear to have an 1758
important role in controlling the location of the physiographic transition, as the location in the 1759
Bhote Koshi river above which GLOFs represent anomalously high discharge corresponds 1760
closely with the location of the PT. We also find evidence of a possible inverse correlation 1761
between normalized channel wideness index and the amount of glaciated drainage area a river 1762

89

captures. This may be a further indicator of enhanced incision in river channels experiencing 1763
more frequent GLOFs. Channel width generally increases due to aggradation, and aggradation 1764
inhibits erosion by armoring the bedrock, so channels that have the erosive power to rapidly clear 1765
out accumulating sediment and expose the channel bed are likely to experience more rapid 1766
incision. Broadly, these observations combine to form the framework of a “threshold channel 1767
model” that carries the principles of the threshold hillslope model downstream. The pace of 1768
GLOF incision in mainstem rivers with glaciated headwaters drives steepening of tributary 1769
basins, leading to increasing debris flow dominance in these watersheds. Like hillslopes, debris 1770
flow channels show threshold behavior, having a threshold bed slope (~11˚) above which debris 1771
flow incision is effective and below which it is not. Whereas increasing stream power in the 1772
fluvial domain is represented in slope-area space by an increase in the steepness of the channel, 1773
the corresponding increase in debris flow erosion is measured instead by how far it can 1774
propagate downstream. For this reason, we conclude that, in a basin where channels have 1775
achieved threshold slopes for debris flow incision, increased coupled uplift-erosion is 1776
accommodated by an increase in the frequency of debris flows. This is logically consistent with 1777
the analogous response of threshold hillslopes experiencing increased landslide rates to 1778
accommodate increased uplift-erosion, as these landslides deliver the sediment necessary to 1779
supply these debris flows and drive fluvial incision. Therefore, both in mainstem rivers with 1780
glaciated headwaters and in their debris flow-dominated tributary basins, we find strong 1781
evidence to support top-down controls on landscape evolution in glaciated mountain belts. 1782
1783
1784

90

3.4 METHODS 1785
Mapping of pre- and post-seismic debris flows and extraction of channel data was 1786
conducted as described in Dahlquist and West, in review. We used the Shuttle Radar Topography 1787
Mission (SRTM) 30-meter digital elevation model (DEM) for topographic analysis. Voids in 1788
SRTM data were patched with the Advanced Spaceborne Thermal Emission and Reflection 1789
Radiometer (ASTER) 30-meter DEM. All topographic analysis uses this patched DEM. We 1790
calculated ksn, and chi using the Topotools toolkit for Matlab (Forte and Whipple, 2018), and 1791
used the TopoToolbox 2 toolkit for processing the DEM, defining streams, and calculating 1792
channel gradients (Schwanghart and Scherler, 2014). We used the Matlab function “nlinfit” to fit 1793
a power law function to the channel width versus discharge relationship. For comparisons 1794
between tributary and trunk stream ksn, we used the ArcGIS function “join”. For analyses 1795
involving basin-wide metrics, we selected drainage basins having drainage areas between 10 and 1796
100 km
2
that are tributaries to rivers capturing >100 km
2
. We found this range of drainage areas 1797
generated maps that covered the area of interest without leaving too large gaps, while containing 1798
basins that exhibited the suite of process transitions we aimed to examine along the length of 1799
their channels. For ksn and chi calculations, we use a minimum contributing drainage area to 1800
define streams of 480,000 m
2
(Roback et al., 2018), and a reference concavity of 0.35, close to 1801
the best-fit value for the study area, and a smoothing distance of 1000 meters. Choices for 1802
reference concavity can vary by ~± 0.2 from the best-fit value without adversely affecting the 1803
calculation. We assigned a regional base of 300 meters above sea level for the chi calculation. 1804
For our a1 calculation, we find a best-fit value for a2 in the study area of 0.62, based on the 1805
average of the values for all basins. We allow S0 to vary among basins, as a1 is not expected to 1806
covary with the initial valley slope (Penserini et al., 2017). Our calculations may not capture the 1807

91

transition between hillslopes and colluvial valleys, as topography begins to show convergent 1808
geometry (defining the position of unchanneled valley heads) at drainage areas smaller than the 1809
area of a single DEM cell. We used Tropical Rainfall Measurement Mission (TRMM) data 1810
averaging mean annual precipitation between 1998 and 2009 (Bookhagen, in review; 1811
Kummerow et al., 1998). Using modern precipitation data is justified in this case, as width of the 1812
active channel can adjust rapidly (at the time scale of a single significant flood) and so should 1813
reflect modern discharge. While TRMM data are not perfect, and in particular may 1814
underestimate light precipitation, this represents a substantial improvement for our normalized 1815
channel wideness analysis over simply using contributing drainage area as a proxy for discharge. 1816
Channel width was measured in 3,525 locations along major north-south draining rivers with 1817
headwaters containing glaciated terrain using Google Earth imagery. We attempted to avoid any 1818
taking measurements where channel width had been recently affected by landslides or debris 1819
flows, and in locations where these processes have recently been active, we used historical 1820
imagery predating the events. All colormaps in figures use quantile schemes. 1821
1822
3.5 REFERENCES 1823
Allen, G. H., J. B. Barnes, T. M. Pavelsky, and E. Kirby, 2013, Lithologic and tectonic controls 1824
on bedrock channel form at the northwest Himalayan front: Journal of Geophysical 1825
Research: Earth Surface, v. 118, no. 3, p. 1806–1825, doi:10.1002/jgrf.20113. 1826
1827
Amatya, K. M., and B. M. Jnawali, 1994, Geological map of Nepal, scale 1:1,000,000, Dep. of 1828
Mines and Geol., Gov. of Nepal, Kathmandu. 1829
1830
Asahi, K., 2010, Equilibrium-line altitudes of the present and Last Glacial Maximum in the 1831
eastern Nepal Himalayas and their implications for SW monsoon climate: Quaternary 1832
International, v. 212, no. 1, p. 26–34, doi:10.1016/j.quaint.2008.08.004. 1833
1834
1835
1836

92

Attal, M., and J. Lavé, 2006, Changes of bedload characteristics along the Marsyandi River 1837
(central Nepal): Implications for understanding hillslope sediment supply, sediment load 1838
evolution along fluvial networks, and denudation in active orogenic belts, in Special Paper 1839
398: Tectonics, Climate, and Landscape Evolution: Geological Society of America, p. 143– 1840
171, doi:10.1130/2006.2398(09). 1841
1842
Baer, P., C. Huggel, B. W. McArdell, and F. Frank, 2017, Changing debris flow activity after 1843
sudden sediment input: a case study from the Swiss Alps: Geology Today, v. 33, no. 6, p. 1844
216–223, doi:10.1111/gto.12211. 1845
1846
Barnard, P. L., L. A. Owen, R. C. Finkel, and K. Asahi, 2006, Landscape response to 1847
deglaciation in a high relief, monsoon-influenced alpine environment, Langtang Himal, 1848
Nepal: Quaternary Science Reviews, v. 25, no. 17–18, p. 2162–2176, 1849
doi:10.1016/j.quascirev.2006.02.002. 1850
1851
Bathurst, J. C., and M. Ashiq, 1998, Dambreak flood impact on mountain stream bedload 1852
transport after 13 years: Earth Surface Processes and Landforms, v. 23, no. 7, p. 643–649, 1853
doi:10.1002/(SICI)1096-9837(199807)23:7<643::AID-ESP889>3.0.CO;2-3. 1854
1855
Baynes, E. R. C., M. Attal, S. Niedermann, L. A. Kirstein, A. J. Dugmore, and M. Naylor, 2015, 1856
Erosion during extreme flood events dominates Holocene canyon evolution in northeast 1857
Iceland: Proceedings of the National Academy of Sciences, v. 112, no. 8, p. 2355–2360, 1858
doi:10.1073/pnas.1415443112. 1859
1860
Benda, L., 1990, The influence of debris flows on channels and valley floors in the Oregon Coast 1861
Range, U.S.A.: Earth Surface Processes and Landforms, v. 15, no. 5, p. 457–466, 1862
doi:10.1002/esp.3290150508. 1863
1864
Benda, L., and T. Dunne, 1997, Stochastic forcing of sediment supply to channel networks from 1865
landsliding and debris flow: Water Resources Research, v. 33, no. 12, p. 2849–2863, 1866
doi:10.1029/97WR02388. 1867
1868
Benda, L., M. A. Hassan, M. Church, and C. L. May, 2005, Geomorphology of Steepland 1869
Headwaters: The Transition From Hillslopes to Channels: Journal of the American Water 1870
Resources Association v. 41 p. 835-851. 1871
1872
Bookhagen, B., in review, High resolution spatiotemporal distribution of rainfall seasonality and 1873
extreme events based on a 12-year TRMM time series. 1874
1875
Bookhagen, B., and D. W. Burbank, 2010, Toward a complete Himalayan hydrological budget: 1876
Spatiotemporal distribution of snowmelt and rainfall and their impact on river discharge: 1877
Journal of Geophysical Research, v. 115, no. F3, doi:10.1029/2009JF001426. 1878
1879
Bovis, M. J., and M. Jakob, 1999, The role of debris supply conditions in predicting debris flow 1880
activity: Earth Surface Processes and Landforms, v. 24, no. 11, p. 1039–1054, 1881
doi:10.1002/(SICI)1096-9837(199910)24:11<1039::AID-ESP29>3.0.CO;2-U. 1882

93

Brayshaw, D., and M. A. Hassan, 2009, Debris flow initiation and sediment recharge in gullies: 1883
Geomorphology, v. 109, no. 3–4, p. 122–131, doi:10.1016/j.geomorph.2009.02.021. 1884
1885
Burbank, D. W., A. E. Blythe, J. Putkonen, B. Pratt-Sitaula, E. Gabet, M. Oskin, A. Barros, and 1886
T. P. Ojha, 2003, Decoupling of erosion and precipitation in the Himalayas: Nature, v. 1887
426, no. 6967, p. 652–655, doi:10.1038/nature02187. 1888
1889
Burbank, D. W., J. Leland, E. Fielding, R. S. Anderson, N. Brozovic, M. R. Reid, and C. 1890
Duncan, 1996, Bedrock incision, rock uplift and threshold hillslopes in the northwestern 1891
Himalayas: Nature, v. 379, no. 6565, p. 505–510, doi:10.1038/379505a0. 1892
1893
Chen, Yaning, C. Xu, Yapeng Chen, W. Li, and J. Liu, 2010, Response of glacial-lake outburst 1894
floods to climate change in the Yarkant River basin on northern slope of Karakoram 1895
Mountains, China: Quaternary International, v. 226, no. 1–2, p. 75–81, 1896
doi:10.1016/j.quaint.2010.01.003. 1897
1898
Church, M., 2002, Geomorphic thresholds in riverine landscapes: Freshwater Biology, v. 47, no. 1899
4, p. 541–557, doi:10.1046/j.1365-2427.2002.00919.x. 1900
1901
Cook, K. L., C. Andermann, F. Gimbert, B. R. Adhikari, and N. Hovius, 2018, Glacial lake 1902
outburst floods as drivers of fluvial erosion in the Himalaya: Science, v. 362, no. 6410, p. 1903
53-57. doi:10.1126/science.aat4981. 1904
1905
Cook, K. L., J. M. Turowski, and N. Hovius, 2013, A demonstration of the importance of 1906
bedload transport for fluvial bedrock erosion and knickpoint propagation: Earth Surface 1907
Processes and Landforms, v. 38, no. 7, p. 683–695, doi:10.1002/esp.3313. 1908
1909
Costa, J. E., and R. L. Schuster, 1988, The formation and failure of natural dams: Geological 1910
Society of America Bulletin, v. 100, no. 7, p. 1054–1068, doi:10.1130/0016- 1911
7606(1988)100<1054:TFAFON>2.3.CO;2. 1912
1913
Croissant, T., D. Lague, P. Steer, and P. Davy, 2017, Rapid post-seismic landslide evacuation 1914
boosted by dynamic river width: Nature Geoscience, v. 10, no. 9, p. 680–684, 1915
doi:10.1038/ngeo3005. 1916
1917
Dahlquist, M. P., A. J. West, and G. Li, 2018, Landslide-driven drainage divide migration: 1918
Geology, v. 46, no. 5, p. 403–406, doi:10.1130/G39916.1. 1919
1920
Dahlquist, M. P. and West, A. J., in revision, Initiation and runout of post-seismic debris flows: 1921
Insights from the 2015 Gorkha earthquake. 1922
1923
Dhital, M. R., 2015, Geology of the Nepal Himalaya: Regional perspective of the classic collided 1924
orogen: Geol. Nepal Himalaya Reg. Perspect. Class. Collided Orogen, 1–498, 1925
doi:10.1007/978-3-319-02496-7 1926
1927

94

DiBiase, R. A., K. X. Whipple, A. M. Heimsath, and W. B. Ouimet, 2010, Landscape form and 1928
millennial erosion rates in the San Gabriel Mountains, CA: Earth and Planetary Science 1929
Letters, v. 289, no. 1–2, p. 134–144, doi:10.1016/j.epsl.2009.10.036. 1930
1931
Dunning, S. A., A. R. G. Large, A. J. Russell, M. J. Roberts, R. Duller, J. Woodward, A.-S. 1932
Meriaux, F. S. Tweed, and M. Lim, 2013, The role of multiple glacier outburst floods in 1933
proglacial landscape evolution: The 2010 Eyjafjallajokull eruption, Iceland: Geology, v. 1934
41, no. 10, p. 1123–1126, doi:10.1130/G34665.1. 1935
1936
Egholm, D. L., S. B. Nielsen, V. K. Pedersen, and J.-E. Lesemann, 2009, Glacial effects limiting 1937
mountain height: Nature, v. 460, no. 7257, p. 884–887, doi:10.1038/nature08263. 1938
1939
Emmer, A., 2017, Geomorphologically effective floods from moraine-dammed lakes in the 1940
Cordillera Blanca, Peru: Quaternary Science Reviews, v. 177, p. 220–234, 1941
doi:10.1016/j.quascirev.2017.10.028. 1942
1943
Fan, L., P. Lehmann, B. McArdell, and D. Or, 2017, Linking rainfall-induced landslides with 1944
debris flows runout patterns towards catchment scale hazard assessment: Geomorphology, 1945
v. 280, p. 1–15, doi:10.1016/j.geomorph.2016.10.007. 1946
1947
Flint, J. J., 1974, Stream gradient as a function of order, magnitude, and discharge: Water 1948
Resources Research, v. 10, no. 5, p. 969–973, doi:10.1029/WR010i005p00969. 1949
1950
Forte, A. M., and K. X. Whipple, 2018, Criteria and tools for determining drainage divide 1951
stability: Earth and Planetary Science Letters, v. 493, p. 102–117, 1952
doi:10.1016/j.epsl.2018.04.026. 1953
1954
Harvey, A., 2001, Coupling between hillslopes and channels in upland fluvial systems: 1955
implications for landscape sensitivity, illustrated from the Howgill Fells, northwest 1956
England: Catena, v. 42, no. 2-4, p. 225–250, doi:10.1016/s0341-8162(00)00139-9. 1957
1958
Hodges, K. V., J. M. Hurtado, and K. X. Whipple, 2001, Southward extrusion of Tibetan crust 1959
and its effect on Himalayan tectonics: Tectonics, v. 20, no. 6, p. 799–809, 1960
doi:10.1029/2001TC001281. 1961
1962
Hodges, K. V., C. Wobus, K. Ruhl, T. Schildgen, and K. Whipple, 2004, Quaternary 1963
deformation, river steepening, and heavy precipitation at the front of the Higher 1964
Himalayan ranges: Earth and Planetary Science Letters, v. 220, no. 3–4, p. 379–389, 1965
doi:10.1016/S0012-821X(04)00063-9. 1966
1967
Howard, A. D., and Kerby, G., 1983, Channel changes in badlands: Geological Society of 1968
America Bulletin, v. 94 p. 739-752. 1969
1970
Hungr, O., 1995, A model for the runout analysis of rapid flow slides, debris flows, and 1971
avalanches: Canadian Geotechnical Journal, v. 32, no. 4, p. 610–623, doi:10.1139/t95-063. 1972
1973

95

Hungr, O., S. McDougall, M. Wise, and M. Cullen, 2008, Magnitude–frequency relationships of 1974
debris flows and debris avalanches in relation to slope relief: Geomorphology, v. 96, no. 3– 1975
4, p. 355–365, doi:10.1016/j.geomorph.2007.03.020. 1976
1977
Iverson, R. M., M. E. Reid, and R. G. LaHusen, 1997, Debris-flow mobilization from landslides: 1978
Annual Review of Earth and Planetary Sciences, v. 25, no. 1, p. 85–138, 1979
doi:10.1146/annurev.earth.25.1.85. 1980
1981
Jansen, J. D., D. Fabel, P. Bishop, S. Xu, C. Schnabel, and A. T. Codilean, 2011, Does 1982
decreasing paraglacial sediment supply slow knickpoint retreat? Geology, v. 39, no. 6, p. 1983
543–546, doi:10.1130/G32018.1. 1984
1985
Kirby, E., and K. Whipple, 2001, Quantifying differential rock-uplift rates via stream profile 1986
analysis: Geology, v. 29, no. 5, p. 415, doi:10.1130/0091- 1987
7613(2001)029<0415:QDRURV>2.0.CO;2. 1988
1989
Kirby, E., and K. X. Whipple, 2012, Expression of active tectonics in erosional landscapes: 1990
Journal of Structural Geology, v. 44, p. 54–75, doi:10.1016/j.jsg.2012.07.009. 1991
1992
Kirchner, J. W., R. C. Finkel, C. S. Riebe, D. E. Granger, J. L. Clayton, J. G. King, and W. F. 1993
Megahan, 2001, Mountain erosion over 10 yr, 10 k.y., and 10 m.y. time scales: Geology, v. 1994
29, no. 7, p. 591, doi:10.1130/0091-7613(2001)029<0591:MEOYKY>2.0.CO;2. 1995
1996
Korup, O., A. L. Densmore, and F. Schlunegger, 2010, The role of landslides in mountain range 1997
evolution: Geomorphology, v. 120, no. 1–2, p. 77–90, 1998
doi:10.1016/j.geomorph.2009.09.017. 1999
2000
Korup, O., and F. Tweed, 2007, Ice, moraine, and landslide dams in mountainous terrain: 2001
Quaternary Science Reviews, v. 26, no. 25–28, p. 3406–3422, 2002
doi:10.1016/j.quascirev.2007.10.012. 2003
2004
Kummerow, C., W. Barnes, T. Kozu, J. Shiue, and J. Simpson, 1998, The Tropical Rainfall 2005
Measuring Mission (TRMM) Sensor Package: Journal of Atmospheric and Oceanic 2006
Technology, v. 15, no. 3, p. 809–817, doi:10.1175/1520- 2007
0426(1998)015<0809:TTRMMT>2.0.CO;2. 2008
2009
Lamb, M. P., W. E. Dietrich, and L. S. Sklar, 2008, A model for fluvial bedrock incision by 2010
impacting suspended and bed load sediment: Journal of Geophysical Research, v. 113, no. 2011
F3, doi:10.1029/2007JF000915. 2012
2013
Lang, K. A., K. W. Huntington, and D. R. Montgomery, 2013, Erosion of the Tsangpo Gorge by 2014
megafloods, Eastern Himalaya: Geology, v. 41, no. 9, p. 1003–1006, 2015
doi:10.1130/G34693.1. 2016
2017
Larsen, I. J., and D. R. Montgomery, 2012, Landslide erosion coupled to tectonics and river 2018
incision: Nature Geoscience, v. 5, no. 7, p. 468–473, doi:10.1038/ngeo1479. 2019

96

Lavé, J., and J. P. Avouac, 2001, Fluvial incision and tectonic uplift across the Himalayas of 2020
central Nepal: Journal of Geophysical Research: Solid Earth, v. 106, no. B11, p. 26561– 2021
26591, doi:10.1029/2001JB000359. 2022
2023
Lemonnier, C., G. Marquis, F. Perrier, J.-P. Avouac, G. Chitrakar, B. Kafle, S. Sapkota, U. 2024
Gautam, D. Tiwari, and M. Bano, 1999, Electrical structure of the Himalaya of central 2025
Nepal: High conductivity around the mid-crustal ramp along the MHT: Geophysical 2026
Research Letters, v. 26, no. 21, p. 3261–3264, doi:10.1029/1999GL008363. 2027
2028
Li, G., A. J. West, A. L. Densmore, Z. Jin, R. N. Parker, and R. G. Hilton, 2014, Seismic 2029
mountain building: Landslides associated with the 2008 Wenchuan earthquake in the 2030
context of a generalized model for earthquake volume balance: Geochemistry, Geophysics, 2031
Geosystems, v. 15, no. 4, p. 833–844, doi:10.1002/2013GC005067. 2032
2033
Li, G., A. J. West, A. L. Densmore, Z. Jin, F. Zhang, J. Wang, M. Clark, and R. G. Hilton, 2017, 2034
Earthquakes drive focused denudation along a tectonically active mountain front: Earth and 2035
Planetary Science Letters, v. 472, p. 253–265, doi:10.1016/j.epsl.2017.04.040. 2036
2037
Livers, B., and E. Wohl, 2015, An evaluation of stream characteristics in glacial versus fluvial 2038
process domains in the Colorado Front Range: Geomorphology, v. 231, p. 72–82, 2039
doi:10.1016/j.geomorph.2014.12.003. 2040
2041
Marc, O., R. Behling, C. Andermann, J. M. Turowski, L. Illien, S. Roessner, and N. Hovius, 2042
2019, Long-term erosion of the Nepal Himalayas by bedrock landsliding: the role of 2043
monsoons, earthquakes and giant landslides: Earth Surface Dynamics, v. 7, no. 1, p. 107– 2044
128, doi:10.5194/esurf-7-107-2019. 2045
2046
Mitchell, S. G., and D. R. Montgomery, 2006, Influence of a glacial buzzsaw on the height and 2047
morphology of the Cascade Range in central Washington State, USA: Quaternary Research, 2048
v. 65, no. 01, p. 96–107, doi:10.1016/j.yqres.2005.08.018. 2049
2050
Montgomery, D. R., 1999, Process domains and the river continuum: Journal of the American 2051
Water Resources Association, v. 35, no. 2, p. 397–410, doi:10.1111/j.1752- 2052
1688.1999.tb03598.x. 2053
2054
Montgomery, D. R., 2001, Slope Distributions, Threshold Hillslopes, and Steady-state 2055
Topography: American Journal of Science, v. 301, no. 4–5, p. 432–454, 2056
doi:10.2475/ajs.301.4-5.432. 2057
2058
Montgomery, D. R., and M. T. Brandon, 2002, Topographic controls on erosion rates in 2059
tectonically active mountain ranges: Earth and Planetary Science Letters, v. 201, no. 3–4, p. 2060
481–489, doi:10.1016/S0012-821X(02)00725-2. 2061
2062
Montgomery, D. R., and W. E. Dietrich, 1994, A physically based model for the topographic 2063
control on shallow landsliding: Water Resources Research, v. 30, no. 4, p. 1153–1171, 2064
doi:10.1029/93WR02979. 2065

97

Montgomery, D. R., and E. Foufoula-Georgiou, 1993, Channel network source representation 2066
using digital elevation models: Water Resources Research, v. 29, no. 12, p. 3925–3934, 2067
doi:10.1029/93WR02463. 2068
2069
Mool, P. K., 1995, Glacier lake outburst floods in Nepal: Journal of Nepal Geological Society v. 2070
11 p. 273-280. 2071
2072
Ouimet, W. B., K. X. Whipple, and D. E. Granger, 2009, Beyond threshold hillslopes: Channel 2073
adjustment to base-level fall in tectonically active mountain ranges: Geology, v. 37, no. 7, p. 2074
579–582, doi:10.1130/G30013A.1. 2075
2076
Penserini, B. D., J. J. Roering, and A. Streig, 2017, A morphologic proxy for debris flow erosion 2077
with application to the earthquake deformation cycle, Cascadia Subduction Zone, USA: 2078
Geomorphology, v. 282, p. 150–161, doi:10.1016/j.geomorph.2017.01.018. 2079
2080
Perron, J. T., and L. Royden, 2013, An integral approach to bedrock river profile analysis: Earth 2081
Surface Processes and Landforms, v. 38, no. 6, p. 570–576, doi:10.1002/esp.3302. 2082
2083
Pratt, B., D. W. Burbank, A. Heimsath, and T. Ojha, 2002, Impulsive alluviation during early 2084
Holocene strengthened monsoons, central Nepal Himalaya: Geology, v. 30, no. 10, p. 911, 2085
doi:10.1130/0091-7613(2002)030<0911:IADEHS>2.0.CO;2. 2086
2087
Pratt-Sitaula, B., D. W. Burbank, A. Heimsath, and T. Ojha, 2004, Landscape disequilibrium on 2088
1000–10,000 year scales Marsyandi River, Nepal, central Himalaya: Geomorphology, v. 58, 2089
no. 1–4, p. 223–241, doi:10.1016/j.geomorph.2003.07.002. 2090
2091
Roback, K., M. K. Clark, A. J. West, D. Zekkos, G. Li, S. F. Gallen, D. Chamlagain, and J. W. 2092
Godt, 2018, The size, distribution, and mobility of landslides caused by the 2015 M w 7.8 2093
Gorkha earthquake, Nepal: Geomorphology, v. 301, p. 121–138, 2094
doi:10.1016/j.geomorph.2017.01.030. 2095
2096
Roe, G. H., D. R. Montgomery, and B. Hallet, 2003, Orographic precipitation and the relief of 2097
mountain ranges: Journal of Geophysical Research: Solid Earth, v. 108, no. B6, 2098
doi:10.1029/2001JB001521. 2099
2100
Ruhl, K. W., and K. V. Hodges, 2005, The use of detrital mineral cooling ages to evaluate steady 2101
state assumptions in active orogens: An example from the central Nepalese Himalaya: 2102
Tectonics, v. 24, no. 4, p. n/a-n/a, doi:10.1029/2004TC001712. 2103
2104
Sakai, H., R. Fujii, and Y. Kuwahara, 2002, Changes in the depositional system of the Paleo- 2105
Kathmandu Lake caused by uplift of the Nepal Lesser Himalayas: Journal of Asian Earth 2106
Sciences, v. 20, no. 3, p. 267–276, doi:10.1016/S1367-9120(01)00046-3. 2107
2108
Schlunegger, F., and K. P. Norton, 2013, Water versus ice: The competing roles of modern 2109
climate and Pleistocene glacial erosion in the Central Alps of Switzerland: Tectonophysics, 2110
v. 602, p. 370–381, doi:10.1016/j.tecto.2013.03.027. 2111

98

Schumm, S. A., 1993, River Response to Baselevel Change: Implications for Sequence 2112
Stratigraphy: The Journal of Geology, v. 101, no. 2, p. 279–294, doi:10.1086/648221. 2113
2114
Schwanghart, W. et al., 2016, Repeated catastrophic valley infill following medieval earthquakes 2115
in the Nepal Himalaya: Science, v. 351, no. 6269, p. 147–150, 2116
doi:10.1126/science.aac9865. 2117
2118
Schwanghart, W., and D. Scherler, 2014, Short Communication: TopoToolbox 2 – MATLAB- 2119
based software for topographic analysis and modeling in Earth surface sciences: Earth 2120
Surface Dynamics, v. 2, no. 1, p. 1–7, doi:10.5194/esurf-2-1-2014. 2121
2122
Seidl, M., and Dietrich, W. E., 1992, The problem of channel incision into bedrock, in Schmidt, 2123
K.-H., and de Ploey, J., eds., Functional geomorphology, Catena Supplement 23: 2124
Cremlingen, Germany, Catena Verlag, p. 101–124. 2125
2126
Seidl, M. A., W. E. Dietrich, and J. W. Kirchner, 1994, Longitudinal Profile Development into 2127
Bedrock: An Analysis of Hawaiian Channels: The Journal of Geology, v. 102, no. 4, p. 2128
457–474, doi:10.1086/629686. 2129
2130
Shobe, C. M., G. E. Tucker, and R. S. Anderson, 2016, Hillslope-derived blocks retard river 2131
incision: Geophysical Research Letters, v. 43, no. 10, p. 5070–5078, 2132
doi:10.1002/2016GL069262. 2133
2134
Sklar, L. S., and W. E. Dietrich, 2001, Sediment and rock strength controls on river incision into 2135
bedrock: Geology, v. 29, no. 12, p. 1087, doi:10.1130/0091- 2136
7613(2001)029<1087:SARSCO>2.0.CO;2. 2137
2138
Sklar, L. S., and W. E. Dietrich, 2004, A mechanistic model for river incision into bedrock by 2139
saltating bed load: Water Resources Research, v. 40, no. 6, doi:10.1029/2003WR002496. 2140
2141
Stock, J. D., and W. E. Dietrich, 2006, Erosion of steepland valleys by debris flows: Geological 2142
Society of America Bulletin, v. 118, no. 9–10, p. 1125–1148. 2143
2144
Stock, J., and W. E. Dietrich, 2003, Valley incision by debris flows: Evidence of a topographic 2145
signature: Water Resources Research, v. 39, no. 4, doi:10.1029/2001WR001057. 2146
2147
Takahashi, T., 1978, Mechanical characteristics of debris flow: Journal of the Hydraulics 2148
Division, v. 104, no. H48, p. 1153-1169. 2149
2150
Thiede, R. C., and T. A. Ehlers, 2013, Large spatial and temporal variations in Himalayan 2151
denudation: Earth and Planetary Science Letters, v. 371–372, p. 278–293, 2152
doi:10.1016/j.epsl.2013.03.004. 2153
2154
Veh, G., O. Korup, S. Roessner, and A. Walz, 2018, Detecting Himalayan glacial lake outburst 2155
floods from Landsat time series: Remote Sensing of Environment, v. 207, p. 84–97, 2156
doi:10.1016/j.rse.2017.12.025. 2157

99

Whipple, K. X., and G. E. Tucker, 1999, Dynamics of the stream-power river incision model: 2158
Implications for height limits of mountain ranges, landscape response timescales, and 2159
research needs: Journal of Geophysical Research: Solid Earth, v. 104, no. B8, p. 17661– 2160
17674, doi:10.1029/1999JB900120. 2161
2162
Willett, S. D., and M. T. Brandon, 2002, On steady states in mountain belts: Geology, v. 30, no. 2163
2, p. 175, doi:10.1130/0091-7613(2002)030<0175:OSSIMB>2.0.CO;2. 2164
2165
Willett, S. D., S. W. McCoy, J. T. Perron, L. Goren, and C.-Y. Chen, 2014, Dynamic 2166
Reorganization of River Basins: Science, v. 343, no. 6175, p. 1248765–1248765, 2167
doi:10.1126/science.1248765. 2168
2169
Wilson, A., N. Hovius, and J. M. Turowski, 2013, Upstream-facing convex surfaces: Bedrock 2170
bedforms produced by fluvial bedload abrasion: Geomorphology, v. 180–181, p. 187–204, 2171
doi:10.1016/j.geomorph.2012.10.010. 2172
2173
Wobus C. Heimsath A. Whipple K. Hodges K. 2005, Active out-of-sequence thrust faulting in 2174
the central Nepalese Himalaya: Nature , v. 434, no. 7036 p. 1008– 1011, doi: 2175
10.1038/nature03499. 2176
2177
Wobus, C. W., K. X. Whipple, and K. V. Hodges, 2006, Neotectonics of the central Nepalese 2178
Himalaya: Constraints from geomorphology, detrital 40 Ar/ 39 Ar thermochronology, and 2179
thermal modeling: Tectonics, v. 25, no. 4, p. n/a-n/a, doi:10.1029/2005TC001935. 2180
2181
Wobus, C., K. X. Whipple, E. Kirby, N. Snyder, J. Johnson, K. Spyropolou, B. Crosby, and D. 2182
Sheehan, 2006, Tectonics from topography: Procedures, promise, and pitfalls, in Special 2183
Paper 398: Tectonics, Climate, and Landscape Evolution: Geological Society of America, p. 2184
55–74, doi:10.1130/2006.2398(04). 2185
2186
Xu, D., 1988, Characteristics of debris flow caused by outburst of glacial lake in Boqu river, 2187
Xizang, China, 1981: GeoJournal, v. 17, no. 4, doi:10.1007/bf00209443. 2188
2189
Yanites, B. J., 2018, The dynamics of channel slope, width, and sediment in actively eroding 2190
bedrock river systems: Journal of Geophysical Research: Earth Surface, v. 123, no. 7, p. 2191
1504–1527, doi:10.1029/2017JF004405. 2192
2193
Yanites, B. J., N. A. Mitchell, J. C. Bregy, G. A. Carlson, K. Cataldo, M. Holahan, G. H. 2194
Johnston, A. Nelson, J. Valenza, and M. Wanker, 2018, Landslides control the spatial and 2195
temporal variation of channel width in southern Taiwan: Implications for landscape 2196
evolution and cascading hazards in steep, tectonically active landscapes: Earth Surface 2197
Processes and Landforms, v. 43, no. 9, p. 1782–1797, doi:10.1002/esp.4353. 2198
2199
Yanites, B. J., G. E. Tucker, H.-L. Hsu, C. Chen, Y.-G. Chen, and K. J. Mueller, 2011, The 2200
influence of sediment cover variability on long-term river incision rates: An example from 2201
the Peikang River, central Taiwan: Journal of Geophysical Research, v. 116, no. F3, 2202
doi:10.1029/2010JF001933. 2203

100

CHAPTER 4: Landslide-driven drainage divide migration 2204
4.1 INTRODUCTION 2205
In regions where topography is actively responding to tectonic or climatic gradients, 2206
extreme events, the processes that drive erosion and deposition also control the evolution of 2207
centers of erosion and deposition in space and time. Put simply, this means that where the 2208
vertical evolution of the land surface is governed by extreme events, over time, the sum of these 2209
events and their effects should drive a landscape toward a configuration where uplift is balanced 2210
by erosion and a topographic steady state may exist. This may involve a steepened tributary 2211
experiencing debris flows at an increased rate to keep pace with rapid incision in a trunk stream, 2212
a river widening to accommodate an increase in sediment load due to landslides, or landslides 2213
cutting back into ridges and allocating increased drainage area to a drainage basin experiencing 2214
more rapid erosion than its neighbors. In this chapter, we examine the third case, where extreme 2215
events are driving drainage reorganization in response to erosion rate gradients. In the study 2216
areas covered here, extreme events, such as major landslide-generating earthquakes and 2217
cyclones, are important erosional agents, but also mediate the long-term response of drainage 2218
divides to tectonics and climate. Their importance to long-term landscape evolution is apparent 2219
even at the timescale of a single earthquake or storm. 2220
4.1.1 Drainage divides and the threshold hillslope model 2221
Fluvial erosion is a primary force shaping most landscapes on Earth (Strahler, 2222
1952; Whipple and Tucker, 1999), countering and influencing uplift by mobilizing and 2223
redistributing mass (Whipple, 2009). A river’s discharge determines in part its ability to erode 2224
bedrock, and is controlled by the area of its drainage basin and the precipitation it receives 2225
(Hack, 1957). The positions of drainage divides and thus basin area are not fixed over time. 2226

101

Divides are thought to migrate via coupling between rivers and hillslopes: river incision 2227
generates oversteepened hillslopes, which fail during landslides, occasionally breaching a ridge 2228
and causing one basin to gain area at another’s expense (Burbank et al., 1996; Harvey, 2001; 2229
Hovius et al., 1998; Larsen and Montgomery, 2012) (Figure 1). Over time, this process should 2230
drive tectonically active landscapes toward steady state, where erosion rates are balanced across 2231
divides (Whipple and Tucker, 1999; Willett and Brandon, 2002; Whipple, 2009; Li et al., 2014). 2232
2233
Figure 1. A landslide in southern Taiwan. Divide migration occurs when a landslide 2234
breaches a ridge, causing the divide to move to a new position at the top of the slip surface. 2235
Rivers are color-coded by upstream-averaged local relief. In this case, the higher relief 2236
basin captures area, consistent with a divide progressing toward steady state. This is one of 2237
the larger migration events we documented, capturing ~11,000 m
2
. 2238
2239
Despite the accepted importance of drainage divide migration in shaping tectonically 2240
active landscapes, this process remains poorly understood, with few direct observations (Bonnet, 2241
2009). Landslide-generating events, such as strong earthquakes and extreme storms, offer natural 2242
experiments for observing divide migration processes in action. The goal of this study is to 2243

102

examine the results of these experiments, using high-resolution satellite imagery and digital 2244
elevation models (DEMs) to identify pre- and post-event divide locations. 2245
River channels span most of the relief in drainage basins and control basin geometry 2246
(Whipple and Tucker, 1999). Recent studies have explored topographic metrics for divide 2247
instability (Willett et al., 2014; Whipple et al., 2017) and have identified relationships between 2248
divide migration and river profile morphology (Yang et al., 2015; Whipple et al., 2017). These 2249
studies have been based predominantly on landscape evolution model outputs and inferences 2250
from river profile analysis. Empirical observations offer the opportunity to test these results 2251
directly and to isolate the effect of landslides in coupling river channels to divides. In this 2252
chapter, we quantify divide migration in three locations and find that patterns of area gain and 2253
loss generally result in divides progressing toward steady state, as predicted by channel and basin 2254
geometry. We use this information to attempt to quantify the impact of storm-triggered 2255
landslides on the long-term evolution of the Central Range in Taiwan. 2256
2257
4.1.2 Study areas 2258
Large earthquakes in Wenchuan, China and central Nepal, and a typhoon in southern 2259
Taiwan, together triggered more than 10
5
landslides, a subset of which caused divide migration 2260
(Figure 1). All three events affected steep, mountainous terrain. The Mw 7.9 2008 Wenchuan 2261
earthquake struck central China on May 12, 2008. Strong shaking generated more than 60,000 2262
landslides (Li et al., 2014). The Mw 7.8 Gorkha earthquake struck central Nepal on April 25, 2263
2015 and caused around 25,000 mapped landslides (Roback et al., 2017). Typhoon Morakot 2264
made landfall in Taiwan on August 7, 2009. The most destructive effect of Typhoon Morakot 2265
was its heavy rainfall, up to nearly 3 meters between August 7-9. This rainfall generated more 2266

103

than 20,000 landslides (Lin et al., 2011) in the steep Central Range of southern Taiwan. The 2267
Taiwan dataset is particularly useful in evaluating divide migration for several reasons: the 2268
landslide density was high, image and DEM quality are good over all affected areas, and 2269
independent estimates of the progression of the Central Range towards topographic steady state 2270
allow us to view our divide migration data within the context of long-term landscape evolution 2271
(Stolar et al., 2007). 2272
2273
4.2 METHODS 2274
4.2.1 Landslide mapping 2275
We mapped ridge-breaching landslides visually using Google Earth Pro and measured the 2276
area captured by each landslide. Google Earth uses SRTM 30-meter topography where available, 2277
filling in areas with limited coverage or poor data quality with other DEMs. The advantage of 2278
Google Earth is that the 3D terrain projection simplifies identifying instances of divide 2279
migration. The topography predates the events, resulting in ridge-breaching landslides appearing 2280
draped over the ridge (Figure 1). We assumed the uppermost extent of the scarp represents the 2281
new divide. In both the satellite images and in field observations from all three locations, we 2282
found that most landslides which appear to initiate at the ridge (see Densmore and Hovius, 2000) 2283
actually initiate a few meters below it, and do not cause divide migration. We included only 2284
landslides which clearly breach a ridge, which comprise only a fraction of the total number that 2285
initiate near ridges. 2286

104

Figure 2. Ridge identification by shadow. Images show a ridge in Taiwan before (2001) and 2287
after (2011) Typhoon Morakot, in the top and bottom panels, respectively. The ridge in the 2288
top photo is easily identified by the shadow it casts, making the divide migration caused by 2289
the landslide in the bottom image easily identifiable even without using topographic data. 2290
Throughout the study regions, we used similar instances of visibly well-defined ridges to 2291
check for accurate positioning of images with respect to topography. 2292
2293

105

2294
Figure 3. Geolocation of ridges and rivers. Image shows a ridge and adjacent river valleys 2295
in Taiwan (top) and an elevation profile of the path marked in blue. Image and topography 2296
are both from Google Earth. The ridge and rivers are marked with arrows of 2297
corresponding colors in the image and elevation profile. The imagery and topography in 2298
this area are properly georeferenced. Similar evaluations were used to screen accurate 2299
georeferencing in all three study regions. 2300
2301
Proper image positioning and rectification is critical for this method. Some 2302
misalignments are visible in Google Earth imagery, so we verified that ridges included in this 2303
study are properly georeferenced to ridges in the DEM. To do this, we used ridges that are easily 2304
identifiable by shadows or, where ridges were not clearly identifiable in images, we checked that 2305
nearby streams are correctly located with respect to topographic minima (Supplemental Figures 2306
2-3). This process enables us to exclude areas where divide locations are suspect, roughly 10% 2307
of the total landslide affected area. We mapped landslides specifically for this study because we 2308
found that using polygons from existing landslide inventories yielded significantly less accurate 2309
georeferencing of landslides with respect to ridges. 2310
2311
2312

106

4.2.2 Verifying geolocation of ridges 2313
For our method of mapping divide migrations to be valid, ridges in photos must be 2314
properly georeferenced to ridges in the topography. Google Earth has some known issues with 2315
georeferencing and orthorectification in some areas that can cause mismatching between images 2316
and topography. Ridges are identifiable in satellite images where the sun angle generates 2317
appropriate shadows (Figure 2), or where a vegetation contrast or cliff edge is apparent, and we 2318
used the correspondence of these visible ridges with the Google Earth base topography to 2319
confirm accurate referencing for the areas analyzed in this study, where possible. 2320
Verifying the location of ridges in this manner was not possible in all images. A more 2321
widely applicable method for verifying that images are properly georeferenced is checking that 2322
streams are properly placed at the lowest points of valleys (Figure 3). We assume that when 2323
streams are properly georeferenced, ridges are as well, such that our divide migration mapping 2324
method is reasonable to use where streams are in place. To determine whether this assumption is 2325
valid, we examine locations in our three field areas where ridges are clearly visible, and verify 2326
that both the ridge and the adjacent streams are properly georeferenced. In steep valleys, a 30- 2327
meter resolution DEM sometimes fails to capture all the fine meanders of small streams, but we 2328
find this does not necessarily indicate a poorly georeferenced image. Rather, it is a systematic 2329
displacement of the stream out of a topographic low that indicates a problematic area where 2330
ridge locations are untrustworthy. We checked more than 150 locations where ridges are clearly 2331
visible and found only 2 where streams are properly located but ridges are out of place. Figure 3 2332
shows an example of a properly located ridge flanked by two properly located streams. Given the 2333
good correspondence between properly referenced streams and ridges, we used in-place streams 2334
to screen areas of accurately referenced imagery for use in our analysis. 2335

107

Figure 4. Schematic
of uncertainty in
area calculations
introduced by DEM
resolution. In the
error calculation, pn
is represented by the
centers of the brown
squares. The
distance between pn
and the actual ridge
shown in red is xn.

2336
4.2.3 Error introduced by DEM resolution 2337
Since we calculate the amount of area captured by a landslide using the position of the 2338
ridge before the landslide occurred and define the position of that ridge based on the topography, 2339
error is introduced due to the 30-meter resolution of the DEM. The satellite photos used to 2340
identify the top of the landslide scarp have a resolution of 0.5-2 meter, so the error introduced in 2341
the area calculation by photo resolution is negligible by comparison. 2342
We have already introduced our method for ensuring that satellite imagery is properly 2343
geolocated to the topography, and we excluded areas from our analysis where imagery was not 2344
accurately georeferenced. We thus estimate error based on properly located ridges. While ridges 2345

108

are not linear features, at the scale of an individual landslide we find it is a reasonable 2346
approximation to define the actual ridge as a line. To estimate the error introduced by the DEM 2347
resolution, we wish to calculate the area between the DEM-defined ridge and the actual ridge. 2348
Approximating the location of a ridge using a 30-meter DEM results in a ridge defined by 2349
a series of points p0,p1,…,pn spaced 30 meters apart, each of which is a distance xn from the 2350
actual ridge (Figure 4). The area between a DEM-defined ridge of length l and the actual ridge it 2351
describes is defined: 2352
a
e
= ∑ �
x
n − 1
+ x
n
2
� �r
2
− (|x
n − 1
− x
n
|)
2
l
r
�
n = 0
, (1) 2353
where r is the resolution of the DEM. For a correctly located ridge, maximum distance x for any 2354
point p is: 2355
x
m ax
= �
r
2
2
. (2) 2356
xmin is zero for a point that lies on the actual ridge. For a 30-meter DEM, we find an average x of 2357
10.61 meters. 2358
For each field area, we measure the total length of the affected divides and find 19,900 2359
meters for Taiwan, 23,100 meters for Wenchuan, and 3,300 for Nepal. Applying equation 1, we 2360
obtain error estimates for our area capture calculations for each site: 1.248±0.245 km
2
, 2361
0.541±0.211 km
2
, and 0.068±0.035 km
2
, respectively. 2362
4.2.4 Topographic analysis 2363
While landslides are a hillslope process, and locations of landslides may be controlled by 2364
local slope, ground shaking, fluid flow, and other factors, the focus of this study is at basin scale. 2365
River incision is thought to be a primary driver of drainage basin evolution, so to contextualize 2366
the data on divide migrations and to test for the role of landslides in river-divide coupling, we 2367

109

analyzed rivers draining the affected basins. We calculated topographic metrics proposed to 2368
capture information about divide stability: Χ (chi), upstream-averaged local relief, and upstream- 2369
averaged channel gradient, all of which may indicate cross-divide differences in erosion rate 2370
(Willett et al., 2014; Whipple et al., 2017). The detachment-limited stream power model relates 2371
the change in the elevation of a channel to its slope and drainage area (Howard et al., 1994): 2372
𝛛𝛛 𝛛𝛛 𝛛𝛛 𝛛𝛛 = 𝐔𝐔 ( 𝐱𝐱 , 𝛛𝛛 ) − 𝐊𝐊 ( 𝐱𝐱 , 𝛛𝛛 ) 𝐀𝐀 𝐦𝐦 �
𝛛𝛛 𝛛𝛛 𝛛𝛛 𝐱𝐱
�
𝐧𝐧 (3) 2373
where z is elevation, t is time, U is rock uplift rate, K is erodibility, A is upstream drainage area, 2374
x is distance upstream, and m and n are constants modifying area and slope, whose values may 2375
vary under different conditions. Transient landscapes are expected to evolve toward a state where 2376
uplift is balanced by erosion (Whipple and Tucker, 1999; Willett and Brandon, 2002). Based on 2377
this stream power model, Perron and Royden (2013) proposed the Χ (chi) metric, an integral of 2378
drainage area along a river, for the interpretation of bedrock river profiles: 2379
𝐗𝐗 = ∫ �
𝐀𝐀 𝟎𝟎 𝐀𝐀 �
𝐦𝐦 𝐧𝐧 𝐝𝐝 𝐱𝐱
𝐱𝐱 𝐱𝐱 𝐛𝐛 (4) 2380
where xb is a point on the channel at base level, and A0 is a reference drainage area which gives 2381
chi dimensions of length. In principle, steady-state divides should have equal chi values on either 2382
side. Differences in chi values across divides are expected to reflect divide instability, with a 2383
lower chi stream expected to capture area from a higher chi stream (Willett et al., 2014). 2384
However, the interpretation of chi differences across divides may be complicated by spatial and 2385
temporal variations in U and K, which are poorly constrained in many regions (Whipple and 2386
Tucker, 1999). Furthermore, in some flow conditions the detachment-limited stream power law 2387
may not apply. 2388
Streams were defined and fluvial metrics calculated using the TopoToolbox 2 and 2389
DivideTools functions for Matlab using SRTM topographic data (Schwanghart and Scherler, 2390

110

2014; Forte, 2016). We used 1 km
2
as the minimum drainage area to define a stream and a 2391
standard reference concavity of 0.45. We set base level to 700 meters in Wenchuan (the 2392
elevation of the Sichuan basin, which clearly demarcates the bedrock-alluvial transition) and 500 2393
meters in Nepal, where many rivers of interest enter basins or begin to widen and form alluvial 2394
valleys. In Taiwan, we calculated chi values both by assuming base level is sea level, as well as 2395
defining base level individually for each basin by visually identifying the bedrock-alluvial 2396
transition. Results were essentially identical. We calculated local relief using a radius of 500 2397
meters, which does not exceed the average hillslope length. 2398
For each instance of divide migration we identified, we compared the metrics of the river 2399
that gained drainage area with those of the river that lost area. If landslides are driving the 2400
landscape towards steady state, we expect to see the majority of divide migrations characterized 2401
by gain in area of rivers with lower chi, higher relief, and high gradient (Willett et al., 2014; 2402
Whipple et al., 2017). To evaluate this hypothesis, we compared values of chi, upstream- 2403
averaged local relief, and upstream-averaged gradient at the point along a stream with the most 2404
direct flow path from the affected divide. 2405
2406
4.3 RESULTS 2407
4.3.1 Area exchanged in each event 2408
We find a total of 365 instances of divide migration, with 56 caused by the Gorkha 2409
earthquake, 156 by the Wenchuan earthquake, and 153 by Typhoon Morakot. From these three 2410
events, we measure 1.857 ± 0.49 km
2
of total drainage area exchanged: 1.248 ± 0.245 km
2
in the 2411
Wenchuan earthquake, 0.552 ± 0.211 km
2
in Typhoon Morakot, and 0.068 ± 0.035 km
2
in the 2412
Gorkha earthquake. A single large landslide triggered by the Wenchuan earthquake affected 4 2413

111

divides and was alone responsible for 0.71 km
2
of area exchange. Reported uncertainties include 2414
estimates of the error introduced by the ~30-meter DEM resolution (Figure 4). 2415
Figure 5. Histograms of divide migrations and area capture plotted by difference in cross- 2416
divide metrics. For consistency with the other two metrics, we flip the sign of chi 2417
differences. Positive differences in metrics indicate progress toward steady state. Solid lines 2418
show numbers of migration instances broken down by site; dashed lines and shaded 2419
histogram show areas for all sites together. We do not include the largest Wenchuan 2420
landslide in the area capture measurement as it caused ~0.7 km
2
to be exchanged. 2421
2422
2423
2424

112

4.3.2 Comparison of divide stability metrics 2425

61.4% of migration directions were towards steady state as predicted by relief differences 2426
(p = 8.18 × 10
-6
, where p is the probability that migration direction is irrespective of the metric), 2427
58.7% by gradient (p = 5.72 × 10
-4
), and 56.4% by chi (p = 7.97× 10
-3
). Local relief and channel 2428
gradient better predict divide migration than chi differences when compared across transient 2429
divides in simulations (Whipple et al., 2017) as well as in this dataset (Table 1). Overall, divide 2430
migrations are roughly normally distributed with respect to differences in chi, relief, and gradient 2431
(Figure 5). The amount of area captured in each landslide does not appear to be dependent on 2432
fluvial geometry, i.e., larger landslides are not necessarily associated with larger differences in 2433
relief, gradient, or chi across divides. Cross-divide differences in mean local relief and mean 2434
gradient are similar at each affected divide, but gradients in chi are only weakly correlated with 2435
gradients in the other two metrics (Figure 6). 2436
2437
4.4 DISCUSSION 2438
4.4.1 Landslides drive divides toward steady state 2439
The results of this study show that event-triggered landslides measurably drive a 2440
landscape toward steady state with respect to the river network. This indicates that, even on the 2441
timescale of an earthquake or storm, mobility of divides is conditioned by river incision, and thus 2442
incision can be coupled to hillslopes and divides by landslides. These links have been inferred in 2443
prior studies (e.g., Stark, 2010; Buscher et al., 2017) but not conclusively demonstrated with the 2444
kind of direct empirical evidence provided here. 2445
2446

113

Figure 6. Cross-divide differences in
chi, relief, and gradient plotted
against each other. Each point
represents the difference in metrics
for a single divide migration site.
The strong correlation between
relief and gradient is indicative that
both metrics represent
straightforward measures of basin
geometry, i.e., a basin where rivers
have a steep gradient should also
have high relief as well. Calculating
chi involves more assumptions and
considers downstream geometry,
which may lead to the lack of
correlation with the other two
metrics.
2447

114

2448
Figure 7. Map of divide migrations used in this study. Main divide of Taiwan Central 2449
Range plotted in red, hypothetical eastern and western divides used in nr calculation in 2450
black. A; Nepal, B; Taiwan, C; Wenchuan. 2451
2452
Mean local relief may predict divide migration well because it is essentially a coarse 2453
measure of hillslope angle. Higher resolution topographic data than the 30-meter SRTM we have 2454
available could allow the effects of local slope to be disentangled from channel geometry, further 2455
clarifying relationships between channels and divide migration. Tributary capture may be 2456
important in basin reorganization on similar timescales, but we do not identify any instances of 2457
tributary capture in these events. 2458
The lack of any strong relationship between chi and the other metrics (Figure 6) 2459
highlights the difficulty in choosing a proxy for basin stability. In Nepal and Taiwan, chi 2460
disparities do not predict the direction of divide migration, but for the Wenchuan earthquake, chi 2461
outperforms the other metrics (Table 1). The reason for this difference is not immediately 2462
obvious, but may suggest that river incision better conforms to the stream power model, upon 2463
which the chi calculation is based, in the Wenchuan region. Interpretation of metrics which 2464
depend on the stream power model may be complicated by nonuniform uplift and erodibility. 2465
Additionally, since chi is an integral from base level, including downstream geometry of a basin, 2466

115

it may be better suited to examining broad, regional trends rather than area exchange between 2467
first order basins. 2468
2469
4.4.2 A metric for divide migration 2470
Drainage divides, like coastlines, have a fractal character, making linear measurements 2471
scale-dependent (Rodriguez-Iturbe and Rinaldo, 2001). We propose a metric for the mobility of 2472
drainage divides and the geometric transience of landscapes that is independent of linear divide 2473
migration rates: 2474
𝐧𝐧 𝐫𝐫 =
𝐀𝐀 𝐞𝐞 𝐀𝐀 𝛛𝛛 𝛛𝛛 �
(5) 2475
where a reorganization number 𝐧𝐧 𝐫𝐫 is defined as the ratio of the area exchanged in a divide 2476
migration event 𝐀𝐀 𝐞𝐞 to the total area of the affected landscape 𝐀𝐀 𝛛𝛛 , divided by the characteristic 2477
timescale, 𝛛𝛛 of divide migration events. This yie lds an absolute measure of divide mobility, not 2478
considering progression toward steady state and irrespective of the scale of the affected basins. 2479
For divide migrations triggered by Typhoon Morakot, we calculate 𝐧𝐧 𝐫𝐫 = 1.06 × 10
-6
yr
-1
based on 2480
a minimum recurrence interval ( 𝛛𝛛) of 200 years (West et al., 2011), area exchanged ( 𝐀𝐀 𝐞𝐞 ) of 0.552 2481
km
2
, and total area affected by divide migration ( 𝐀𝐀 𝛛𝛛 ) of 2.6 × 10
3
km
2
. This 𝐧𝐧 𝐫𝐫 value represents 2482
the fraction of area in a landscape that is exchanged between basins of any order during the 2483
timescale of interest. For Wenchuan, given 𝛛𝛛 of 2300- 3300 years for large earthquakes on the 2484
Longmenshan Fault Zone (Ran et al., 2010), 𝐀𝐀 𝐞𝐞 of 1.294 km
2
, and 𝐀𝐀 𝛛𝛛 of 4.0 × 10
3
km
2
yields 𝐧𝐧 𝐫𝐫 2485
= 1.4 × 10
-7
-9.8 × 10
-8
yr
-1
. Events with different recurrence intervals may cause divide migration 2486
in the same landscapes, but 𝐧𝐧 𝐫𝐫 may be useful for quantifying a single event’s influence on a 2487
landscape, and the overall motility of divides in a landscape. The higher value of 𝐧𝐧 𝐫𝐫 for Taiwan 2488
compared to Wenchuan implies comparatively more rapid reorganization of this landscape 2489

116

during the single events studies here. Moreover, these events are likely to occur more frequently 2490
in Taiwan: the recurrence interval for large earthquakes in Taiwan is ~475 years (Cheng et al., 2491
2007) versus 2300-3300 for the Longmenshan Fault Zone, while Taiwan also experiences greater 2492
rainfall erosivity (Panagos et al., 2017). 2493
While there is much debate as to whether arc-continent collision and orogeny in Taiwan 2494
is progressing from north to south (Suppe, 1981) or occurring simultaneously along strike (Lee et 2495
al. 2015), the topography of the southernmost 125 km of Taiwan along strike ( 𝐀𝐀 𝛛𝛛 ~4,250 km
2
2496
area) appears not to have achieved steady state. Stolar et al. (2007) estimate a duration t of 1.8- 2497
2.3 Myr from subaerial exposure to steady-state topography. We assume that steady state is 2498
achieved by migration of the main divide, and its steady-state position lies between its current 2499
easternmost and westernmost extents in the southern Central Range (Figure 7). To obtain a 2500
maximum estimate of area that must be exchanged between east and west-flowing basins to 2501
achieve steady state, we assume that the divide must migrate from one extreme to the other. 2502
Between the hypothetical easternmost and westernmost divide positions, we measure 𝐀𝐀 𝐞𝐞 ~1,133 2503
km
2
. We find the maximum 𝐧𝐧 𝐫𝐫 required for the main divide to reach steady state in 1.8-2.3 Myr 2504
is 1.2-1.5 × 10
-7
yr
-1
. 9/153 of Typhoon Morakot migrations occurred on the main divide, with an 2505
average area captured of 3600 m
2
. Assuming 64% of migrations result in progress toward steady 2506
state (Table 1, from the value for relief) yields 𝐧𝐧 𝐫𝐫 of 1.8 × 10
-8
yr
-1
. Comparing this value to nr = 2507
1.2-1.5 × 10
-7
yr
-1
estimated for the long-term means that Morakot-type landslides account for a 2508
minimum of 12-15% of the motion of the main divide toward steady state. Tributary capture and 2509
other landslide-generating events such as earthquakes may also contribute to migration of the 2510
central divide. We emphasize that this calculation is based on a small landslide population and 2511
that we use a maximum estimate of the amount of necessary area exchange. The role of typhoon- 2512

117

triggered landslides may thus be more important than we estimate, but this method could provide 2513
an approach for more robustly evaluating the role of landslides given data from a larger number 2514
of events. 2515
2516
4.5 CONCLUSIONS 2517
By examining ridge-breaching landslides triggered by three recent events, we have 2518
demonstrated that event-triggered landslides couple river channels to hillslopes and ridges and 2519
lead to migration of drainage divides toward steady state. A landscape’s progress toward steady 2520
state is thus measurable at the timescale of one earthquake or storm. We compare three channel 2521
morphology proxies for the direction a divide will migrate to achieve steady state: Chi, mean 2522
local relief, and mean channel gradient. All three meaningfully predict the direction of observed 2523
migrations. Chi is an excellent predictor for Wenchuan migrations, but does not perform as well 2524
in the other areas. Cross-divide differences in relief and gradient closely correspond for all three 2525
sites, while differences in chi are weakly correlated with the other two metrics. 2526
We propose a reorganization number as a metric for divide mobility, helping to quantify 2527
the impact of a landsliding event on a landscape. Applying this approach to southern Taiwan, 2528
where the progression of the landscape toward steady state has been widely discussed, we find 2529
that typhoons on the scale of Morakot are likely responsible for a minimum of 12-15% of the 2530
motion of the main divide in the Central Range toward a steady-state position. 2531

118

FIGURE 8A

119

FIGURE 8B

120

FIGURE 8C

121

FIGURE 8D

122

FIGURE 8E

123

FIGURE 8F

124

FIGURE 8G

125

FIGURE 8H

126

FIGURE 8I
Figure 8. A, B, C Maps of chi, local relief, and channel gradient for southern Taiwan. D, E, 2532
F Maps of chi, local relief, and channel gradient for Wenchuan Earthquake affected region, 2533
China. G, H, I Maps of chi, local relief, and channel gradient for Gorkha Earthquake 2534
affected region, Nepal. 2535
2536
2537
2538
2539

127

4.6 REFERENCES 2540
Bonnet, S., 2009, Shrinking and splitting of drainage basins in orogenic landscapes from the 2541
migration of the main drainage divide: Nature Geoscience, v. 2, no. 11, p. 766–771, 2542
doi:10.1038/ngeo666. 2543
2544
Burbank, D. W., J. Leland, E. Fielding, R. S. Anderson, N. Brozovic, M. R. Reid, and C. 2545
Duncan, 1996, Bedrock incision, rock uplift and threshold hillslopes in the northwestern 2546
Himalayas: Nature, v. 379, no. 6565, p. 505–510, doi:10.1038/379505a0. 2547
2548
Buscher, J., A. Ascione, and E. Valente, 2017, Decoding the role of tectonics, incision and 2549
lithology on drainage divide migration in the Mt. Alpi region, southern Apennines, Italy: 2550
Geomorphology, v. 276, p. 37–50, doi:10.1016/j.geomorph.2016.10.003. 2551
2552
Cheng, C.-T., Chiou, S.-J., Lee, C.-T., Tsai, Y.-B., 2007, Study on probabilistic seismic hazard 2553
maps of Taiwan after Chi-Chi earthquake: Journal of GeoEngineering v. 2, no.1, p. 19– 2554
28. 2555
2556
Densmore, A. L., and N. Hovius, 2000, Topographic fingerprints of bedrock landslides: 2557
Geology, v. 28, no. 4, p. 371, doi:10.1130/0091-7613(2000)28<371:tfobl>2.0.co;2. 2558
2559
Forte, A., 2016, DivideTools: Set of Matlab-based tools for analyzing drainage divide stability: 2560
https://github.com/amforte/DivideTools. 2561
2562
Hack, J. T., 1957, Studies of longitudinal stream profiles in Virginia and Maryland: US 2563
Geological Survey professional paper, v. 294-B, p. 45-97 2564
2565
Harvey, A., 2001, Coupling between hillslopes and channels in upland fluvial systems: 2566
implications for landscape sensitivity, illustrated from the Howgill Fells, northwest 2567
England: Catena, v. 42, no. 2-4, p. 225–250, doi:10.1016/s0341-8162(00)00139-9. 2568
2569
Hovius, N., C. Stark, M. Tutton, and L. Abbott, 1998, Landslide-driven drainage network 2570
evolution in a pre-steady-state mountain belt: Finisterre Mountains, Papua New Guinea: 2571
Geology, v. 26, no. 12, p. 1071-1074, doi:10.1130/0091- 2572
7613(1998)026<1071:lddnei>2.3.co;2. 2573
2574
Howard, A. D., W. E. Dietrich, and M. A. Seidl, 1994, Modeling fluvial erosion on regional to 2575
continental scales: J. Geophys. Res.: Solid Earth, v. 99, no. B7, p. 13971–13986, 2576
doi:10.1029/94jb00744. 2577
2578
Larsen, I. J., and D. R. Montgomery, 2012, Landslide erosion coupled to tectonics and river 2579
incision: Nature Geoscience, v. 5, no. 7, p. 468–473, doi:10.1038/ngeo1479. 2580
2581
Lee, Y.-H., Byrne, T., Wang, W.-H., Lo, W., and Rau, R.-J., 2015, Simultaneous mountain 2582
building in the Taiwan orogenic belt: Geology, v. 43, p. 451–454, doi:10.1130/G36373.1. 2583
128
2584 Li, G., A. J. West, A. L. Densmore, Z. Jin, R. N. Parker, and R. G. Hilton, 2014, Seismic
2585 mountain building: Landslides associated with the 2008 Wenchuan earthquake in the
2586 context of a generalized model for earthquake volume balance: Geochemistry,
2587 Geophysics, Geosystems, v. 15, no. 4, p. 833–844, doi:10.1002/2013gc005067.
2588
2589 Lin, C.-W., W.-S. Chang, S.-H. Liu, T.-T. Tsai, S.-P. Lee, Y.-C. Tsang, C.-L. Shieh, and C.-M.
2590 Tseng, 2011, Landslides triggered by the 7 August 2009 Typhoon Morakot in southern
2591 Taiwan: Engineering Geology, v. 123, no. 1-2, p. 3–12,
2592 doi:10.1016/j.enggeo.2011.06.007.
2593
2594 Panagos, P. et al., 2017, Global rainfall erosivity assessment based on high-temporal resolution
2595 rainfall records: Scientific Reports, v. 7, no. 1, doi:10.1038/s41598-017-04282-8.
2596
2597 Perron, J. T., and L. Royden, 2012, An integral approach to bedrock river profile analysis: Earth
2598 Surface Processes and Landforms, v. 38, no. 6, p. 570–576, doi:10.1002/esp.3302.
2599
2600 Ran, Y., L. Chen, J. Chen, H. Wang, G. Chen, J. Yin, X. Shi, C. Li, and X. Xu, 2010,
2601 Paleoseismic evidence and repeat time of large earthquakes at three sites along the
2602 Longmenshan fault zone: Tectonophysics, v. 491, no. 1-4, p. 141–153,
2603 doi:10.1016/j.tecto.2010.01.009.
2604
2605 Roback, K., M. K. Clark, A. J. West, D. Zekkos, G. Li, S. F. Gallen, D. Chamlagain, and J. W.
2606 Godt, 2017, The size, distribution, and mobility of landslides caused by the 2015 Mw7.8
2607 Gorkha earthquake, Nepal: Geomorphology, doi:10.1016/j.geomorph.2017.01.030.
2608
2609 Rodríguez-Iturbe, I., and A. Rinaldo, 2001, Fractal river basins: chance and self-organization:
2610 Cambridge, Cambridge University Press.
2611
2612 Schwanghart, W., and D. Scherler, 2014, TopoToolbox 2 – MATLAB-based software for
2613 topographic analysis and modeling in Earth surface sciences: Earth Surface Dynamics, v.
2614 2, no. 1, p. 1–7, doi:10.5194/esurf-2-1-2014.
2615
2616 Stark, C. P., 2010, Oscillatory motion of drainage divides: Geophysical Research Letters, v. 37,
2617 no. 4, doi:10.1029/2009gl040851.
2618
2619 Stolar, D. B., S. D. Willett, and D. R. Montgomery, 2007, Characterization of topographic steady
2620 state in Taiwan: Earth and Planetary Science Letters, v. 261, no. 3-4, p. 421–431,
2621 doi:10.1016/j.epsl.2007.07.045.
2622
2623 Strahler, A. N., 1952, Dynamic basis of geomorphology: Geological Society of America
2624 Bulletin, v. 63, no. 9, p. 923, doi:10.1130/0016-7606(1952)63[923:dbog]2.0.co;2.
2625
2626 Suppe, J., 1981, Mechanics of mountain building and metamorphism in Taiwan: Memoirs of the
2627 Geological Society of China, v. 4, p. 67–89.
2628
2629

129

West, A. J., C.-W. Lin, T.-C. Lin, R. G. Hilton, S.-H. Liu, C.-T. Chang, K.-C. Lin, A. Galy, R. 2630
B. Sparkes, and N. Hovius, 2010, Mobilization and transport of coarse woody debris to 2631
the oceans triggered by an extreme tropical storm: Limnology and Oceanography, v. 56, 2632
no. 1, p. 77–85, doi:10.4319/lo.2011.56.1.0077. 2633
2634
Whipple, K. X., and G. E. Tucker, 1999, Dynamics of the stream-power river incision model: 2635
Implications for height limits of mountain ranges, landscape response timescales, and 2636
research needs: J. Geophys. Res.: Solid Earth, v. 104, no. B8, p. 17661–17674, 2637
doi:10.1029/1999jb900120. 2638
2639
Whipple, K. X., 2009, The influence of climate on the tectonic evolution of mountain belts: 2640
Nature Geoscience, v. 2, no. 2, p. 97–104, doi:10.1038/ngeo413. 2641
2642
Whipple, K. X., A. M. Forte, R. A. Dibiase, N. M. Gasparini, and W. B. Ouimet, 2017, 2643
Timescales of landscape response to divide migration and drainage capture: Implications 2644
for the role of divide mobility in landscape evolution: J. Geophys. Res.: Earth Surface, v. 2645
122, no. 1, p. 248–273, doi:10.1002/2016jf003973. 2646
2647
Willett, S. D., and M. T. Brandon, 2002, On steady states in mountain belts: Geology, v. 2648
30, no. 2, p. 175, doi:10.1130/0091-7613(2002)030<0175:ossimb>2.0.co;2. 2649
2650
Willett, S. D., S. W. Mccoy, J. T. Perron, L. Goren, and C.-Y. Chen, 2014, Dynamic 2651
reorganization of river basins: Science, v. 343, no. 6175, p. 1248765-1–1248765-9, 2652
doi:10.1126/science.1248765. 2653
2654
Yang, R., S. D. Willett, and L. Goren, 2015, In situ low-relief landscape formation as a result of 2655
river network disruption: Nature, v. 520, no. 7548, p. 526–529, doi:10.1038/nature14354. 2656

130

Chapter 5: Conclusions 2657
5.1 Extreme events drive landscape evolution from channel to ridge 2658
Presented in the preceding chapters is a pair of datasets and accompanying topographic 2659
analyses which attempt to bridge the gaps above (to the ridge) and below (to the major rivers) the 2660
predictions the threshold hillslope model makes about the ways steep landscapes respond to 2661
tectonic and climatic forcing. In particular, we examine the way mass wasting driven by major 2662
earthquakes and storms manifests in the geometry of the landscape and controls process regime 2663
transitions. 2664
In Chapter 2, we presented our findings from an inventory of debris flows associated with 2665
the Gorkha Earthquake. For the first time, we assembled a large (1164 total flows) inventory of 2666
debris flows, filling an important niche in the study of debris flows alongside smaller-scale field 2667
and laboratory studies. We categorized these flows into four groups: Type-1 debris flows initiate 2668
from landslide deposits, mostly co-seismic landslides, that have been deposited in steep, 2669
colluvial channels. These landslide deposits are remobilized later during heavy rainfall. Type-2 2670
debris flows initiate from landslides that initiate in saturated conditions, become channelized, 2671
and continue to flow downstream. Mixed debris flows are a combination of the first two types, 2672
initiating either when a debris flow is sourced from both new and pre-existing landslides, or 2673
when it is sourced from a re-activated pre-existing landslide. We also distinguish Type-2 debris 2674
flows that occurred before the Gorkha Earthquake from those that occurred afterward. All the 2675
flows we identified that occurred before the earthquake were Type-2, as all of the Type-1 flows 2676
we identified were sourced from co-seismic landslides. Between 2009-early 2015 we mapped 2677
104 debris flows, compared to 1060 from 2015-2017, the vast majority of which occurred during 2678
the 2015 monsoon. Type-1 debris flows occurred nearly exclusively during the 2015 monsoon, 2679

131

indicating that landslide deposits that are vulnerable to being remobilized as debris flows are 2680
largely flushed out of steep channels in the first major rainfalls following an earthquake. For 2681
Type-2 debris flows, we find a decrease in the intensity-duration precipitation threshold for 2682
debris flow/shallow landslide initiation commensurate with the reduction that was identified 2683
following the Wenchuan Earthquake in 2008. 2684
We find that Type-1 debris flows initiate in consistently steeper channels than either pre- 2685
or post-seismic Type-2 flows. Type-1 flows appear to have a threshold channel slope for 2686
initiation similar to the 22° value that has been observed for debris flows initiating by a similar 2687
mechanism in flume experiments. We also found that Type-2 debris flows tended to run out for 2688
greater distances than Type-1, although we were unable to find any obvious relationships 2689
between channel geometry, precipitation, and runout distance, despite these relationships having 2690
been observed in prior studies. An important observation from our inventory is that the debris 2691
flows frequently terminate where the channels in which they initiate form a confluence with a 2692
larger river. Likely, the angle of the junction also plays a role. 2693
In the interest of improving our understanding of the cascading hazard chain after major 2694
earthquakes, we develop a simple framework toward a model for Type-1 debris flow hazards. 2695
The ultimate goal of this is to produce a model that can be applied easily and rapidly in the wake 2696
of a major earthquake to identify at-risk areas for debris flows during the next heavy rainfall. 2697
Identifying channels that both exceed a threshold slope and have access to co-seismic landslide 2698
material appears to have some crude predictive ability. We expect this method could be 2699
considerably improved if more factors are accounted for and ought to be expanded to include 2700
hazard prediction for Type-2 debris flows. 2701

132

In Chapter 3, we examine debris flows in the Nepal Himalaya in a different context. 2702
Examining the slope distributions of the debris flow channels, we can distinguish a threshold 2703
slope around 11°, or 0.2, above which channels appear to have their morphology generally 2704
controlled by debris flows, and below which fluvial processes, including monsoon floods and 2705
glacial lake outburst floods (GLOFs) dominate incision. This represents the point below which 2706
debris flows act mainly as a vehicle of sediment delivery, rather than a primarily erosive force. 2707
This observation is critical to inform the rest of the analysis in this chapter. How does this point 2708
move in response to variable coupled uplift-erosion in a basin? We conducted analyses of the 2709
topography in the study area explored in Chapter 2 to elucidate these process regime transitions 2710
and test whether GLOFs might be the dominant erosive force in large, north-south draining 2711
rivers in this region. 2712
Since GLOFs are relatively rare events in an individual river basin, at least with respect 2713
to the annual monsoon, or the hundreds to thousands of landslides and debris flows that might 2714
occur due to a single large storm or earthquake, it was necessary to select a proxy for the 2715
magnitude of GLOF incision. We assumed that GLOF frequency in a basin is proportional to the 2716
amount of drainage area above the equilibrium line altitude (ELA) (the elevation above which 2717
snow and ice accumulate year to year) that it captures. For analyses involving river long profiles, 2718
we use the last glacial maximum ELA, and for channel width, which is much quicker to adjust. 2719
We used a variety of metrics of river morphology to in conjunction with GLOF frequency 2720
to examine the way tributary channels have evolved in response to differing rates of GLOF 2721
incision in the mainstems. Firstly, we calculated normalized steepness index (ksn), both as a basin 2722
average and as ksn differences between 1
st
and 2
nd
order tributaries and the 4
th
or higher order 2723
rivers they empty into, the metric a1, which describes how wide the range of drainage areas 2724

133

debris flow incision is in a basin, and the proportion of channels steeper than 11° as a proxy for 2725
the amount of the channel network susceptible to debris flow erosion. We find that gradients in 2726
ksn differences, a1, and the proportion of channels steeper than 11° correspond well with the 2727
physiographic transition (PT) in Nepal. 2728
We found strong correlation between GLOF frequency in a trunk stream and the ksn 2729
difference between the trunk stream and its tributaries. This is indicative of steepening in the 2730
tributary streams to keep up with rapid incision in GLOF dominated streams. This steepening 2731
appears to be coupled with debris flow incision progressing further downstream, as basin a1 and 2732
ksn differences are also strongly correlated. This indicates that in steeper basins, as debris flows 2733
begin to dominate in channels further and further downstream, it is reflected as a reduction in ksn 2734
differences between tributary streams and trunk streams. Therefore, ksn differences are an 2735
effective means of distinguishing process domains. Channel width, normalized to the power-law 2736
trend it follows with discharge, appears to decrease with GLOF frequency. This relationship is 2737
not immediately straightforward, as GLOFs have been shown to drive lateral incision and carry 2738
heavy bedloads, but we interpret the decrease in the width of the active channel as illustrating the 2739
power of GLOFs to evacuate alluvium and expose the channel bed to erosion. Additionally, we 2740
find that basin a1 corresponds closely with detrital (U-Th)/He (r
2
= 0.48). As a metric for basin- 2741
averaged exhumation rate, a1 appears to work well, better than ksn, even though these ages are 2742
integrated over several million years. 2743
Finally, applying the chi integration to channel profiles in the study area, we find that 2744
broad chi disparities suggest that some catchments capturing large amounts of glaciated drainage 2745
area are eroding more rapidly and capturing drainage area from their neighbors. These many 2746
observations in concert highlight the importance of GLOFs in setting the pace of river incision, 2747

134

and in non-glaciated channels, channels steepen and allow debris flows, usually generated by 2748
extreme events, to drive incision farther downstream. This is supportive toward our vision of a 2749
model similar to the threshold hillslope model, where a downstream process sets base level for 2750
the processes that occur upstream to it, mediated by thresholds above or below which the process 2751
is no longer a dominant agent of geomorphic change. The extent of GLOF incision downstream 2752
may also be an important control on the location of the PT, as the crossover point where 2016 2753
and 1981 GLOFs in the Bhote Koshi river no longer have discharges that exceed that of a 2754
monsoon flood with a similar recurrence interval corresponds closely with the location of the PT. 2755
Finally, in Chapter 4 we examined the role of extreme events in driving the redistribution 2756
of drainage area among basins. We assembled a dataset of landslides triggered in the Gorkha 2757
Earthquake, the Wenchuan Earthquake, and Typhoon Morakot that caused drainage divides to 2758
migrate, the first inventory of such events. We identified clear preference for migration 2759
directions that drove divides toward an expected steady-state configuration. This is a successful 2760
test of the threshold hillslope model, demonstrating coupling between channel incision and 2761
landsliding on the bordering hillslopes, even at the timescale of a single extreme event. We 2762
compared three proxies proposed for divide stability: Chi, mean local relief, and mean channel 2763
gradient. All three meaningfully predicted the direction of observed divide migrations. However, 2764
chi is inconsistent. Chi performed very well in predicting Wenchuan Earthquake divide 2765
migrations, but failed for the other two sites. This is also exemplified by the relationships 2766
between cross-divide differences in the different proxies. Channel gradient and relief, which are 2767
very simple physical measurements, are closely correlated, while chi differences correlate poorly 2768
with either of the other metrics. 2769

135

We also proposed a reorganization number as a metric for divide mobility, attempting to 2770
quantify the impact of a landsliding event on a landscape. We applied this approach to Southern 2771
Taiwan, a region which has been proposed to be in transient adjustment as the arc-continent 2772
collision driving uplift in Taiwan progresses southward. This metric indicates that typhoons with 2773
a scale and recurrence interval similar to Typhoon Morakot are likely responsible for a minimum 2774
of 12-15% of the motion of divides toward a steady state configuration, with the rest 2775
accommodated by co-seismic landsliding, storms of different intensities, or major drainage 2776
capture events. 2777

Abstract (if available)

Abstract Geologists have recognized the potential influence of extreme events on landscapes since the early 20th century, when the Channeled Scablands of Washington were revealed to be carved during the Pleistocene by cataclysmic lake outburst floods. However, the relative rarity of extreme earthquakes, floods, and storms, makes their cumulative effect on landscapes difficult to study empirically. I use a combination of debris flow and landslide mapping from satellite imagery and topographic analysis to examine different aspects of extreme events on mountain belts. Large datasets of post-seismic debris flows are necessary to predict their initiation and runout characteristics, yet such datasets have remained scarce. I use satellite imagery to compile an inventory of >1000 debris flows associated with the 2015 Gorkha Earthquake in Nepal. The inventory suggests distinct initiation and runout conditions for debris flows sourced from co-seismic landslide debris, versus those sourced from post-seismic, monsoon-triggered landslides. This inventory shows confluence angles exerted an important control on flow runout distance, and most vulnerable landslide deposits formed debris flows during the first monsoon following the earthquake. Glacial lake outburst floods (GLOFs) are also powerful erosional agents, though they operate on different spatial and temporal scales than debris flows. Regional-scale analysis of the roles of these processes in landscape evolution has been limited. I use the inventory of Gorkha Earthquake debris flows in conjunction with several topographic metrics to better understand how relationships between different process domains (debris flow vs. GLOF vs. monsoon-driven fluvial) contribute to shaping the landscape, and what topographic signatures are characteristic of different processes. This analysis shows basins that capture glaciated headwaters (where GLOFs can originate) have distinct geometry versus those that do not, and rapid GLOF-driven incision in glaciated rivers appears to drive steepening in their non-glaciated tributaries. The distinct, structurally controlled physiographic transition between the Lesser and High Himalayas may also reflect a transition in dominant geomorphic processes. The evolution of watershed boundaries is also subject to the influence of extreme events. Drainage divide migration reorganizes river basins, redistributing erosive energy and contributing to feedbacks between tectonics, erosion, and climate. However, conditions governing divide migration and the timescales on which it occurs are poorly understood. By connecting channels to hillslopes in steep landscapes, landslides play a central role in divide migration. I examine landslides triggered by three events (two earthquakes and a tropical cyclone) that breached ridges, causing area to be exchanged between drainage basins. Using several proposed metrics for divide stability based on river channel morphology, I find that patterns of area gain and loss between basins are consistent with landscapes progressing toward steady state (where uplift is balanced by erosion). I also propose a metric to quantify divide migration and the contribution of an event toward topographic steady state. Restricting analysis to the main drainage divide and using estimates of recurrence interval and the rate of topographic evolution in Taiwan, landslides triggered by large typhoons should account for a minimum of 12-15% of southern Taiwan’s progress toward steady state.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Tectonic control on landsliding revealed by the 2015 Gorkha earthquake

PDF

Observations of temporal and spatial patterns of strain accommodation and earthquake occurrence along strike-slip faults of New Zealand and southern California, USA

Asset Metadata

Creator Dahlquist, Maxwell Philip Boulet (author)

Core Title Extreme events as drivers of landscape evolution in active mountain belts

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Geological Sciences

Publication Date 10/24/2020

Defense Date 02/25/2019

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag debris flows,drainage divides,geohazards,Geomorphology,landscape evolution,Landslides,mountain belts,OAI-PMH Harvest,remote sensing,river networks,topographic analysis,watersheds

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor West, A. Joshua (committee chair), Dolan, James F. (committee member), Nutt, Steven R. (committee member)

Creator Email dahlquis@usc.edu,mpdahlquist@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-148219

Unique identifier UC11660183

Identifier etd-DahlquistM-7283.pdf (filename),usctheses-c89-148219 (legacy record id)

Legacy Identifier etd-DahlquistM-7283.pdf

Dmrecord 148219

Document Type Dissertation

Format application/pdf (imt)

Rights Dahlquist, Maxwell Philip Boulet

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA