THE CONTRIBUTION OF LARGE, SLOW-MOVING LANDSLIDES TO
LANDSCAPE EVOLUTION
by
BENJAMIN HUNTER MACKEY
A DISSERTATION
Presented to the Department of Geological Sciences
and the Graduate School of the University of Oregon
in partial fulfillment of the requirements
for the degree of
Doctor of Philosophy
December 2009
11
University of Oregon Graduate School
Confirmation of Approval and Acceptance of Dissertation prepared by:
Benjamin Mackey
Title:
"The Contribution of Large, Slow-Moving Landslides to Landscape Evolution"
This dissertation has been accepted and approved in partial fulfillment of the requirements for
the Doctor of Philosophy degree in the Department of Geological Sciences by:
Joshua Roering, Chairperson, Geological Sciences
Ilya Bindeman, Member, Geological Sciences
Dean Livelybrooks, Member, Physics
Ray Weldon, Member, Geological Sciences
W. Andrew Marcus, Outside Member, Geography
and Richard Linton, Vice President for Research and Graduate Studies/Dean of the Graduate
School for the University of Oregon.
December 12,2009
Original approval signatures are on file with the Graduate School and the University of Oregon
Libraries.
© 2009 Benjamin Hunter Mackey
iii
An Abstract of the Dissertation of
Benjamin Hunter Mackey for the degree of
in the Department of Geological Sciences to be taken
IV
Doctor of Philosophy
December 2009
Title: THE CONTRIBUTION OF LARGE, SLOW-MOVING LANDSLIDES TO
LANDSCAPE EVOLUTION
Approved: ~~ _
Joshua J. Roering
This dissertation discusses the contribution of deep-seated landslides and
earthflows to the morphology, erosion, and evolution of mountainous landscapes,
focusing on the northern California Coast Ranges.
In active landscapes, channel incision is necessary to create relief but also
increases stresses in adjacent hillslopes, ultimately leading to slope failure. While
conceptually simple, the spatial relationships between channel incision and landsliding
have not been well quantified. Along the South Fork Eel River, I mapped the distribution
of deep-seated landslides using light detection and ranging (LiDAR) derived maps.
Landslide density increases in regions subject to late Pleistocene - Holocene channel
incision and particularly in response to lateral incision at the apex of meander bends.
vWavelet analysis of channel sinuosity reveals hillslopes are most sensitive to meander
wavelengths of 1.5 km.
Argillaceous lithology generates abundant earthflow activity along the main stem
Eel River, yet spatial and temporal patterns of earthflow movement are poorly
understood. I undertook a detailed study ofthe Kekawaka Earthflow using LiDAR,
meteoric lOBe in soil, orthorectified historical aerial photographs, and field surveys.
Inventories of lOBe in soil pits increase systematically downslope, indicate an average
movement rate of 2.1 ± 1.3 m1a over the past 150 years, and establish a minimum
earthflow age of 1700 years. The Kekawaka earthflow has a systematic history of
movement, both spatially, with greatest movement in the narrow transport zone, and
temporally, as velocities peaked in the 1960's and have slowed since 1981.
I used LiDAR and aerial photographs to map earthflow movement and calculate
sediment flux across 226 km2 of the main stem Eel River. From 1944 - 2006, 7.3% of the
study area was active, and earthflows account for an erosion rate of 0.53 ± 0.04 mm/a,
over half the regional average sediment yield. Velocity time series on 17 earthflows
suggest temporal earthflow behavior is influenced by decadal-scale changes in
precipitation, temperature, and river discharge, although local topographic factors can
overwhelm this climatic signal. When active, earthflows erode an order of magnitude
faster than surrounding terrain; however, source supply limitations appear to govern long-
term earthflow evolution.
This dissertation includes previously published coauthored material.
vi
CURRICULUM VITAE
NAME OF AUTHOR: Benjamin H. Mackey
PLACE OF BIRTH: Blenheim, New Zealand
GRADUATE AND UNDERGRADUATE SCHOOLS ATTENDED:
University of Oregon
University of Canterbury
DEGREES AWARDED:
Doctor of Philosophy, 2009, University of Oregon
Bachelor of Science (1 st Class Honours - Geological Sciences), 2003, University
of Canterbury
Bachelor of Laws (1st Class Honours), 2002, University of Canterbury
AREAS OF SPECIAL INTEREST:
Landscape Evolution
Tectonic Geomorphology
Landslides and Natural Hazards
PROFESSIONAL EXPERIENCE:
Graduate Teaching Fellow, University of Oregon, 2004-2009
GRANTS, AWARDS AND HONORS:
Fulbright-EQC Graduate Award in Natural Disaster Research, Fulbright
New Zealand, 2004
University Prize in Science, University of Canterbury, 2003
PUBLICATIONS:
Mackey, RH., Roering, lJ., and McKean, lA., 2009, Long-term kinematics and
sediment flux of an active earthflow, Eel River, California: Geology, v. 37, p.
803-806.
Vll
V111
ACKNOWLEDGMENTS
Foremost thanks to my advisor, Josh Roering, who guided me through the trials
and adventures of graduate school, and has been an inspiration to work with. Josh was
always generous with time and resources, has opened my eyes to the fascinating world of
quantitative geomorphology, all the while retaining a sharp sense of humor and great
outlook on life. I can only hope some of your insight and skill as a scientist has rubbed
off on me over the past five years. I am similarly grateful to Jim McKean and Jarg
Pettinga for fostering my interest in geomorphology as an undergraduate in New Zealand,
and for continued advice, counsel, and encouragement during my time at Oregon.
The graduate students at Oregon have ensured time in Eugene has been both
thoroughly enjoyable and academically stimulating; whether in formal group discussion,
a Friday evening at Rennie's, or around a campfire in the Cascades. Special thanks to the
fellow students, past and present, in Josh's Earth Surface Processes Laboratory.
Additionally, I remain indebted to the Geological Sciences faculty and administrative
staff for their continued help, advice, and accessibility.
This research was funded by an NSF Grant (EAR-0447190) to J. Roering, and a
NASA Earth Surface and Interior Program grant to J. Roering and D. Schmidt (UO). I
had additional support via the 'Fulbright EQC award in Natural Disaster Research' from
Fulbright and the Earthquake Commission in New Zealand.
IX
The LiDAR data of the South Fork Eel River analyzed in Chapter II was provided
by the National Center for Earth-surface Dynamics. Thanks to Peter Steel and the Angelo
Coast Range Reserve for access. Sean Bemis and Belle Philibosian helped prepare
radiocarbon samples. The main stern Eel River LiDAR data underpinning Chapters III
and IV was acquired by the National Center for Airborne Laser Mapping. John Stone,
Greg Ba1co and the University of Washington Cosmogenic Isotope Laboratory provided
invaluable advice and assistance with meteoric lOBe processing. I am especially grateful
to Calvin and Wendy Stewart of Stewart Ranch for field access, advice, and generous
hospitality over the past several years. Additional field access along the Eel River was
provided by Lone Pine Ranch and Mr. Ted Winters. Sean Bemis, Adam Booth and Laura
Stimely provided assistance with field surveying. I thank Humboldt County Public
Works, Six Rivers National Forest, USGS, and the libraries at Humboldt State University
and UC Berkeley for help with accessing aerial photos. Discussions with Harvey Kelsey
and Jay Stallman enhanced our understanding of this spectacular landscape.
None of this would have been possible without the encouragement of my
extended family and friends in New Zealand, and new friends I have made in the US.
Eternal thanks to my parents, Ian and Lachlan, and sisters, Sarah and Rachael, who have
been unwavering in their support during my time at both Canterbury and Oregon. Finally,
many thanks to Austin for your all you have done over the past year.
For Jim and Nancy
x
xi
TABLE OF CONTENTS
Chapter Page
I. INTRODUCTION 1
II. DEEP-SEATED LANDSLIDES AND HILLSLOPE BASE LEVEL CONTROL:
SPATIAL ANALYSIS USING AIRBORNE LIDAR MAPPING 5
1. Introduction............................................................................................. 5
1.1. Landslides and Landscape Evolution.............................................. 5
1.2. Channel Incision 6
1.3. Testing Landslide Response to Channel Incision 7
2. South Fork Eel River, California.............................................................. 8
3. Methods................................................................................................... 11
3.1. Geomorphic Mapping 11
3.2. Wavelet Analysis of Channel Sinuosity.......................................... 15
4. Results..................................................................................................... 19
4.1. Distribution of Landsliding 19
4.2. Distribution of Terraces.... 23
4.3. Channel Sinuosity and Landslide Location..................................... 23
5. Discussion 25
5.1. Incision-Landslide Correlation....................................................... 25
5.2. Terrace Distribution 25
5.3. Scale of Sinuosity 26
5.4. Long Term Landslide-Channel Coupling........................................ 28
5.5. Channel Incision, Landsliding and Long Term
Hillslope Evolution.......................................................... 29
6. Conclusion. 31
xii
Chapter Page
III. LONG-TERM KINEMATICS AND SEDIMENT FLUX OF AN ACTIVE
EARTHFLOW, EEL RIVER, CALIFORNIA 32
1. Introduction............................................................................................. 32
2. Eel River, Northern California 33
3. Historical Air Photos and Surveying 34
4. Meteoric lOBe and the Transport Zone "Soil Conveyor" 37
5. Results..................................................................................................... 39
6. Discussion and Conc1usions..................................................................... 40
IV. DECADAL-SCALE VARIATIONS INEARTHFLOWMOVEMENT AND
SEDIMENT PRODUCTION USING HISTORICAL AERIAL PHOTOS
AND AIRBORNE LIDAR, EEL RIVER, CALIFORNIA 43
1. Introduction 43
1.1. Earthflows - Slow Moving 'Earth Glaciers' 44
1.2. Mass Movements and Landscape Evolution 44
1.3. Earthflows - A Unique Form of Landslide 47
1.4. Key Questions.............. 48
2. Study Area - Eel River, California........................................................... 50
2.1. Introduction.......................................................... 50
2.2. Geology and Structure.................................................................... 50
2.3. Uplift Rates and Erosion 52
2.4. Landslide and Earthflow Studies.................................................... 53
2.5. Main Stem Eel River Study Area............... 54
3. Methods - Mapping Earthflow Activity with High Resolution Topography
and Photogrammetry 56
3.1. Background - Landslide Mapping.................................................. 56
3.2. LiDAR........................................................................................... 59
3.3. Photo Orthorectification................................................................. 59
3.4. Objectively Quantifying Earthflow Movement............................... 60
xiii
Chapter Page
3.5. Earthflow Displacement Time Series.............................................. 64
3.6. Climate 65
3.7. Sediment Production...................................................................... 66
4. Results..................................................................................................... 67
4.1. Spatial Distribution of Earthflows 67
4.2. Temporal Distribution of Earthflows 76
4.3. Climate 77
4.4. Earthflow Sediment Flux.... 83
5. Discussion............................................................................................... 86
5.1. Erosion and Sediment Delivery....................................... 86
5.2. LiDAR and Aerial Photos 87
5.3. Spatial Characteristics 87
5.4. Temporal Characteristics................................................................ 91
6. Conclusion............................................................................................... 94
APPENDICES 97
A. 14-C DATING OF THE SKUNK CREEK LANDSLIDE COMPLEX,
SOUTH FORK EEL RIVER, CALIFONIA 97
B. SUPPLEMENTARY MATERIAL FOR CHAPTER III... 103
C. EVIDENCE FOR A LANDSLIDE DAMMED PALEO-LAKE ALONG
THE MAIN STEM EEL RIVER........................ 109
D. AERIAL PHOTOGRAPHS USED IN CHAPTER IV 118
REFERENCES....................................................................................................... 120
Page
XIV
LIST OF FIGURES
Figure
Chapter II
1. Location map of northern California and study site............................................ 9
2. South Fork Eel River longitudinal profile, landslides, terraces and sinuosity...... 13
3. South Fork Eel River cumulative angular deflection and residuals..................... 16
4. First order Gaussian wavelet.............................................................................. 17
5. Representation of how we use wavelets to capture channel sinuosity................. 18
6. Quantifying channel curvature and bank-dependent landslide area 20
7. South Fork Eel River channel from 28 to 38 km................................................ 21
8. Cross-correlation between landslide area and wavelet coefficient 24
Chapter III
1. Kekawaka Earthflow displacement.................................................................... 35
2. Tree velocities for the three kinematic zones of the earthflow 37
Chapter IV
1. Location of the study area along main stem the Eel River 49
2. Simplified lithological map................................................................................... 51
3. Active vs. dormant earthflow morphology.............................................................. 58
xv
Figure Page
4. Unfiltered shaded relief map of the Boulder Creek earthflow 61
5. Field evidence for earthflow activity. 64
6. Oblique view of the Penstock Earthflow looking southeast 67
7. Map of the 226 krn2 study area, active earthflows, and tree displacements........... 68
8. Earthflow areas 71
9. Histogram of 998 tree displacements................................................................ 72
10. Histogram is slope of all 4 m cells within the study area 73
11. Aspect of earthflows and the study area... 75
12. Temporal velocity and maps of earthflows with sustained velocities 78
13. Temporal velocity and maps of earthflows with declining velocities 79
14. Temporal velocity and maps of earthflows with a slow, steady velocity...................... 80
15. Plots of annual water year (Oct-Sept) climate statistics recorded on the Eel River ....... 81
16. Histogram ofx2 coefficients of polynomial fits to random sampling........................... 82
17. Monthly PD~ values from 1900........................................................................ 84
18. Combined annual sediment delivery for 15 large earthflows.............................. 85
19. Conceptual earthflow evolution model................................................................... 89
-------- ------- ----
XVI
LIST OF TABLES
Table Page
1. Proportion of lithology and active earthflows (EF) across the study area..................... 74
1CHAPTER I
INTRODUCTION
A major contemporary focus in the geosciences is the study of earth surface
processes, with the objective of providing quantitative explanations and predictions for
how environmental perturbations may affect surface processes. This focus encompasses
the processes of erosion in the broadest sense - where and how sediment is produced, the
rate and mechanics of transport through the fluvial system, and how it is delivered to
floodplains, the coastal environment, and ultimately the ocean depths. Beyond simply
dictating the physical appearance of a landscape, surface processes exert primary controls
on hydrocarbon and mineral genesis and storage, the locations of arable land, building
sites and transport corridors, and can pose significant natural hazards through flooding
and landslides. In particular, significant effort is directed at constraining how surface
processes were affected by the climates of the past, and crucially, how they will be
affected by future climate changes.
For much of the twentieth century, the study of earth surface processes was
largely qualitative. Researchers were concerned with descriptions of landscapes (Chorley
et aI., 1991), and influenced by pioneering works such as Davis (1899) and Penck (1953).
In recent decades however, aided by significant advances in computing, geochronology,
and geodesy, there has been a shift towards a more quantitative understanding of the
mechanics and rates of processes, so that we can better predict landscape form and
dynamics (Dietrich et aI., 2003). For many surface processes, such as soil production,
slope dependent soil creep, and river incision, researchers have developed field verified
and physically testable mathematical statements that can quantify both the mass flux
2(erosion) attributable to a surficial processes, and the resulting landscape morphology.
When parameterized in this way, a given surface process can be incorporated into
powerful numerical landscape evolution models, which are used to test fundamental
theories about how the earth surface evolves (e.g., Perron et aI., 2009).
In non-glaciated mountainous landscapes, sediment production and hillslope form
are controlled by mass wasting processes, such as landslides and earthflows. Beyond their
popular notoriety as a formidable natural hazard, landslides exert a considerable influence
on the supply rate, grain size, and timing of sediment to the channel network and
morphology of the landscape. Landslides and earthflows have proven challenging to
analyze using the quantitative framework described above, as they do not occur
predictably in time or space, and range in size from local stream bank collapse to
mountain-scale slope failure. Given these difficulties, landslide processes are frequently
overlooked in models of landscape evolution, and none account for slow sustained mass-
movement such as earthflows (Dietrich and Perron, 2006).
Research into deep-seated landsliding typically focuses either on detailed
engineering-style studies of individual slides, or broad regional landslide inventories
(Malamud et aI., 2004). Earthflow studies have adopted a narrower focus, with research
directed almost exclusively on seasonal slide mechanics (e.g., Coe et aI., 2003; Iverson
and Major, 1987), and we know little of earthflow behavior beyond individual slopes, or
at timescales exceeding a decade (cf. Bovis and Jones, 1992; Kelsey, 1978; Maquaire et
aI., 2003; Trotter, 1993). Consequently, we struggle to quantify the sediment yield,
morphology and dynamics of an earthflow prone landscape.
With this motivation, I seek to quantify the factors which influence deep-seated
slope instability and modulate sediment yield. I address a need for extensive, detailed
maps of mass movement, which capture landslide and earthflow behavior over significant
spatio-temporal scales. Specifically, I investigate how and why deep-seated landslide and
earthflow movement varies in time and space, with a particular focus on large, slow
moving earthflows in the Eel River catchment of northern California. Methodologically, I
3underpin the research presented in the next three chapters with high-resolution digital
topography acquired via airborne light detection and ranging (LiDAR).
The upper South Fork Eel River has experienced sustained incision since the late
Pleistocene, as reflected in extensive strath terraces and abundant ancient, deep-seated
landslides (Seidl and Dietrich, 1992). In Chapter II, I document the landslide response to
this incision, and map the spatial distribution of landslides, terraces and channel incision.
I observe extensive landsliding in the incised section of the river, although the amount of
landsliding does not scale with the magnitude of incision, but rather reflects the
progressive hillslope evolution of up to 80 m of sustained incision. The lateral migration
of the channel within the valley also exerts a strong influence on landslide location,
which typically cluster at the outer edge of meander bends. Wavelet analysis of channel
sinuosity reveals that landslides are most sensitive to meanders with a wavelength of 1.5
km. This chapter provides rare field data on landslide response to channel incision, a
commonly assumed by seldom quantified process. Chapter II was co-authored with Josh
Roering, Jim McKean and Bill Dietrich.
Extensive argillaceous melange along the main stem Eel River generates abundant
slow-moving landslides spanning all states of activity and form, from small <100 m2
slumps, to huge earthflow complexes spanning 900 m of relief and extending up to 5 km
from channel to ridge (Brown and Ritter, 1971; Dwyer et aI., 1971). Chapter III describes
a study of the 1.5 km long Kekawaka earthflow. This project was designed to better
understand long-term earthflow activity and evolution, and I incorporate LiDAR,
orthorectified aerial photographs, field survey data, and a novel application of meteoric
lOBe in soil pits to measure earthflow displacement. Inventories of meteoric lOBe increase
systematically downslope, indicate an average movement rate of 2.1 ± 1.3 mla over the
past 150 years, and establish a minimum earthflow age of 1700 years. Orthorectified
aerial photos show earthflow velocities peaking in the 1960's at 3--4 mla. I calculate a
20th Century erosion rate approaching 10 rnm/a, 10-20 times greater than the estimated
regional sediment yield over that period. This study emphasizes the variability of
emthflow movement at the decadal scale, and the localized and vigorous role of active
4earthflows in landscape evolution. Chapter III was published in the journal Geology in
September 2009, volume 37(9), pages 803-806, and co-authored with Josh Roering and
Jim McKean.
Chapter IV, co-authored with Josh Roering, describes a regional study of large,
slow moving earthflows in the Eel River catchment. A 226 km2 LiDAR dataset and a
series of orthorectified aerial photos enables documentation of 122 individual active
earthflow features, including 17 large (>1.5 km long) failures with sufficient
displacement (>30 m) to generate a time series of earthflow movement. Over the period
1944-2006,7.3% of the terrain was active, and earthflows account for a 0.53 ± 0.04
mm/a erosion rate across the study area, over half the regional average sediment yield. I
suggest that when active, earthflows erode an order of magnitude faster than surrounding
terrain, however source supply limitations appear to govern long term earthflow
evolution. The displacement data, determined by systematically tracking the position of
shrubs growing on earthflow surfaces, cluster into 3 distinct temporal patterns. One group
exhibits peak velocities from 1944 to 1964 with little subsequent movement, while a
second group of earthflows sustain rapid velocities of -4m/a until the late 1970's. The
third group shows slow steady movement (-1 rnla), with occasional surges. The temporal
patterns of movement correspond with decadal-scale climatic changes recognized as the
recent cool-warm cycle of the Pacific Decadal Oscillation (1947-2008). I document
persistently cooler, wetter conditions from 1950 to the late 1970's, synchronous with the
majority of earthflow movement.
5CHAPTER II
DEEP-SEATED LANDSLIDES AND HILLSLOPE BASE
LEVEL CONTROL: SPATIAL ANALYSIS USING
AIRBORNE LIDAR MAPPING
This chapter is in preparation for submission to Journal ofGeophysical
Research - Earth Suiface. This paper is co-authored with Josh Roering,
who provided funding and advisorial help. Josh Roering, Jim McKean and
Bill Dietrich provided editorial support.
1. Introduction
1.1. Landslides and Landscape Evolution
Deep-seated landsliding is a dominant erosional process in the evolution of
mountainous landscapes - landslides can impart large perturbations on sediment budgets,
establish or modify drainage patterns, and alter hillslope morphology [Hovius et al.,
1998; Korup, 2005b; Korup, 2006a]. Despite the significance of deep seated landslides to
landscape evolution, when compared to other erosional processes (eg river incision, soil
production, or slope dependent soil creep), we continue to be challenged when attempting
to predict the landscape form and sediment flux associated with large slope failures
[Dietrich and Perron, 2006]. Much ofthis knowledge gap can be attributed to the
stochasticity of landslide behavior. Difficulties arise when extrapolating geomechanical
theories beyond an individual hillslope to the landscape scale, where the hillslope
6material properties, vegetation, destabilization processes, rate of base level change, and
triggering factors (meteorological and seismic events) are inherently spatially and
temporally variable [Dietrich et aI., 2003; Palmquist and Bible, 1980].
Many geotechnical and engineering investigations document the behavior and
characteristics of individual landslides, yet comparatively few studies have explored the
erosional and morphologic role of deep-seated landslides in a context significant to
geomorphic temporal and areal scales. Such an approach, for example, would look to
explain where and when landslides occur within drainage basins [Densmore and Hovius,
2000; Hovius et aI., 1998; Korup, 2005b; Palmquist and Bible, 1980], and ultimately
their contribution to sediment flux and impact on landscape morphology [Densmore et
aI., 1998; Hovius et aI., 2000; Meunier et aI., 2007; Ouimet et aI., 2007; Roering et aI.,
2005].
1.2. Channel Incision
Because river incision sets the pace of erosion in non-glaciated landscapes [e.g.,
Whipple, 2004], and deep-seated landslides ultimately respond to changes in base level
via channel incision, increased understanding of the nature of coupling between the
channel and deep-seated landslides may help us predict the geomorphic impact of such
landslides (such as their size, location and timing) after a given channel perturbation.
Alternatively, we would not expect landslides generated by other factors, such as
earthquakes, to show a strong signal of coupling with the channel, as they preferentially
occur on ridges where seismic accelerations are greatest [Densmore and Hovius, 2000;
Meunier et aI., 2007]. In landscapes prone to deep-seated slope failure, landsliding should
occur following channel incision, although the response may not be immediate as stress
and/or strength changes may require time to propagate upslope. Physical models give
some insight into the extent to which landslide behavior is coupled to channel incision
[Densmore et aI., 1997; Hasbargen and Paola, 2000]; however, catchment-scale deep-
seated landslide response to channel incision remains poorly documented in real
landscapes [cf. Azanon et aI., 2005; Griffiths et aI., 2005; Hovius et al., 2000; Hovius et
aI., 1998; Ouimet et aI., 2007].
7It is well established that river incision can cause hillside strength to decrease,
and/or shear stresses within adjacent slopes to increase, leading to slope failure [Terzaghi,
1950]. An important, but often elided consideration in many studies of river incision is
that rivers do not just incise vertically - lateral river incision will similarly increase
stresses on adjacent hillslopes. As a simple example, on a slope of 45°, 1 m of lateral
incision is effectively the same as 1 m vertical incision. Lateral incision can occur more
rapidly, and will be preferentially imposed on the outside of meander bends [Stark,
2006]. This pattern has been observed with small historical landslides clustering on the
outer edge of meander bends due to fluvial undercutting [Hugenholtz and Lacelle, 2004].
Although deep-seated slope failure in response to lateral channel incision may
seem axiomatic, and specific examples have been well documented [e.g., Kelsey, 1978],
it has rarely been systematically quantified throughout a landscape. There is insufficient
data to test theories relating the locations of deep seated landslides to progressive
incision, and no studies have assessed landslide response to varying scales of channel
curvature. We are not equipped to answer fundamental questions regarding how
hillslopes respond to a channel undergoing sustained incision. For example, is a channel
with a very short wavelength and tighter transient meanders more effective at generating
slope instability than river with longer wavelength, less-mobile meanders? How do
landslide prone slopes evolve under sustained channel incision?
With research focused largely on contemporary landslide movement, there is little
theory about what happens to a slope following failure [Mather et al., 2003], or how the
slope might evolve through a series of landsliding episodes [Cendrero and Dramis,
1996]. Given the long time scales of slope evolution, and the difficulty of reconstructing
pre-failure morphology, the process of slope failure in response to sustained channel
incision has not been well quantified.
1.3. Testing Landslide Response to Channel Incision
We focus on one particular hillslope erosion process, mass wasting through deep-
seated (bedrock) landslides in geotechnically weak meta-sedimentary rocks, and ask a
8first order question: does the location of landslides correspond with recent channel
incision, as recorded by strath terraces? As a null hypothesis, we assume landsliding
along the channel has no systematic pattern. To test this, we employ light detection and
ranging (LiDAR) to document the spatial pattern of landsliding and strath terraces along
the South Fork (SF) Eel River, Northern California. We explore whether the extent and
location of landsliding correlates with the relative magnitude and lateral directivity of
channel incision.
2. South Fork Eel River, California
We used -200 km2 of LiDAR data (average bare earth data density about 2.6 m- I ,
gridded to 1 m resolution) encompassing the upper South Fork Eel River catchment in
northern California, centered on the Nature Conservancy's Angelo Coast Range Reserve
(Figure la,b). The area has a Mediterranean climate, and vegetation comprises a mixture
of redwood, fir and oak forest. The lithology, summarized in Figure 2a, consists of the
Central and Coastal Belts of the Franciscan Complex [Jayko et ai., 1989], a Jurassic-
Cretaceous accrectionary prism complex renowned for its susceptibility to deep seated
landsliding [Iverson and Major, 1987; Kelsey, 1978]. Tectonically, the Neogene-to-
present migration of the Mendocino Triple Junction along the coast of northern California
has generated a double humped zone of uplift which migrates north at approximately 5
cm/yr [Furlong and Schwartz, 2004]. This manifests in the landscape as transient changes
in uplift rates and regional-scale catchment reorganization [Lock et ai., 2006].
9N
A
o 2.5
CA
Eel River
:)
an Francisco
Landslides
Terraces
r"\J SF Eel River
() River km
Figure L Location map of northern California and study site. (a) Eel River
system and the area of LiDAR coverage in the headwaters of the SF Eel River
catchment. (b) Shaded relief map of the Lidar coverage illuminated from the
NW. SF Eel River (black line) runs south to north. Mapped deep seated
landslides (red) and SF Eel River terraces (yellow) are indicated. Mapping was
confined to the slopes bounding the SF Eel River. White circles mark every 10
km of channel distance.
10
From the late Pleistocene to the present, the SF Eel River has experienced several
periods of incision, recorded by multiple levels of strath terraces that typically converge
upstream [Fuller et a!., 2009; Seidl and Dietrich, 1992]. To approximate the relative
depth of incision along the SF Eel River, we focus on a distinctive upper strath terrace
(locally equivalent to Terrace 3 in Seidl and Dietrich [1992]) whose apparent remnants
can be traced for approximately 30 km along the river profile (top of green shading in
Figure 2a). Optically stimulated luminescence dating on gravels at the base of this terrace
tread (river km 32.5) gives an age of 20.5 ± 3.3 kyr [Fuller et a!., 2009], which equates to
a bedrock incision rate of 0.3-0.4 mm/yr. Fuller et a!. [2009] use lOBe in terrace gravels
to show catchment averaged erosion rates were higher in the late Pleistocene than the
Holocene. They invoke increased deep seated landslide activity to explain these higher
erosion rates, and suggest elevated levels of sediment supply armored the channel and
promoted strath planation.
Incision into the strath terraces extends approximately half way up our study
reach, and invites comparison of landslide distribution between the un-incised (0-26 km)
and incised (26-53 km) sections of channel. Abundant dormant deep-seated landslide
features are preserved throughout the upper SF Eel catchment [Lashermes et a!., 2007],
as evident in Figures 1 and 7. There is minimal evidence of modern deep-seated activity,
such as ground fissures or bowed trees.
The landslides along the SF Eel River span a range of failure styles from large
deep-seated rotational slumps, slump-flows, to complex, multi-part earthflows. Some
ridges exhibit Sackung or antiscarp features illustrating the large-scale slope instability.
Due to the predominance of the slump failure style, on many slides the landslide mass has
not been completely eroded, and persists as a restabilized mass or bench below the
headscarp. We frequently observe extensive channel or gully incision and subsidiary
slumps on the landslide deposit. One section of the river (41-50 km) differs, in that many
of the landslides deposits have been eroded, to expose the approximate landslide failure
surface.
11
Excavations and creek exposures reveal the landslide material as a very poorly
sorted, largely unstructured colluvium, with individual clasts up to tens of meters per
side. Overlying the landslide deposits are mature redwood and fir forests, which
presumably favor the deep soils and favorable groundwater conditions afforded by the
landslide mass. The depth to the failure plane is unknown, although given scaling
relationships between slide area and volume [Hovius et aI., 1997; Malamud et al., 2004],
potentially ranges from 4 to 50 m.
As the slides are currently largely inactive, modern conditions may not be
conducive to promoting deep seated slope failure. This lack of modern activity is
plausibly attributable to various factors, including a dryer climate, a change in the rate or
directivity of channel incision, changed seismic regime, or the establishment of the dense
redwood and conifer forest in the late Holocene [Barron et aI., 2003; Pisias et aI., 2001].
Alternatively, the landslides may reflect previous perturbations, such as a rapid period of
channel incision.
3. Methods
3.1. Geomorphic Mapping
High-resolution LiDAR data enables us to map landslides and quantify the spatial
relationship between landslides and river channels across large areas. The pervasive
landslide features throughout the SF Eel River catchment are problematic to map
conventionally: dense conifer forests obscure slide features, and surficial erosion acts to
attenuate and modify the characteristic landslide morphology [Wieczorek, 1984; Wills
and McCrink, 2002]. The 'bare-earth' digital elevation model (DEM) generated by
LiDAR therefore serves as an efficient and effective tool to map subtle landslide features
across an extensive, mostly forested catchment [Van den Eeckhaut et aI., 2007].
We concentrated our analysis on the upper 53 km of the SF Eel River (extent of
LiDAR coverage - Figure 1b), and delineated the margins of landslide features located
on hillslopes bounding the river. Our mapping did not extend up tributary drainages;
12
tributaries frequently have a knickpoint or channel step at the tributary junction, and thus
do not experience the same extent of channel incision as the SF Eel River. We digitized
landslide boundaries guided by slope, contour, drainage area, surficial roughness
[McKean and Roering, 2004] and variably-rendered shaded relief maps via characteristic
landslide morphologic features including headscarps, benches, undrained depressions, toe
boundaries, drainage patterns, and residual hummocky slide mass deposits [eruden and
Varnes, 1996]. Terrain not affected by deep-seated landslides is generally restricted to the
inside of meander bends. This terrain typically has steep, well organized 'ridge and
valley' topography, conventionally associated with incision by debris flows [Stock and
Dietrich, 2006], distinguishable from the hummocky, low drainage density terrain
attributed to deep-seated landsliding [e.g., Roering et a!., 2005]. We caution that mapping
ancient deep-seated landslides in highly erodible melange terrain, even from high-
resolution LiDAR-derived DEM's, remains an interpretive exercise [Van Den Eeckhaut
et al., 2005] and acknowledge that this is the most subjective aspect of our methodology.
We grouped slide scarps and masses for each landslide within a single polygon,
and mapped complex slides as one contiguous feature. Similarly, we mapped individual
terrace treads from the flights of terraces flanking the SF Eel River, and visited dozens of
landslide and terrace sites in the field to verify correct feature classification. A
geographical information system (GIS) generated areal and elevation statistics for each of
the landslide and terrace polygons, and we projected mapped landslides and terraces to
the channel longitudinal profile, as shown in Figure 2a-c. We quantify the pattern of
terraces both in the context of height above the river (terrace elevation as the median
elevation of the individual terrace polygon), and in terms of bank-dependent average
terrace width down the channel (Figure 2b). Additionally, in Figure 2 we distinguish
landslides which impinge on the modern SF Eel River from those landslides which do not
interact with (and are disconnected from) the modern channel (e.g. landslides higher on
hillslope or adjacent to a terrace).
13
Figure 2. South Fork Eel River longitudinal profile, landslides, terraces and
sinuosity. Figure 2a-d is displayed as a function of distance (kIn) along the SF
Eel River.
(a) Stream profile of the SF Eel River. Terraces of the SF Eel River are denoted
by horizontal dashes, with elevation the median elevation of a terrace polygon.
The region shaded green from -26 kIn represents vertical incision into the upper
strath surface. Landslides are projected onto the stream profile, and are
represented by vertical bars related to elevation range, while the circle at the top
indicates relative landslide size, with the largest 0.817 kIn2 at 51 kIn. Slides
colored red have potential interaction with the modem SF Eel channel, whereas
those colored grey are disconnected. We summarize the lithology transected by
the SF Eel along the base [Jayko et aI., 1989].
(b) Individual terrace surfaces were projected to the channel and area binned
every 500 m of channel length. Terraces greater than 500 m in length were
compartmentalized and projected in pieces. Positive area is the left bank, and
negative area the right bank. Approximate incision into the upper strath surface is
shown in green. Note the anti-correlation between terraces and landslides from
12-27 kIn.
(c) All landslides are projected to the SF Eel channel. A positive area represents
the slides on the true left bank, a negative area those on the right, with the same
coupling coloring described in 2a. The curves are the coefficients of the 1.5 kIn
(solid) and 6.2 kIn (dashed) wavelets, used to approximate planform curvature at
different scales. The box denotes the section of channel depicted in Figure 7.
(d) Average slope (degrees) ± 1 standard deviation of hiIlslopes bounding the SF
Eel River. All terrace surfaces were excluded from slope calculations.
>-'
->::-
ill
-400 ~
~
-800
80 Sc
40Sl
o "0
E
800 UJ
C
<D
400 "0
~
0
- 0 u
50
50
(0"817km'~
41 42 43 44 <t.5 45 (7 48 49
40
40
~. 6 2km wavelength
-Jm
_____--'Al!~n~atjnQyaoIWCoastal.B"'e"tt'--- _
Yager compnses thin-bedded arkosIc sandstone and argilhte
(More competent than Coaslal Belt)
3\ 32 33 34 35 36 37 3S 3:9
30
30
Elder Ck
0
'
-- ~
Slo(io'l'~'VU;,t:n
l';}"!IC'
21 22 23 21, 2S 2S 27 23 29
20
Longitudinal Channel Distance (km)
20
L- "-lnse~1 m Figure 6
Longitudinal Channel Distance (km)
eo<
~'citJd
~rve~!:llrt'l:
Key . \
1 5km \V3Velength
G<q
d~'CIM
Coastal Belt Franciscan
Interbedded arkosic sandstone and argillite
Branscomb
1112131.:1 1S 16 17 13150
r
10
10
, >
.;{I-=-=-1
1 .. :3 01 S 6
qc~ l~rrr~ rJITtJ' C1~-rr~' L?--'lTI-'~_ ~~~ -~-, ~_J:L:::r~1 - -- ~
,--'-1--'1'1-- - 1 I I I t I I I I I I I, 1 I, I I I , I
1 1 :3 <4 5 6 7 S <) 11 11 13 14 15 16 11 18 19 1~ 2'2 13 101 15 2t 17 18 29 31 Z1 n 34 35 36 31 33 39 .I, 42 "3 4':' 45 <t6 .7 ~ ," 5\ 52 53
Left Bank
Right Bank
o
l"-- "" "- "-,,-, " " "" '" ,,""
51 52 53
400
200~Central B~lt Franciscan
Penetratlvely sheared
Ar9iUit~ matnx melange
800
- 600
.§.
"~
ro
>
CD
LD
(b)
01
ro -
<D E
rog
<DoD
0-roN
l:: E
~ 6 ·01
(c) 04 1Left Bank
N 02
E
~
ro
~
«
<D
:g
:gj -0 2
c
ro
IRight Bank-J
-04
(d) ~ 40
Q)"
~: 20
(/),
!-
(a)
15
3.2. Wavelet analysis of channel sinuosity
In quantitative geomorphology, wavelet analysis has been used to analyze
changes in river corridor width [Gangodagamage et ai., 2007], morphological roughness
of landslides [Booth et ai., 2009], and longitudinal variation in thalweg depth [McKean et
ai., 2008]. Here we use wavelets to characterize scales of planform river curvature. For
our purposes, wavelet analysis has advantages over other signal processing techniques
(e.g., Fourier transform) in that frequency and time data can be obtained simultaneously
(although here we substitute time for channel distance). Wavelet analysis enables us to
examine channel curvature patterns over a variety of wavelengths at all sections of the
channel, in order to determine the scale of channel sinuosity that best correlates with the
along-channel distribution of landsliding.
We constructed a channel path for the SF Eel River by routing water down the
path of steepest descent. From this profile, we sampled points every 20 m along the
thalweg of the SF Eel River (Figure 1), and connect successive points with a chord of
azimuth e(Figure 3). We calculated the change of direction (118) between successive
chords to generate a series of direction change {B }, described as !1B i =B i - Bi-I
[Ferguson, 1975], and defined positive change as a clockwise rotation. We added each
successive angle of direction change to the sum of the previous, to generate a record of
cumulative change in deflection. We used a very stiff smoothing spline to de-trend the
regional planform curvature from the signal and undertook wavelet analysis on the
residuals of the signal of cumulative change in deflection (Figure 3) to find characteristic
wavelengths of sinuosity.
16
Angular deflection (c.9) between successive chords
20m
Pomts
200
o
-200 -t---------,---------,----------,---------,,-----------,----
o
10 3020
Channel Distance (km)
40
50
Figure 3. South Fork Eel River cumulative angular deflection and residuals.
Inset diagram shows angular deflection (L~8) between chords constructed from
points every 20 m of channel length. Cumulative angular deflection (L,1.8 - grey
line) with spline fit (Lambda = 10 (3). Note how the trend mimics the channel
planform in Figure la. The residuals of spline fit (black) remove the regional
channel curvature, and we analyze this signal with wavelet analysis.
Using a first order Gaussian wavelet (Figure 4), we analyzed the residual
direction change data (Figure 3) with a 1-D continuous wavelet transform. Figure 5
details how the wavelet captures channel sinuosity. Compared with other conventional
wavelet shapes, the simple and smoothly varying form of the Gaussian (Figure 4) best
approximates the form of a planform channel 's-bend'. To relate wavelet length-scale
(amount of dilation of the wavelet) to the sampling period of the signal (i1, the 20 m
sampling interval) and therefore back to channel length, we calculated the 'center
frequency' (Fe) and corresponding 'pseudo period' (Ta) of the Gaussian wavelet as
illustrated in Figure 4 [Abry, 1997; Zhou, 2003]. The center frequency is the frequency
maximizing the Fourier transform of the mother wavelet, and equates to 0.2 signal units
for the 1st order Gaussian. Fe is related to pseudo-frequency (Fa) at a given scale or
dilation of the wavelet (a) and sampling period (i1) by the relationship:
17
F=~
a a./);. (1)
Finally, the pseudo period, Ta, is simply the inverse of Fa. For example, in our
analysis a wavelet scale (a) of20, stretches the wavelet over 200 signal units, or 4 km of
longitudinal channel length with /);. =0.02 km. The a of 20 and Fe of 0.2 generates a
pseudo frequency of 0.5 km- I , and pseudo-period (or wavelength) of 2 km. The pseudo-
period relates to channel distance as the wavelength of an idealized planform s-bend, and
is how we henceforth quantify and term wavelet length scale. We stress this is a different
definition to the conventional definition of wavelength, being the direct distance between
similar points on the waveform.
11 st Order Gaussian Wavelet I
5
"-
\
\
\
\
43
/
I
I
I
I
21a
Signal Units
-1
0.8
0.4
a I
\ I
\ I
-0.4 \ J
"- /
"
,/
-0.8
-5 -4 -3 -2
Figure 4. First order Gaussian wavelet (solid line - effective support [-5, 5]).
The center frequency based approximation has a frequency of 0.2, and pseudo-
period of 5 (dashed line). We appropriate pseudo-period as the channel length of
an 's-bend' at a given wavelet scale.
18
Idealized planform channel Bend 2
(a)
~ CD ](b) ~ <l r
::J
u
c <J)
.~ ~(c) ~ ~
ro ro(95:
(e) l~ ].-~8
Bend 1
Channel Distance
Bend 2
Figure 5. Representation of how we use wavelets to capture channel sinuosity,
illustrated at two channel bends. Sa is in map view, whereas the x-axis of plots
5b-e is channel distance.
(a) Planform view of an idealized sinuous channel, with 2 opposite bends
highlighted.
(b) Plot of cumulative change in direction between successive chords
(degrees) along the channel. At the apex of a meander bend, the rate of
change in direction is greatest, and these points correspond with
inflection points on this plot of cumulative L10.
(c) The wavelet calculates coefficients at all points along the channel, but
here we show a representation of the wavelet at 2 points correlating to
the two channel bends from Sa. At Bend I the left wavelet is in phase
with the cumulative azimuth signal and returns high coefficients,
whereas at Bend 2 the wavelet is out of phase and returns large negative
coefficients
(d) Coefficients of the wavelet transform - note the high coefficients relating
to the apex of bends, such as at Bend I and Bend 2.
(e) Due to the way we defined directionality (clockwise channel turn
positive), the wavelet coefficients finally need to be inverted to give a
representation of the channel sinuosity at a given wavelet length scale.
19
Large absolute wavelet coefficients indicate that the local river planform is well
approximated by the Gaussian wavelet at the designated wavelet length scale. Positive
wavelet coefficients (clockwise river bend) correlate with positive (true left bank)
landslide area, and vice versa (Figure 6). To determine which scale of sinuosity best
correlates with the along channel distribution of landsliding, we cross-correlated wavelet
coefficients with bank dependent landslide area for a range of wavelet length-scales. We
excluded the slide at 52 km from the correlation as its large size exerted too much
influence on the regression. We analyzed both the whole channel length, and the un-
incised (0-26 km) and incised sections (26-53 km) separately. While it is ambitious to
expect a linear relationship between coefficients and landslide area, if landslides do
cluster on the outer edge of meander bends we would expect points to plot in opposite
quadrants, as shown conceptually in Figure 6. The top right quadrant relates to landslides
on the left bank, and clockwise river bends, whereas landslides on the right bank and
anti-clockwise river bends will fall within the lower left quadrant. Linear regression is a
simple way to quantify the clustering of points in the opposite quadrants, even though the
correlation coefficients are generally low.
4. Results
4.1. Distribution of Landsliding
We mapped 87 deep-seated landslide features along the 53 km length of the SF
Eel River as shown on Figures 1 and 2. Along 0-26 km, we identified 29 landslides,
which have an average area of 0.047 km2 There is a notable concentration of landslides
from 0 to 10 km where the SF Eel crosses the highly erodable Central Belt of the
Franciscan Complex (Figure 2a). In contrast, along 11-27 km, there are only 9 mapped
landslides, and this section coincides with low channel gradient, and extensive broad
terraces. From 26 to 53 km, the section of the profile with evidence of channel incision,
we mapped 58 deep seated landslides, with an average area of 0.096 km2.
20
Left
bank
(positive)
Regression
line
•
Landslide Area
•
Q)
.§ OJ
i5-g:Ei
-§ ~ ~
:6 ~ C---. . --Y
~ ~ Right
bank
(negative)
Right bank
landslides
Left bank
landslides
Figure 6. Quantifying channel curvature and bank-dependent landslide area. The
cartoon on the left shows an idealized section of river with landslides clustering
at the apex of meander bends. The graph represents how this would look when
curvature is plotted against landslide area at a specific wavelet lengthscale. We
use the correlation coefficient to quantify the correlation between landslides and
varying scales of planform curvature.
As colored red in Figure 2, we observed that slides which adjoin the modern
channel occur preferentially from 0 to 12 km, and along the incised channel section from
27 to 52 km. From 12 to 27 km, only one of the landslides has interaction with the
modern channel (at 20.1 km). Towards the far end of the canyon (50-53 km) are several
large slumps, the largest at 51 km with an area of 0.817 km2. Qualitatively, the landslides
between 28 and 36 km appear the most recently active, based on freshest (or roughest)
surface features (Figure 7).
21
Figure 7. South Fork Eel River channel from 28 to 38 km. The channel is color
coded by the absolute coefficients of the 1.5 km wavelet, with the warmer colors
denoting larger coefficients. The inset shows the relevant part of Figure 2c, and
highlights the effectiveness with which wavelet coefficients can replicate the
channel planform, and the high correlation between landslides and the outer edge
of meander bends. Note how the 1.5 km scale of sinuosity is superimposed on the
larger scale (-6 km) valley sinuosity.
22
SF Eel River
~ 1.5km Wavelet
coefficients I
(Warm = high
absolute coefs)
o Riverkm
0.4
.1 Left Bank - 800 N.
- AiN 0.2 1.5km 400
E wavelet6
C'il
Ql 0
-<
Ql
:2
(jj
-02
-0
c \ sC'il
...J
I 6.2km-0.4 wavelet - ·800
I Right Bank
t
28 29 30 31 32 33 34 35 36 37 38
Channel Distance (km)
23
4.2. Distribution of Terraces
Terraces persist along much of the SF Eel River (Figure 2a,b), and are most
areally extensive along 10-26 km, where they occur on both sides of the river.
Downstream of 26 krn, terrace remnants are smaller and less continuous than upstream,
and characteristically alternate from bank to bank. Some comparatively high but isolated
terrace features are preserved sporadically along the length of the river (e.g., the terrace
60 m above the channel at 25 krn), but these likely persist in the landscape from periods
of channel migration that long pre-date the late Pleistocene-Holocene focus of this study.
Figure 2(a) implies some incision into the gently sloping broad terraces between 11 krn
and 23 krn, although much of this is a methodological artifact and primarily attributable
to the median elevation of broad, gently sloping terraces, as opposed to channel incision.
Between 26 km and 39 krn the channel becomes more sinuous, the depth of
incision into the upper strath terrace gradually increases, and landsliding is more
prevalent. At 39.3 krn, the SF Eel River's drainage area approximately doubles at the
confluence ofTen Mile Creek, and the SF Eel gradient increases sharply, as does the
elevation difference between the channel and the reference strath terrace (Figure 2a,b).
Downstream of 39 krn the SF Eel River is confined to a steep, sinuous canyon (Figures
1b, 2d), and only isolated remnants of the terrace surfaces remain.
4.3. Channel Sinuosity and Landslide Location
Figure 8 shows the results of cross-correlation between wavelet coefficients and
landslide location over a range of wavelet length-scales. We analyzed the whole channel,
and the upper and lower sections of the channel separately, to see if the pattern of
sinuosity and landsliding differed in the section which has been incised (downstream of
26 krn). The upstream and downstream reaches exhibit different patterns of cross-
correlation with changing meander wavelength. Along the 0-26 krn section of the profile
the channel is generally straight, with small-scale intra-floodplain sinuosity. Small
wavelengths (pseudo-period 0.5 krn) have the highest correlation with landslide location.
Downstream of 26 krn, the channel is more sinuous, and the cross-correlation exhibited a
peak at a wavelength of 1.5 km. The spatial pattern along the whole channel length
24
mimics the 26-53 km pattern due to the larger number of slides in the regression. Figure
2c illustrates the optimal agreement between the 1.5 km period wavelet and the bank-
dependent landslide location along the lower section of the river. The channel in Figure 7
is color coded by absolute wavelet coefficients of the 1.5 km wavelength wavelet (warm
colors == high coefficient), and the warm colors at the apex of river bends frequently
correspond with landslide locations.
-600 --+---,-----+------,-----1
,-------------------
0.5
Left bank
0.25
o
Landslide Area (km2)
-0.25
-0.5
IfCl
600 •~ , Right bank
:9.. 300 ... • //
. ..--,.....1:: • .//
Q) • ./;g .~ • ~//.
Q) o-t--------.h''''-------~_
o /~.r
(.) . /.-.
Q) ",.-e/ ••~.i -300 ./ ., ·i',
0.8
:s 0.6
C
Q)
·13
~
o() 0.4
c
o
~
~8 0.2
o 4 8 12
Wavelet Lengthscale or Pseudo Period (km)
16
Figure 8. Cross-correlation between landslide area and wavelet coefficient over a
range of wavelet lengthscales. The regression of the 1.5 km wavelet over channel
distance 26-53 km is shown in the inset.
The largest absolute wavelet coefficients are associated with wavelet pseudo-
period 6.2 km, centered on channel distance 36.8 km. This is where the river planform
25
best approximates the Gaussian derivative, and corresponds to a section of the river with
entrenched large-scale sinuosity, as seen of Figures 1 and 7. The inset in Figure 7 shows
the finer scale channel sinuosity captured by the 1.5 krn wavelength, which is super-
imposed on the larger 6.2 km wavelength.
5. Discussion
5.1. Incision-Landslide Correlation
The pattern of deep-seated landsliding along the SF Eel River, as documented
from LiDAR, supports the notion that base level fall via channel incision has increased
the incidence of landsliding. This is illustrated by sections of channel with late
Pleistocene channel incision having a greater occurrence of along-channellandsliding.
Upstream portions of our study area that have not been subject to recent channel incision
(0-26 krn) do not exhibit widespread or systematic landsliding when compared to the
incised section of the SF Eel River (26-53 km), which has twice as many landslide
features per unit of channel length. We note a strong tendency for landslides to cluster on
the outer edge of river bends where the river incises laterally. The high incidence of
landslides on the outer edge of channel bends may result from lateral incision at the base
of the slope. Other considerations include the greater area of available hillslope in these
locales upon which landsliding can occur, and hydrological considerations such as the
concentration of groundwater flows and increased pore pressures in areas of topographic
convergence.
5.2. Terrace Distribution
The apparent linearity of the upper terrace treads from 26 to 52 km suggests the
treads are remnants of a once extensive and possibly continuous terrace surface. We use
this as a reference surface to quantify channel incision along the river. The systematic
decrease in the amount of river incision into this surface from 52 to 26 krn suggests that a
pulse of incision has migrated headward, although the cause of this incision (tectonic,
eustatic or climatic in origin) remains unknown. Within the zone of incision there are
26
lower flights of terraces which also converge upstream, indicating that the SF Eel River
has experienced several pulses of incision and strath cutting subsequent to the planation
of the upper strath. In addition to the change in gradient at the Ten Mile Ck confluence,
we note the more subtle inflection in the river profile at 24 km, where incision into the
upper strath terrace pinches out.
The paucity of landsliding from 12 to 28 km corresponds with extensive terraces.
Along this reach of the channel the terraces have effectively isolated most of the
hillslopes from any recent or future channel incision.
5.3. Scale of Sinuosity
We found wavelet analysis to be an effective new method by which to
characterize different scales of sinuosity along the channel profile. Our technique could
be replicated on any chmmel for which high-resolution topographic data is available.
The Gaussian wavelet with pseudo-period (wavelength) of 1.5 km had the most
meaningful correlation with the along channel distribution of mapped landslides (Figure
2c). River meanders with a scale of -1.5 km (longitudinal channel length of an 'S-bend')
may reflect a critical length scale for generating landsliding in the actively incising
section of the SF Eel River. This length scale likely reflects a compromise between
minimizing the radius of channel curvature that can persist for geomorphic timescales,
whilst maximizing the width of hillslope base that is exposed to lateral incision on the
outer edge of a meander bend, thereby having the greatest probability of undercutting and
destabilizing a significant section of terrain. A meander with a small wavelength can have
a very small radius, and exert a lot of hydraulic energy at the outer bank, but only over a
short continuous length of the slope base. Very small radius meanders are also more
likely to be modified by natural river evolution, and are unlikely to persist in the
landscape. Conversely, a large wavelength meander may abut a significant length of
slope, but as the bend has a larger radius, the migration rate and hydraulic energy directed
at the slope will be comparatively less than a bend with smaller radius. Currently, we
have little ability in predicting where deep seated landslides will occur in a landscape
27
beyond basic considerations of slope, drainage area and rock properties [Densmore et aI.,
1998; Korup et aI., 2007; Niemi et aI., 2005]. Identifying critical length scales of
meandering in a particular landscape may become an important and readily quantified
metric in landscape evolution models.
The most sinuous section of river (33-42 km) has a characteristic wavelength of
6.2 km. Although this scale of sinuosity did not correlate as well with landsliding as did
smaller wavelengths, we note along 31-38 km (Figure 7) there is good agreement
between the 6.2 km coefficients and landslide location. The -6 km wavelength of
sinuosity in this stretch represents the entrenched river course, especially from 30 to 45
km, and would not have changed significantly since the late Pleistocene, as the strath
terraces follow a similar pattern. In contrast, smaller-scale sinuosity (1.5 km) is
superimposed on the -6 km scale of valley meandering, and at this smaller scale the
channel can incise into the strath terraces, undercutting adjacent hillslopes within the
architecture of the larger scale valley sinuosity. This scale of lateral channel activity
appears to impart a dominant control on the location of mappable landslides.
Very short wavelengths of sinuosity best correlated with landsliding from 0 to 26
km (Figure 8), although we do not view this result as significant. The correlation was
strongest at the 0.5 km length-scale, but this is approaching the scale of intra-channel
noise, and the high correlation is likely a reflection of increased degrees of freedom with
the smaller meander scale as opposed to a physical control on landslide location. There
were only 29 landslides along the 0-26 km river section, and the correlation is not as
robust in comparison to the lower half of the profile which has a greater population of
landslides. As such, we downplay the role of modern lateral channel migration as a
significant control on landslides along the 0-26 km section of the river. Additionally,
there is little evidence of recent vertical incision along 0-26 km, significant terraces and
flood plains isolate much of the hillslopes from the river from 10 to 24 km, and only
rarely could the river be considered to influence the adjacent slopes (e.g., 20.1 km). The
high concentration of landsliding from 0 to 10 km corresponds with the very weak
argillaceous Central Belt melange, which has a very different character to the Coastal
28
Belt and Yager lithology encountered by the SF Eel downstream of 10 km (Figure 2). For
much of the 0-10 km section, the river is confined in a narrow valley, in constant
connection with the weak adjacent hillslopes. This tight connection between the SF Eel
River and adjacent landslides likely illustrates how weak rocks are unable to resist
failure, as opposed to a signal of active channel incision. As such, weak lithology can
potentially overwhelm other landscape signals.
5.4. Long Term Landslide-Channel Coupling
Most studies of hillslope-channel coupling to date have looked at smaller scale
catchments, with active, quantifiable movement, where the interactions between the
channel and slope failures are readily discerned (See for example Harvey [2001] and
references therein). Our study is complicated by an extensive study area, large, mostly
dormant slope failures (lacking in chronology), a time frame tentatively spanning the
Late Pleistocene - Present, and poor constraints on the rate or homogeneity of incision.
Despite these challenges, the linkages we find between channel incision and deep-seated
landsliding are instructive when considering channel-landslide coupling over longer time
periods.
Landslides along the lower reach of our study area in the SF Eel River
preferentially occur in areas subject to vertical and/or lateral channel incision, so base
level fall appears a primary factor in the landslides' ultimate failure, as we would predict
from fundamental geomechanics [Terzaghi, 1950]. In Figure 2a we distinguished the
slides that extend onto the floodplain of the modern channel. We consider these slides to
be influenced by recent incision of the SF Eel River, as the landslide mass extends below
the projected terrace elevation. Of the 26 slides not considered coupled in Figure 2a, 22
appear to have evidence of previous interaction with the SF Eel River. For example, most
slides occur at the edge of strath terraces (see for example the two small slides at -28.5
km, Figure 7), and there are very few landslides which are restricted to higher parts of the
hillslope, away from the direct influence of the SF Eel River during its strath cutting
period. The comparative scarcity of landslides higher up on the slope would suggest that
base level exerts a greater control on deep-seated landslide location than seismic
29
acceleration in this landscape [Densmore and Hovius, 2000]. The hillslopes adjacent to
the SF Eel terraces would have been undercut by the river at the time of strath planation
as the river eroded laterally, but have since become isolated as the river has incised, the
terrace effectively acting as a buffer between channel incision and hillslope. Without age
control on these landslides or the time of terrace formation it is speculative to say the two
events were coeval, but the presence of many qualitatively old slide features at the
margins of strath terraces may reflect tight coupling between the SF Eel and hillslopes
during strath cutting. The controls on the cyclical process of strath terrace planation and
subsequent channel incision are yet to be fully resolved, but once a river starts cutting a
strath and undermining hillslopes, the supply of sediment from activated landslides may
represent a positive feedback, with coupled landslides at the strath margin providing tools
and abundant sediment to inhibit vertical incision and promote further strath cutting
[Finnegan et aI., 2007; Fuller et aI., 2009].
There are many large dormant landslides with toes encroaching onto the modern
SF Eel channel, such as those seen in Figure 7. If these landslides were active, we would
expect to observe perturbations in the channel profile or bed-form attributable to
landslide toe deposits [Korup, 2005a]. The lack of deep seated landslide activity under
modern environmental conditions is a defining feature of this study area. We observe a
network of gullies and small creeks eroding into the dormant mass of these landslides,
and this secondary reworking of the landslide deposit may be the primary means of
removing sediment from the system, rather than direct sediment delivery from the
landslide to river.
5.5. Channel Incision, Landsliding and Long Term Hillslope Evolution
Poor constraints on incision/landslide chronology inhibit predictions as to whether
there is a critical amount of incision required to generate slope failure, or whether there is
a characteristic lag period between incision and slope failure. The most recent incision
(e.g., 15-20 m at 28-30 km) corresponds with the terrace dated to 20.5 kyr [Fuller et aI.,
2009], and what qualitatively appear to be some of the most recent landslides along the
profile. This observation suggests that -15-20 m of channel incision adjacent to a
30
hillslope may be sufficient to generate large-scale slope failure in this landscape,
although it is difficult to attribute causality given the changing climate and vegetation
over the Holocene. Farther downstream, the amount of incision into the upper strath
increases, and the character of the landslide deposits change. In the channel reach below
the Ten Mile Creek confluence (41-50 kIn), most of the landslide deposits have been
eroded. Slopes along this section of the river are steep and planar (Figure 2d), and deep-
seated slumps are rare. Near the end of the profile, the slumping failure style returns as
seen in the large slides at 51 and 52 kIn.
The strath terraces which converge upstream indicate that incision is migrating
headward. Therefore, the farther down the profile, the longer a section of the river has
been undergoing channel incision into the upper strath. This observation allows a
simplistic space-for-time substitution [e.g., Hilley and Arrowsmith, 2008], and we can
speculate about the long term cyclical hillslope response to channel incision. We propose
that initial channel incision generates large slope slumps where the river abuts the
hillslope (28-38 km). Once the landslide has slumped and stabilized, the landslide
deposit is gradually removed by subsidiary erosion processes. As channel incision
proceeds and the channel migrates within the valley, the strath is incrementally removed,
all original landslide debris is eroded, and the hillslopes are progressively steepened by
shallower planar landsliding (41-48 kIn). During this time, we speculate that weathering
proceeds into the recently exposed bedrock as channel incision increases. Eventually,
stresses in the bedrock have accumulated and/or weathering has decreased mountain
scale rock strength sufficiently [Schmidt and Montgomery, 1995], to generate a second
generation of large slump failures, such as we see at 50-52 km. Qualitative hillslope
evolution models have been proposed in the past [e.g., Cendrero and Dramis, 1996;
Hugenholtz and Lacelle, 2004; Palmquist and Bible, 1980], but generally conclude at
slope failure, and don't explore slope evolution under sustained incision.
31
6. Conclusion
Predicting the size, location and timing of deep-seated landslides remains a
challenging problem, and much research needs to be done before we can articulate a
comprehensive geomorphic transport law for deep-seated landslides. We constrained our
analysis to the spatial pattern of deep-seated landslides in response to channel incision, as
opposed to determining rates of landslide generated sediment flux - the quantity a
sediment transport law would seek to address. Our key findings, that deep-seated
landsliding correlates well with channel incision and the outer edge of meander bends are
not unexpected, but these relations have seldom been systematically quantified in a
manner that could eventually be incorporated into an expression for geomorphic transport
by deep seated landslides. Channel meanders with a wavelength of 1.5 km best explained
the distribution of along-channel landsliding in this catchment, and we suspect this
optimal wavelength will vary between different landscapes, drainage areas, lithology, and
river systems.
We propose a long term model of hillslope evolution in a slump-prone landscape
undergoing continual base level fall. Initial incision induces deep-seated slumping. The
landslide mass stabilizes, and is removed by smaller scale erosion processes, exposing the
failure surface. Weathering penetrates into the hillslope, and after continual channel
incision we predict an additional episode of deep seated slumping.
32
CHAPTER III
LONG-TERM KINEMATICS AND SEDIMENT FLUX OF AN ACTIVE
EARTHFLOW, EEL RIVER, CALIFORNIA
This paper was published in the journal Geology in September 2009,
volume 37(9), pages 803-806. This paper was co-authored with Josh
Roering and Jim McKean. Josh provided funding, advisorial, and editorial
support. Jim provided editorial and field support.
1. INTRODUCTION
Large, slow moving «2 rn/yr) glacier-like earthflows are pervasive in many
rapidly eroding landscapes world wide, most commonly in fine grained marine
sediments. They are extremely problematic for land management, transport corridors, and
represent major point sources of sediment to the channel network, yet we have little
knowledge of earthflow behavior beyond the decadal scale. While factors such as
precipitation, temperature, topographic loading or toe erosion are widely recognized to
influence motion (Keefer and Johnson, 1983), the geomorphic role of this style of slow,
persistent mass wasting as an erosional process remains poorly constrained, especially
when compared to other processes such as bedrock channel incision, debris flows or soil
creep (Dietrich and Perron, 2006). The majority of earthflow research has been directed
at interpreting earthflow mechanics by constraining movement patterns over seasonal-to-
annual monitoring periods, and this approach yields excellent data on contemporary
earthflow behavior (e.g., Coe et aL, 2003; Iverson and Major, 1987; Zhang et al., 1991).
33
With notable exceptions (Bovis and Jones, 1992; Kelsey, 1978), comparatively
little focus has been applied to the longer term evolution of individual earthflows. Active
earthflows can deflate their source basins orders of magnitude faster than regional erosion
rates (Kelsey, 1978), and consequently, earthflow movement at a given location will
likely be episodic. As such, active earthflows may compose a small percentage of the
landscape, yet account for a large fraction of the regional sediment flux. Attesting to the
ephemeral nature of earthflow activity, much of an earthflow-prone landscape is
imprinted with ubiquitous subtle headscarps, toes, and deflated transport zones from
multiple generations of earthflows and landslides in various stages of dormancy (e.g.,
Bovis and Jones, 1992). Factors controlling long-term earthflow evolution include
availability of easily mobilized source material, base level control and slope buttressing,
climatic variability, and land-use practices. To better understand the relative importance
of such factors on earthflow transport history and evolution, we aim to constrain the long
term sediment flux, kinematics, and topographic development of an individual earthflow.
2. EEL RIVER, NORTHERN CALIFORNIA
The coastal ranges of northern California are composed of penetratively sheared
meta-sedimentary interbedded sandstone and argillite of the Franciscan Melange, a
Jurassic-Cretaceous accretionary prism complex, renowned for earthflows and deep
seated landsliding (Iverson and Major, 1987; Kelsey, 1978). Some of the most extensive
earthflow activity in northern California occurs along the main stem of the Eel River and
its tributaries between Dos Rios and Alderpoint (Brown and Ritter, 1971). Highly
erodible Central Belt Melange and high seasonal rainfall predispose this section of the
Eel River to ubiquitous earthflows. Here, we observe slow-moving landslides spanning
all states of activity and form, from small <100 m2 slumps, to huge earthflow complexes
spanning 900 m of relief and extending up to 5 km from channel to ridge.
We focused on a 1.5 km-Iong active earthflow (0.2 km2 area) entering Kekawaka
Creek, an 85 km2 tributary of the Eel River (Fig. 1). This active earthflow has classic
morphology (Keefer and Johnson, 1983), including: a steep, amphitheater-like source
34
area; a narrow, elongate transport zone which over-rides several bedrock steps; and
termination at the creek as a bulbous toe (Fig. 1- Appendix B). Specifically, we study the
upper half of the earthflow where the surface has minimal disturbance and little lateral
input from creeks and other earthflows. Based on morphology, we sub-divide it into three
kinematic units; a steep headscarp or source area, a long narrow transport zone, and a
mid-slide compressional zone with reduced topographic gradient (Fig. 1). The headscarp
of the earthflow currently abuts a broad ridgeline, underlain by sandstone.
3. HISTORICAL AIR PHOTOS AND SURVEYING
We use several approaches to study movement of the Kekawaka earthflow.
Underpinning the analysis is high resolution (1m grid spacing) digital topography derived
from airborne laser scanning (LiDAR) that reveals earthflow morphology in
unprecedented detail (Fig. 1, Fig 1 - Appendix B). To determine modem earthflow
movement, we placed 27 0.9 m rebar stakes on the slide, and recorded their position
using a total station for 3 successive summers (2006-2008).
To quantify decadal scale slide deformation, we rectified historical air photos to
remove lens distortions and topographic effects (Wolf and Dewitt, 2000) and measured
offset features in sequential photographs (Baum et aI., 1998, Walstra et aI., 2007). We
obtained aerial photos taken in 1944, 1964, 1968, 1981, and 1991, in addition to the
LiDAR coverage flown in 2006, defining 5 time intervals of landslide change (Table 1,
Appendix B). We identified 53 individual trees growing on the earthflow surface and
tracked their positions (and thus velocities) from 1944 to 2006 (Figs. 1-2). The sequential
location of trees on stable ground adjacent to the earthflow enabled us to constrain errors
due to imperfect rectification or vegetation change (Fig. 2, Appendix B).
35
Figure 1. Kekawaka Earthflow displacement. Colored lines show trajectories of
trees growing on the earthflow from 1944 to 2006, which we demarcate into 3
kinematic units. The inset shows depth profiles of meteoric l~e concentration.
Vertical error bars show the sample interval, and horizontal bars the analytical
error. The background image is LiDAR derived slope (l m resolution). Inset map
shows site location in northern California.
36
:t
+ +
.,
+
Age: 400 +1· 80
Age: 250 +1- 50
2X107 3x107
, '-'=
+
Age: 620 +1· 130
+
+
-+-,
t• •Age: >1700 +1· 400
+
.
Age: >12000 +1· 2500
Pit 1: 'OBe (atoms.em.J)
50
200
Pit 3: 'oBe (atoms.em"')
OX10o 107 2x107 3x107
O-t---'---'---------'----'------,,±,-
200
Pit 2: 'oBe (atoms.em.J)
Ox100 107 2x107 3x107
o +---'-------===---
*
E 50 /
~
..c: 100Q.
ID
a 150
E 50 +
~
..c: 100Q.
l3 150
Pit 4: 'oBe (atoms.em"')
OX10o 4x107 8x107E 0
~
..c: 50
Q.
l3 100
•
,
~J
I
.1/
Vl?·S~
'v ~ ~
(1) '$.$
e S·~ •;:j.::iS
'" / . 0 &bRoad
,.... -' . &j .:r ~i: {J~:~'~:' /r-------"'s'=-oj--=,p-it---------,
Meteoric 10Se inventories
Crown: 'OBe (atoms.em.J)
~1~ ~1~ ~1~
o +-..L--'---'----'-----L..,~-'--'
•
+
40.111 0
- 23.48850
•§~IV;I
~~
·0 I .~*~
J ' J* ,J 1/ ~
I 4,' ...... iii~~
'. ~:S
...... ..0 §
..... .. ~
100.... ·· 200 m
Stable feature
Velocity
(m/yr)
0.0 - 1.0
1.0 - 2.0
2.0 - 3.0
3.0 -- 4.0
SOm contour~
Soil pit *
Flow direction 0
Kinematic
boundary
Displacement track
1944 '64~ '81~ _ 2006
\. '68 t ·91....
PosItion Displacement
in photo vector
37
4
Source zone
-- Transport zone
-- Com ression zone
3
10Se derived 150 year
transport-zone velocity
_ _ _ _ "'pe~e!.n .£it~..2 ~c!.]
1
o
1940 1950 1960 1970 1980
Year
1990 2000
Total station
sUlveying
(2006-2008)
~
2010
Figlue 2, Tree velocities for the three kinematic zones of the earthflow for each
photo interval, in addition to the 2006-2008 total station survey data. We plot the
individual velocity of each tree (faint lines), and the mean and standard error for
each kinematic unit (bold lines).
4. lVlETE RIC lOBE AND THE TRANSPORT ZONE "SOIL CONVEYOR"
Based on field and LiDAR observations, we envisage the earthflow transport zone
as a relatively undeformed soil conveyor that progressively translates material generated
and mixed in the source area to the compressional zone. Because it can take 100s to
1,000s of years for soil to move through an earthflow, this novel application of meteoric
lOBe has the potential to substantially extend historical records of earthflow movement.
Meteoric lOBe is produced in the atmosphere by the spallation (fragmentation) of
atmospheric gas nuclei from cosmic rays (Monaghan and Elmore, 1994). Once produced,
IQBe (half life of 1.51 x 106 years) attaches to sub-micron aerosol particles in the
atmosphere. These particles are scavenged by precipitation, and delivered to the Earth's
38
surface. The lOBe nuclide is rapidly adsorbed and sequestered when it contacts fine
grained soils, which typically have a high surface area per volume, and high cation
exchange capacity (McKean et aI., 1993; Pavich et aI., 1986; You et aI., 1989). lOBe can
be lost through groundwater leaching or clay translocation to lower soil horizons (Pavich
et aI., 1986) and is readily mobilized under acidic soil conditions. Soils derived from the
Central belt Franciscan melange are ideal reservoirs for lOBe because they are clay rich
and relatively non-acidic.
By measuring the inventory of lOBe through soil profiles and dividing by the
annual lOBe delivery rate, we calculated the age of several earthflow surfaces. Below the
base of the steep source zone, which experiences pervasive block rotation, ravel, and soil
slips, the reconstituted earthflow material exhibits a broad, planar morphology,
demarcating the top of the transport zone "soil conveyor." At this location, we expect
concentrations of lOBe to be low given the extent of source zone reworking and thus
dilution of previously stable lOBe enriched soil incorporated into the earthflow. Through
the transport zone, we assume that the surface remains relatively intact such that local
inventories of lOBe progressively increase downslope. The ratio of the distance between
two sample locations and the difference in their surface ages provides an estimate the
long-term velocity of the transport zone.
For this approach, we assume the following: 1) rainfall and lOBe production are
relatively constant with time, 2) minimal lOBe is lost once it adheres to the soil (e.g., via
leaching), 3) the cation adsorption capacity of the soil is not exhausted, and 4) slide
movement occurs by simple plane strain. Below the headscarp, incomplete mixing of
previously stable surface soils with high lOBe concentrations may cause pockets of
concentrated lOBe to persist. Such pockets should be readily identifiable because the
progressive accumulation of lOBe in stable, clay-rich surfaces tends to generate well-
behaved exponential concentration profiles.
We dug 4 soil pits along the axis of the earthflow, and 1 on unfailed terrain above
the headscarp crown (Pit C, Fig. 1). The earthflow material is colluvium (poorly sorted
gravelly clayey sand with occasional large rocks 1-5 m in diameter), whereas the soil
39
above the headscarp was a loamy sand. We sampled soil at regular depth intervals in each
pit, and collected intact samples to determine representative bulk density of the earthflow
colluvium (2.1 ± 0.1 g/cm\ We extracted lOBe from the soil samples following the soil
fusion technique described in Stone (1998) at the University of Washington Cosmogenic
Isotope Laboratory. The lOBelBe ratio was analyzed via mass spectrometry at PRIME
Lab, Purdue University. By integrating through the concentration profile, we calculated
the inventory of lOBe for each sample site (atoms/cm2). We used the average annual
rainfall recorded at Alderpoint (15 km NW) of 1.3 m ± 0.06 (mean ± s.e., 1940-1980)
and an estimate for mid-latitude meteoric lOBe production of 1.3 x 106 ± 20%
atoms/cm2/yr per meter of rain (Pavich et aI., 1985; Pavich and Chadwick, 1997) to
quantify the surface exposure age of the earthflow soil.
5. RESULTS
Our surveying and photographic data show that earthflow velocities vary
significantly between the 3 kinematic units (Figs. 1-2). Maximum slide velocities
averaged across each kinematic zone are lowest in the upper source zone «1 mlyr),
fastest through the narrow transport zone (2.5 mlyr), and moderate through the mid-slide
compressional zone (1.8 mlyr). Since 1944, the average velocity in the transport zone is
1.7 m/yr, although temporal variation was substantial. Velocities were highest in the late
1960's and 1970's and have steadily decreased to <1 mlyr since. This temporal pattern is
broadly consistent over all three kinematic units (Fig. 2). The headscarp exhibited
negligible expansion over the photographic record, whereas the mid-slope toe advanced
over 40 ill.
Our lOBe analysis (Fig.l, Table 2 - Appendix B) demonstrates that the unfailed
soil above the headscarp (Pit C) has the highest inventory of lOBe and a minimum surface
age of 12ka. On the earthflow, Pits 1-3 show the majority of lOBe is retained in the upper
40 cm of soil, although Pit 1 in the source zone has an anomalous high concentration at
90 cm. Pits 2 and 3 in the transport zone show a difference in surface age of 150 ± 90 yr,
which can be used to derive a long-term slide velocity of 2.1 ± 1.3 mlyr (Figs. 1-2). Pit 4,
40
in the compression zone, retained 1700 ± 450 yr of lOBe, although this age remains a
minimum estimate, as our deepest sample interval shows high concentrations of lOBe
implying the isotope has penetrated to depths below 90 cm.
6. DISCUSSION AND CONCLUSIONS
The spatial pattern of movement, as revealed by photo-derived tree displacement
vectors (Fig. 1), shows a systematic change in displacement rates from the headscarp to
the lower portion of the earthflow. The upper source zone exhibits slowing velocities on
the order of 1 m/yr, indicative of slow steady input of source material from headscarp
slumping. The maximum earthflow velocities are obtained through the narrow (80 m
wide) section of the transport zone. This section also has the smoothest surface, largely
free of folding or extensional slumping, implying plug sliding with minimal internal
deformation (Fig. 1- Appendix B). The mid-slope compressional zone exhibits slower
movement than the transport zone, but follows a similar temporal pattern. Here the slide
widens and has a series of folds and thrusts that reflects deformation from the advancing
transport zone.
Although we do not have well-constrained data on the basal sliding depth, field-
based estimates from the channel margin of the Kekawaka slide are -6 m, and similar
sized earthflows have depths in the range of 4-8 m (Iverson and Major, 1987; Zhang et
aI., 1991). Taking an average velocity of 2 m/yr, the earthflow translates over 1000 m3fyr
through the transport zone. Given that the combined area of the earthflow source and
transport zones is 1.0 x lOs m2, sliding at 2 m/yr equates to an average lowering rate of
10 mm/yr, -20 times faster than the estimated regional erosion rate (Fuller et aI., 2009).
The evacuated volume of the earthflow source area and transport zone is -1.0 x 106 m3,
suggesting that -1000 yr are required to evacuate the earthflow basin given average
transport zone sliding of 2 m/yr.
The inventories of lOBe provide crucial insight about the long term evolution of
the landslide. Pit C on the unfailed surface above the headscarp has the highest lOBe
41
inventory as would be expected for stable ground. Given the sandy nature of the soil at
this location, some of the lOBe may have been lost from this pit, probably by leaching. Pit
1 is within the steep, unstable part of the accumulation zone amidst back-rotated
headscarp slumps and disintegrating blocks of detached argillaceous bedrock. As such,
we interpret much of the Pit 1 inventory and concentration spike at 80-90 cm to be
residual headscarp soil that has yet to be fully mixed or buried, indicating that Pit 1
reflects a transient state between headscarp failure to reconstituted earthflow surface.
Owing to their location in the transport zone, Pits 2 and 3 are useful for
determining long-term rates of earthflow movement. Both show the majority of lOBe is
retained within the upper 40 cm, and these pits have negligible lOBe below this depth. We
argue that lOBe at these sites began accumulating when the surface moved into the
transport zone, which is dominated by translational movement as opposed to the mixing
in the source zone. Our lOBe-derived, pre-historic velocity (2.1 m/yr) exceeds the 62 year
average transport zone velocity obtained from the photos (1.7 m/yr) and to account for
the reduced velocities over the past 30 yr, the transport zone may have sustained
velocities approachirlg 2.5 m/yr prior to 1944.
Pit 4 has 1700 ± 400 yr of lOBe. The Pit 4 lOBe depth profile shows lOBe
concentration decreasirlg irI a noisy exponential fashion, and the isotope has penetrated to
depths well below 40 cm, and probably below our deepest sample at 90 cm. The Pit 4 age
represents a minimum age of the earthflow. The transport rate between Pits 3 and 4 is
-0.15 m/yr, and the earthflow has longitudinally shortened through this section via
compression and widening, which may account for much of the deceleration. Air photo-
derived velocities in this area exceed 1 m/yr, suggestirlg that earthflow movement is
episodic and the earthflow surface proximate to Pit 4 may have origirlally been formed
during a previous episode of activity, and has subsequently been reactivated, all the while
continuing to accumulate lOBe.
The acceleration and subsequent gradual slowirlg of the 1944-2006 vectors
illustrates the non-steady nature of this earthflow' s movement at the decadal scale,
probably due to climatic variation, supportirlg the notion of episodic or surgirlg behavior
42
superimposed on the long-term earthflow evolution. Historical increases in earthflow
activity have colloquially been attributed to anthropogenic activity such as grazing, but
the dramatic slowing of the Kekawaka earthflow does not support this contention. The
peak velocities recorded from 1964 through 1981 coincide with both an increase in
rainfall and discharge across the Eel River catchment in the mid 20th Century (Syvitski
and Morehead, 1999), and a large storm event in 1964.
Regardless of climatic forcing, a supply of readily mobilized source material is
requisite for continued earthflow movement. The Kekawaka earthflow headscarp has
retrogressed to a sandstone ridgeline, likely limiting or greatly slowing future expansion
and the source zone has received minimal fresh material since the 1940s. With continued
earthflow movement (and assuming no downward propagation of the failure surface), the
slide thickness must have decreased causing a reduction in shear stress. This supply
limitation may be a primary factor in the long-term velocity decline revealed by our
analysis, although the buttressing effect of accumulation in the low gradient mid slope
compressional zone may also contribute. The surface of the earthflow is 5-lOm lower
than the surrounding slopes, consistent with progressive deflation and slowing of the
active slide. Given this supply limitation, the Kekawaka earthflow appears set to become
a dormant feature, typical of this slide-dominated landscape.
43
CHAPTER IV
DECADAL-SCALE VARIATIONS IN EARTHFLOW MOVEMENT AND
SEDIMENT PRODUCTION USING HISTORICAL AERIAL PHOTOS
AND AIRBORNE LIDAR, EEL RIVER, CALIFORNIA
This chapter is in preparation for submission to the journal Geological
Society ofAmerica - Bulletin. This chapter was co-authored with Josh
Roering, who provided financial, advisorial and editorial support.
1. INTRODUCTION
In non-glaciated mountainous landscapes, sediment production and hillslope form
are primarily controlled by mass wasting processes, including deep seated landsliding and
earthflows. These large slope failures can establish or modify drainage patterns, alter
hillslope morphology, and impart large perturbations on sediment budgets by regulating
the timing, magnitude, frequency, spatial distribution, and grain size of sediment entering
the channel or river network (Hovius et aI., 1998; Korup, 2005b, 2006b). This has direct
implications for flooding, channel network evolution, sediment transport, aquatic habitat,
and infrastructure. While the importance of landsliding in mountainous landscapes has
been long recognized, understanding how landsliding, and in particular earthflows,
operate as a geomorphic process has proven more challenging.
44
1.1. Earthflows - Slow Moving 'Earth Glaciers'
Earthflows are a class of landslide (Cruden and Varnes, 1996) characterized by
continuous movement along transient shear surfaces with a degree of internal
deformation or flow. They span a range of landslide failure styles, and have been variably
described as landslide complex (Iverson, 1986a), earth-slide (Cruden and Varnes, 1996),
and mud-slide (Chandler and Brunsden, 1995; Glastonbury and Fell, 2008). Here, we use
the term earthflow as advocated by Hungr et aI. (2001) to describe large, slow moving
landslides. These features typically resemble 'earth glaciers'. They are slow moving «4
m/a), large (>500 m long), deep-seated (>5 m thick), clay rich, and behave in a plastic or
visco-plastic manner (e.g., Baum et aI., 1993; Bovis and Jones, 1992; Chandler and
Brunsden, 1995; Coe et aI., 2003; Iverson and Major, 1987; Kelsey, 1978; Malet et aI.,
2002; Zhang et aI., 1991). Earthflows classically have an hour-glass planform, with an
amphitheatre-like accumulation zone, an elongate narrow transport zone, and a lobate toe
(Keefer and Johnson, 1983). Commonly earthflows exhibit seasonal movement patterns
and can require several months of cumulative rainfall before the onset of movement.
Earthflows can potentially dominate sediment delivery to channels in erosive landscapes
(Kelsey, 1978; Swanson and Swanston, 1977), yet seldom fail catastrophically. We
specifically distinguish earthflows from large displacement, catastrophic single event
failures, such as rockslides, slumps, or translational bedrock slides.
1.2. Mass Movements and Landscape Evolution
A common approach to quantify the role of a surface process in landscape
evolution is to develop a physically based mathematical statement which describes the
sediment flux or erosion, and resulting landscape form, attributable to the process over
geomorphically significant time and space scales (Dietrich et aI., 2003). This statement
can then be incorporated into numerical landscape evolution models to quantify the
process's contribution to landscape evolution and change (e.g., Perron et aI., 2008).
While this approach has worked successfully for surface processes which operate
45
predictably across a landscape (such as channel incision, slope dependent soil creep, and
debris flow incision), landslides have proven more challenging to describe using this
framework. Difficulties arise when extrapolating geomechanical theories of slope
stability beyond an individual hillslope to the landscape scale, where the hillslope
material properties, vegetation, destabilization processes, rate of base level change, and
triggering factors (meteorological and seismic events) are inherently spatially and
temporally variable (Dietrich et aI., 2003).
Despite these challenges, some progress has been made with single-event
catastrophic deep-seated landsliding. Landsliding of this type is somewhat predictable,
based on structure, bedding, defects, critical strength discontinuities (Korup et aI., 2007;
Pettinga, 1987), and considerations of mountain scale strength (Schmidt and
Montgomery, 1995). Landscape evolution models incorporating a simplified process of
deep seated bedrock failure can generate synthetic landscapes which statistically replicate
real topography. These models incorporate probabilistic event failure based on a critical
slope (Densmore et aI., 1998) and employ detachment limited transport of resulting
landslide debris.
1.2.1. Earthflows and Landscape Evolution
No landscape-scale theory exists for slow, sustained mass movements such as
earthflows, and there is little data or theory pertaining to earthflow behavior beyond the
decadal scale. As a consequence, we cannot confidently explain where earthflows occur
in a landscape, what percentage of terrain is active at a given time, the duration of
activity, and what governs long-term movement. A key question is whether earthflow
movement is systematic or predictable in space and time.
Perhaps more important than the lack of theory describing catchment-scale
earthflow generated sediment flux, we also lack a conceptual model describing the long
term evolution of an earthflow prone landscape, and even the long-term evolution of an
individual earthflow. For clay rich, earthflow-prone hillslopes, there is nothing
46
comparable to the colluvial hollow and debris flow framework applicable to soil mantled
uplands (Dietrich and Dunne, 1978; Lehre and Carver, 1985).
1.2.2. The Focus on Earthflow Mechanics
Although we lack earthflow data for the long term, extensive work has been
undertaken on the behavior of individual earthflows at the seasonal scale (Coe et aI.,
2003; Iverson and Major, 1987; Malet et aI., 2002; Van de Grift and Sack, 2004), specific
earthflow mechanics such as shear zone dilatency and strengthening (Iverson, 2005;
Schulz et aI., 2009), and the evolution of material strength over time (Maquaire et aI.,
2003). Many attempts have been made to model earthflow movement and rheology (e.g.,
Angeli et aI., 1996; Baum et aI., 1993; Bruckl and Sehidegger, 1973; Iverson, 1986c;
Savage and Wasowski, 2006; van Asch et aI., 2007; Vulliet and Hutter, 1988), primarily
to describe or replicate the behavior of specific earthflows.
Despite the abundance of site specific geomechanical studies focusing on seasonal
earthflow movement, knowledge of earthflow activity across a landscape trails our
understanding of other styles of landsliding. This is in part attributable to the continuous,
sometimes inconspicuous, failure style of earthflows. Whereas single event landslide
failure can be easily bracketed between sequential sets of aerial photos or site visits,
earthflows can continue moving for many decades, at rates heterogeneous both over time
and spatially within an individual earthflow complex (Swanson and Swanston, 1977).
One of the few attempts to study longer term earthflow behavior was undertaken by
Bovis and Jones (1992), who used radiocarbon dating and dendrochronology to link
periods of accelerated earthflow movement with sustained wet periods across several
hundred years of climatic changes in British Columbia. Bovis and Jones (1992) stressed
the sensitivity of large slow moving earthflows to multi-year climatic trends, rather than
individual storms or wet seasons.
47
1.3. Earthflows - A Unique Form of Landslide
1.3.1. Scaling Relationships
Inventory maps of catastrophic landslides reveal that landslide frequency often
decays as the inverse power of landslide area, both for a total landslide inventory, or an
event specific distribution (Guzzetti et aI., 2002; Malamud et aI., 2004). From this fractal
relationship, long-term landslide erosion rates can be estimated (Hovius et aI., 1997). In
comparison, few, if any, earthflow inventory maps sufficiently document earthflow area,
distribution, and populations to allow investigation of comparable scaling relationships
and estimation of earthflow derived sediment flux.
1.3.2. Long Term Erosional Features
Dormant earthflow features can persist in the landscape for thousands of years
(Mackeyet aI., 2009). Therefore, the history of earthflow evolution and movement must
be integrated across a range of climatic conditions and topographic changes, rather than a
single causative event such as an earthquake or storm (Bovis and Jones, 1992). Bedrock
slopes, in contrast, can adjust suddenly to incision or climatic change (Burbank et aI.,
1996; Ouimet et aI., 2009), making the effect of landscape change immediate and
obvious. In both landscape evolution models and calculations of sediment flux,
catastrophic slope failure is often amenable to simplification as a detachment limited
erosion problem (Densmore et aI., 1998).
Low-gradient, fine-grained hillslopes have longer response times. Such hillslopes
generally respond by gradual retrogressive failure (Iverson and Major, 1987; Skempton,
1964), so after a perturbation such as base level fall, an earthflow prone hillslope may
take hundreds or thousands of years to respond and fully adjust (Carson and Pelley,
1970). Landslides can be slope clearing, removing all material down to the failure
surface, and erasing much of the pre-failure topography (Densmore et aI., 1997).
Earthflows frequently stabilize after a period of movement preserving characteristic
earthflow morphology on the slope (Keefer and Johnson, 1983), essentially redistributing
mass on the hillslope which may not immediately reach the channel network. While this
48
complicates estimates of earthflow derived sediment flux, dormant earthflow morphology
can be used to reconstruct the history of earthflow movement (Booth et aI., 2009;
McKean and Roering, 2004).
1.3.3. Position in the Landscape
Predictably, earthflows predominately occur in fine-grained, clay-rich lithologies
(Hungr et aI., 2001; Keefer and Johnson, 1983), especially in argillaceous or altered
volcaniclastic sediments (Glastonbury and Fell, 2008). Topographically, Ohlmacher
(2007) shows small shallow earthflows preferentially occur in regions of planar slope
with low planform curvature. Iverson (l986b) notes that earthflows typically extend from
the channel to the ridge, and Kelsey (1978) suggests earthflows in northern California
tend to have a southerly aspect, where dry summers inhibit conifer growth on the slopes.
Beyond these observations we have little data establishing where earthflows
characteristically occur within a drainage basin and on hillslopes. For example, do
earthflows require a critical drainage area or slope? Are they driven by failure high on the
slope which coalesces and moves downhill, or does a stream bank failure progressively
retrogress to the ridgeline?
1.4 Key Questions
We undertook extensive work in the Eel River catchment in northern California
(Fig. 1), making detailed earthflow maps with the aid of LiDAR, aerial photography and
field inspection. We ask four questions. Firstly, how much of the terrain is active? Do
earthflows have a characteristic spatial attributes, or a preferential position in the
landscape, such as slope, lithology, or aspect? Thirdly, do temporal patterns of movement
correlate to climatic variables such as rainfall or temperature? Finally, we seek the
contribution of earthflows to regional sediment flux and erosion. More specifically, how
much of the sediment delivery to the channel network is attributable to earthflow
activity?
49
:0
Pacific
Ocean
20 40km
CA
Figure 1.. Location of the study area along main stem the Eel River, northern
California. LiDAR data was acquired over the 230 km2 region shaded in red.
Towns mentioned: Alderpoint (AP), Dos Rios (D.R.), Scotia (Sco), and
Garberville (Ga). Note the strong northwest trending structural grain visible in
the shaded rei ief background map.
50
2. STUDY AREA - EEL RIVER, CALIFORNIA
2.1. Introduction
The northern California Coast Ranges are comprised of the Franciscan Complex,
a Jurassic-Cretaceous penetratively sheared meta-sedimentary accretionary prism
complex subject to uplift since the Neogene. Rainfall averages 1.2 mia, primarily falling
between October and May. The combination of weak melange lithology, active tectonics,
and heavy seasonal rainfall makes the Eel River watershed especially prone to landslides
and slope instability, and an ideal location to study active earthflow processes.
2.2. Geology and Structure
The Franciscan Complex is comprised of three structurally separated belts, the
Eastern, Central and Coastal belts. These terrains young to the west, reflecting the
cumulative accretion of oceanic sediments to western North America (Jayko et aI., 1989;
McLaughlin et aI., 2000). The Central belt is especially prone to landsliding. It runs
through much of the Eel River catchment, and is structurally bound on the east by the
Grogan-Red Mountain fault zone, and on the west by the Coastal belt thrust. It consists of
an extensive Late Jurassic to Middle Cretaceous argillaceous melange matrix,
encompassing blocks and slabs of sandstone and shale turbidite sequences (McLaughlin
et aI., 2000). This assemblage was obliquely obducted to North America from 88 to 40
Ma, prior to emplacement of the Coastal Belt. Large blocks of older, meta-sandstone,
meta-basalt and high blueschist-grade rocks have been incorporated into the penetratively
sheared matrix of the Central belt from the older eastern units during oblique dextral
translation. These more competent, erosion resistant units have a significant local
influence on the topography, persisting as erosion resistant topographic highs amid the
melange (Fig. 2). The northwest trending structural grain dominates the modern
topography, with major axial drainages and ridges trending northwest (McLaughlin et aI.,
1982).
51
N
+
ewr
Legend
cm1- Meta-argillite
cm2 - Meta-argillite
cb1 - Sandstone
cb2 - Sandstone
cwr - Sandstone
chr - Sandstone
bs - Basaltic rock
cc - Chert
yb - Sandstone
b - Melange block
gs - Greenstone
c - Chert
m - Blueschist
sp - Serpentinite
cm2
4km
\
b
b
yb
chr
Figure 2. Simplified lithological map modified from Jayko et aJ. (1989) and
McLaughlan et al. (2000). See Table I for more detailed lithological descriptions.
Active earthflows and the study area extent are outlined in black, and
predominantly occur in the melange unit cm I.
52
The post-emplacement tectonics of northern California Coast Ranges has been
dominated by the northerly migration of the Mendocino triple junction (MTJ) since the
Miocene. The Pacific and Gorda plates are translating north relative to North America
along the Cascadia Megathrust, creating the San Andreas Fault to the south (Furlong and
Schwartz, 2004). This creates a double-humped zone of uplift which migrates north at
approximately 5 cm/a (Furlong and Govers, 1999), and a series of northwest-trending
emergent fault systems cut through the northern California Coast Ranges associated with
the advance of the MTJ (Kelsey and Carver, 1988).
This migrating zone of uplift has had a profound influence on the landscape,
causing river capture and drainage reversals. Most notably, the headwaters of the Russian
River were captured by the Eel River at 2 Ma when river incision was unable to keep up
with the transient wave of increased uplift rates (Lock et aI., 2006), generating the
unusual 'fish hook' drainage pattern of the Eel River tributaries (Fig. 1).
2.3. Uplift Rates and Erosion
Due to the high rates of erosion, many geomorphic studies have focused on the
Northern California Coast Ranges. Wahrhaftig and Curry (1967) warned the rates of
erosion in the Eel River (0.84 mmJa) were abnormally high in comparison to other North
American river systems, and called for further research into sediment sources and
monitoring. After an intensive period of hydrologic and suspended sediment data
collection by the USGS in the mid century, Brown and Ritter (1971) established that
from 1957-1967, the Eel River had the highest sediment yield of any non-glacial
continental river in the United States at 10,000 tons per square mile (3900 tons/km\
Extensive research has been undertaken in the offshore environment focusing on
the rate and distribution of shelf sedimentation (e.g., Nittrouer, 1999). Notable are the
changing rates of sedimentation over the Holocene (Sommerfield and Wheatcroft, 2007)
and in historical times (Sommerfield et aI., 2002; Sommerfield and Nittrouer, 1999). In
the 20th Century, the Eel River catchment experienced significant logging and road
53
building activities, in addition to major storms in 1955 and 1964 (Brown and Ritter,
1971; Sloan et aI., 2001). Determining the relative contribution of natural versus
anthropogenic effects on erosion and sediment yield has proven challenging
(Sommerfield and Wheatcroft, 2007).
Rapid rates of uplift and channel incision are reflected in fluvial and marine
terraces. Merritts and Bull (1989) and Merritts (1996) documented variable rates of
coastal uplift peaking at 5 rnrn/a at the coast near to the MTJ. Uplift rates are thought to
decay inland to approximately 1 rnrn/a near Garberville (Bickner, 1985), located on
Figure 1.
A variety of approaches have been used to estimate background levels of erosion.
Fuller et aI. (2009) used cosmogenic isotopes and optically stimulated luminescence
dating to calculate long term erosion rates for the upper South Fork Eel River of
approximately 0.3 rnrn/a. Similarly, Gendaszek et al (2005) measured concentrations of
lOBe in sediments from northern California rivers, and document long term catchment
averaged erosion rates of 0.8 rnrn/a in the upper Van Duzen, and 0.4 rnrn/a for the Eel
River catchment near Scotia. Modern erosion rates can be estimated with suspended
sediment, and Wheatcroft and Sommerfield (2005) re-analyzed suspended sediment data
(1950-2000) and calculated a sediment yield of2232 tonnes/kd/a across the 8063 kd
Eel River catchment, which equates to a catchment averaged erosion rate of 0.9 rnrn/a
(assuming a bedrock density =2.5 t/m\ This period includes both anthropogenic effects
and the mid-century storms.
2.4. Landslide and Earthflow Studies
The most comprehensive study of earthflow processes in the Eel River catchment
was undertaken by Kelsey (1977; 1978; 1980) in the Van Duzen river basin, a large
tributary of the Eel River at the northern end of the drainage basin (Fig. 1). Kelsey
studied 19 active earthflows over a several year period in the 1970's through a
combination of field surveying and aerial photo analysis, and calculated a regional
I~
54
sediment budget. Although covering just 1% of the Van Duzen basin area, earthflows
contributed 10% of sediment to the channel. Kelsey estimated that earthflow translation
accounts for approximately 50% of the sediment flux from active features, the remainder
removed by active gullies on the surface of the earthflow.
Another extensive study was undertaken farther north in the Redwood Creek
catchment, looking at all aspects of sediment production and transport (e.g., Harden et aI.,
1995; Nolan and Janda, 1995). Nolan and Janda (1995) studied the sediment flux from
two Redwood Creek earthflows, but concluded fluvial processes accounted for only 10%
of the long term erosion of the earthflow complex, the majority coming from mass
movement. They also stressed the temporal variability of earthflow activity across
northern California, due to localized lithology and the varying topographic development
of individual earthflows. Iverson (1986a; 2005; Iverson and Major, 1987) undertook
extensive work on the Minor Creek earthflow in the Redwood Creek Basin, and
developed a model to explain earthflow movement by the slow infiltration of winter
rainfall through the landslide mass. Iverson (2005) argues movement is regulated by
shear zone dilatancy, which decrease pore pressures, thus preventing catastrophic failure.
Kelsey (1980) and Muhs (1987) highlighted the role of contrasting geomorphic
processes operating in different lithologies. The harder, competent, steep sandstone
lithologies feature well organized ridge and valley drainage networks, with erosion
dominated by fluvial debris flow incision. In contrast, the weaker melange units have a
dense but poorly developed ephemeral drainage network, longer low gradient slopes, and
erosion is dominated by earthflow and mass movement processes. Despite this weak
lithology, rivers frequently steepen when they cross weak melange, as earthflows deliver
boulders to the channel which armor the bed (Kelsey, 1977, 1978).
2.5. Main Stem Eel River Study Area
Our study focuses on a section of the main stem Eel River between Dos Rios and
Alderpoint (Fig. 1), which has one of the highest concentrations of earthflow activity in
55
the Eel River catchment (Brown and Ritter, 1971). Lithology in this area is
predominantly argillaceous melange of the Central belt Franciscan Complex (Fig. 2),
characterized by long, low gradient slopes and extensive slope instability. The melange
was open oak grassland at the time of European settlement in the 1850's, attracting
ranchers, and the primary land use remains low-density cattle ranching. Conifer growth
and forestry are generally limited to isolated sandstone outcrops. The Northwestern
Pacific Railway follows the main stem Eel River canyon, connecting Eureka to San
Francisco, but this line was abandoned in 1998, partially due to the high cost of landslide
related maintenance.
Descriptions of the large landslides and earthflows along the Eel River canyon
between Dos Rios and Alderpoint have largely been confined to a variety of engineering
reports for the California Department of Water Resources (Dwyer et aI., 1971; Scott,
1973; Smith et aI., 1974), primarily in anticipation of dam construction over much of the
Eel River catchment (California Department of Water Resources, 1965). These were
largely reconnaissance studies with descriptions of the landslides and analysis of the
earthflow material. They included estimates of landslide generated sediment flux,
although with limited information on slide depth and long term movement rates. Malhase
(1938) described the Mile 201 landslide after a wet winter in 1938: "the surface of the
slide takes on the characteristic form of mud glaciers, with lateral terminal moraines, with
rolls and pressure ridges forming valleys and hummocks. Great cracks open up into
which water from the surface enters."
More recently, Mackeyet al. (2009) studied an earthflow on Kekawaka Creek, a
tributary to the Eel River, and highlighted the variability of temporal movement and the
longevity of earthflows in the landscape. Roering et al. (2009) used satellite-based InSAR
to monitor seasonal movement of the Boulder Creek landslide, and estimated a minimum
erosion rate of 1.5 mm/a across the earthflow source area. Brown and Ritter (1971)
calculated a 2 mm/a erosion rate for the 1701 km2 catchment area of the Eel between Dos
Rios and Fort Steward between 1965 and 1967, although this period was in the aftermath
of the 1964 flood and may be elevated beyond background rates.
56
Very little data exists on the depth of large landslides and earthflows along the Eel
River. The exceptions are 2 landslides in the lower reaches of the Middle Fork Eel River
(California Department of Water Resources, 1970), approximately 34 km southeast of
our study area, but in comparable Central belt Franciscan lithology. Boreholes in the 1.6
km long, 0.45 km2 Salt Creek landslide, 3.5 km upstream of Dos Rios, revealed landslide
colluvium thickening from 6.7 m near the headscarp, to depths of 33 and 35 m towards
the toe. The larger 0.64 k~ 1 km long Salmon Creek landslide is 16 km upstream from
Dos Rios. A drill hole in this landslide was sheared off at 34 m depth, although the
inclinometer also detected small amounts of movement at 57 m, attributable to deep
seated failure within the sheared shale bedrock.
3. METHODS - MAPPING EARTHFLOW ACTIVITY WITH HIGH
RESOLUTION TOPOGRAPHY AND PHOTOGRAMMETERY
3.1. Background - Landslide Mapping
Landslide mapping has traditionally been undertaken by a combination of stereo
pair aerial photo and topographic map analysis, in concert with some field verification.
Aerial photo analysis can efficiently cover a large area, but at the expense of accuracy-
especially in forested terrain, and problems arise in accurately relocating features on a
photograph to the base topographic map (Malamud et al., 2004). Field mapping yields
greater accuracy (especially with modern GPS technology), but at the expense of high
labor costs and the limited extent of terrain that can be covered (Wills and McCrink,
2002). In recent years satellite-based InSAR mapping has proven effective in locating
slow sustained mass movement (Hilley et al., 2004; Leprince et al., 2008). The C-band
PALSAR ALOS satellite has achieved success mapping earthflows in mountainous and
vegetated terrain (Roering et al., 2009), although data only extends back to 2006.
The increasing availability of high resolution topography acquired through
airborne laser swath mapping (Light detection and ranging or LiDAR) has greatly
enhanced the ability of researchers and practitioners to map mass movement features
57
across a broad area. LiDAR based mapping allows greater accuracy in both correct
feature identification and location than is available with traditional aerial photo and field
mapping approaches. This has been particularly evident in landscapes with deep seated
landslides and slump features occurring in glaciated or generally uncomplicated
topography, such as the Puget Sound area in Washington (Schulz, 2007), and the layered
basalts in southern Idaho (Glenn et aI., 2006). In these landscapes, the margins of
landslides are easily recognized on LiDAR imagery as they have a distinctly different
visual and statistical surficial texture in comparison to the surrounding unfailed terrain
(Booth et aI., 2009; McKean and Roering, 2004).
In landscapes prone to pervasive slope instability with multiple generations of
landsliding, distinguishing between historically active and long dormant mass movement
can be difficult, even with the LiDAR derived meter-scale topographic maps. While the
Lidar maps have been a great improvement on previous approaches, confidently mapping
mass movement in terrain with multiple episodes of deep seated failure remains a
challenging and largely subjective exercise (Van Den Eeckhaut et aI., 2005).
Despite this uncertainty, qualitative estimates of earthflow activity state or age
can be made from landslide morphology. Over time, morphological features attributable
to landslide movement (e.g sharp headscarps, tension cracks, compressional folding and
lateral margins) become smoothed out and attenuated by small scale surficial or diffusive
processes (e.g soil creep, bioturbation) when a landslide stops moving (Gonzalez-Diez et
aI., 1999; Wieczorek, 1984). Figure 3 shows earthflows adjacent to the Eel River which
qualitatively appear to span a range of activity states. Two currently active features with
fresh kinematic structures adjoin a dormant landslide, where much of the small scale
morphology has been erased, leaving subtle headscarps and lateral margins.
58
Figure 3. Active vs. dormant earthflow morphology. Two active earthflows
show sharp morphological features, whereas a large dormant earthflow to the
south has had much of the finer scale morphology attenuated by surficial
processes. The Northern Pacific Railroad runs along the southwest side of the Eel
River
Visual inspection of bare earth LiDAR shaded relief maps in the Franciscan
melange of northern California reveals a wide range of landslide and earthflow features.
Slope failure ranges from small slumps to huge earthflow complexes (Brown and Ritter,
1971; Kelsey, 1978; Mackey et al., 2009). Complicating the mapping process, on a single
hillslope, earthflows often exhibit complex cross-cutting and nested relationships, and
span a range of size, activity state and failure style.
59
3.2. LiDAR
The National Center for Airborne Laser Mapping (NCALM) acquired a high
resolution LiDAR dataset of the study area in September 2006 (Fig. 1). The raw data was
processed by NCALM, who produced ESRI-compatible grids of elevation in January
2007. The elevation grids included both unfiltered elevation (trees and buildings are
recorded and show up in the data), and filtered data, where algorithms filter out
vegetation and structures, leaving the bare earth elevation (Carter et aI., 2007; Slatton et
aI., 2007).
3.3. Photo Orthorectification
The earliest extensive aerial photo set covering our field area in the Eel River
catchment was flown in 1944 by the Department of Defense. These 1:24000 scale photos
are of high quality; trees, buildings, and subtle topographic features are readily
identifiable. We acquired medium resolution (800 dpi) scans of the photos extending over
the field area (Appendix D).
For quantitative analysis, aerial photographs must be referenced to the ground,
and rectified to remove lens distortions and topographic effects (Wolf and Dewitt, 2000).
This requires the calibrated focal length of the lens, the coordinates of fiducial markers
on the edge of the photo, a digital elevation model of the terrain, and ground control
points to co-locate features on the photo with features on the ground.
We were not able to locate camera information or calibration reports for the 1944
series of photos, but estimated the focal length (210 mm) from a database of typical
camera specifications (Slama, 1980). The coordinates of fiducial marks around the
margins of the photos were measured manually on true scale scans in Corel Draw
graphics software to within 0.005mm. We assumed a perfect principal point of focus (x,y
=0,0) to construct the photo coordinate system. We used the 1 m2 LiDAR bare earth
elevation as the elevation model, and the unfiltered 1 m2 LiDAR shaded relief as a
reference image to rectify the photographs within ERDAS Imagine 9.3 software. For
60
ground control points, we identified features such as small trees, buildings, or rock
outcrops on unequivocally stable terrain (such as ridges and terraces) and co-located the
identical features on both the LiDAR shaded relief image and the photo.
We compared the position of stable features on the LiDAR with the rectified
photos across the study site, to estimate error. Through this process, we could rectify the
photos with a high degree of accuracy (-2 m), and relocate features on the photos as they
were placed on the ground in 1944.
3.4. Objectively Quantifying Earthflow Movement
The earthflows of the Eel River catchment commonly have oak trees and bushes
growing on the earthflow surface. These trees continue to grow on the landslide mass as
it translates downhill, and are readily identifiable in sequential photographs, on unfiltered
LiDAR maps, and in the field. By comparing differences between the 1944 photos and
the unfiltered LiDAR data acquired in 2006, we were able to track the locations of
individual trees and construct a series of displacement vectors for the 62 year time
interval. We use this approach to objectively map historically active earthflows, and
discriminate between stable and moving terrain.
The distribution of trees varies across the study area, and between different
earthflows. Some earthflows have many long-lived trees that can be used to construct a
detailed vector field of displacement (Fig 4.), whereas on other earthflows trees are
sparse and the extent of movement is more difficult to objectively discern with tree
displacement vectors alone. To delineate the margins of earthflows we used an iterative
approach, carefully comparing the rectified 1944 photos and the LiDAR imagery. The
primary guide to mapping the earthflow margin was the comparison of stable and moving
trees, which show a clear offset on active earthflows. The margins of active earthflows
typically feature a subtle morphological structure, such as a headscarp or lateral levee, or
toe lobe thrust, and we used these morphological features to guide fine scale margin
construction.
61
Figure 4. Unfiltered shaded relief map of the Boulder Creek earthflow. Margins
of active earthflows outlined in black, with individual tree displacements in red.
Inset shows displacement at the mid-section of the flow, where it crosses a
structural barrier running obliquely across the transport zone - note the large toe
advancing from the northeast. The Lone Pine earthflow (L.P.) is to the northwest.
Eel River runs south to north along the toe of the Boulder Creek earthflow.
Coordinates are UTM Zone ION.
62
Q
Q
Q
....
<D
..,.
Q
Q
Q
o
<D
..,.
Q
o
o
en
I()
..,.
Q
Q
o
co
I()
..,.
OOOLl:PP
OOOLl:PP
0009l:PP
0009f:PP
OOOSl:PP
OOOSl:PP
OOOPl:PP
OOOPl:PP
OOOl:l:PP
Q
o
Q
....
<D
..,.
o
Q
o
o
<D
....
Q
<::I
<::I
co
I()
..,.
<::I
<::I
<:I
...
~
63
In areas with few or no trees suitable for quantifying displacement, we could often
observe movement based on changing earthflow morphology. For example, advancing
toe lobes or headscarp retrogression can be readily identified by comparing the photos
with the LiDAR maps. We took a conservative approach to mapping: if we could not
confidently map either vegetation displacement or morphological change, we did not
include the terrain as an active earthflow. Where we mapped displacements in different
parts of one contiguous earthflow, we amalgamated these into one larger contiguous
feature when justified by morphology. Additionally, we distinguished stable patches
within a larger earthflow from the neighboring mobile terrain.
We visited earthflow features in the field over 4 field seasons to confirm the
reliability of our technique. On the ground, active earthflows exhibit fresh headscarps,
exposed 'mole-track' lateral margins with slickensides (Fig. Sa), disturbed, densely
cracked hummocky terrain (Keefer and Johnson, 1983; Kelsey, 1978), and occasionally
trees which have been stressed, toppled or killed by ground disturbance (Fig. Sb).
Conversely, stable or dormant earthflows, while retaining much of the hummocky terrain
and characteristic earthflow morphology, do not exhibit the same degree of fresh ground
disturbance (Fig. 3).
We mapped earthflow features as detailed polygons using ESRI ArcMap 9.2
software, and constructed a statistical database on the individual earthflows. By
combining LiDAR, rectified photos and field inspection, we objectively mapped
landslide movement (>S m) over the 62 year interval.
64
Figure 5. Field evidence for earthflow activity. a) View up the active Penstock
earthflow, showing the disturbed ground and distressed trees. b) Lateral margin
of an active earthflow showing slicken Jines.
3,5, Earthflow Displacement Time Series
On 16 earthflows with sufficient displacement (>30 m over the 62 year period),
we used sequential aerial photos to construct a time series of earthflow movement,
following the approach of Mackey et a1. (2009). We calculated velocities for trees in the
transport zone, the sections of the earthflows which are most sensitive to climatic
changes (Bovis and Jones, 1992), and where displacements are greatest. Although the
aerial photo coverage varies across the study site, we were able to reconstruct the
65
temporal pattern of movement at approximately the decadal scale. With our resolution
(generally 4-10 years between photos), we expect to observe longer term earthflow
behavior, such as decadal-scale climatic patterns (Bovis and Jones, 1992), rather than the
effect of individual storms or seasonal movement.
In the aerial photo rectification process there is frequently a small positional error,
which can propagate through calculations of earthflow velocity. To correct for this, in
addition to tracking the position of trees on the earthflow through time, we also tracked
the apparent position of stable trees through the same photo series. We were then able to
differentially correct the apparent movement of the stable feature from the displacement
record of the moving tree.
3.6. Climate
One of the primary factors governing movement of many landslides is climatic
change, notably changes in precipitation, groundwater levels, and temperature. The
longest climatic record near the field area is located on the Eel River at Scotia (Fig. 1),65
km to the northeast of the study site (USGS Site 11477000 Latitude 40°29'30", Longitude
124°05'55"). Continuous daily temperature, precipitation and river discharge records date
back to the 1920's.
We calculated statistics each water year (starting October 1st) for discharge,
temperature and precipitation. Groundwater level is both a critical control on landslide
movement, and a primary factor in sustaining channel baseflow. As such, we use the 25th
percentile of daily annual discharge (Q2S) as a proxy for base flow. Additionally, lower
percentile flow measurements have been shown to be a more sensitive recorder of
climatic trends (Lins and Slack, 1999; Luce and Holden, 2009).
66
3.7. Sediment Production
We distinguished earthflows which discharge sediment directly into a channel or
major gully from flows which are unconnected to the channel network and do not
represent active sediment sources. To calculate the sediment delivery attributable to mass
movement from the earthflow to the channel, we multiplied the average annual toe
movement rate (from the photo-derived vectors) by the width and depth of the earthflow
at the toe (Fig. 6). We have no direct data as to the depth to the failure surface (e.g. from
drill holes), but can estimate from the height of the steep toe face where an earthflow
reaches the channel. We used the depth data from the Middle Fork Eel River earthflows
to guide depth estimates of the larger features in our study area. We assigned
uncertainties of 25% for the depth, 10% for the width, and 10% for the velocity when
calculating sediment flux from each individual earthflow.
By integrating the sediment contributed by the earthflows across the study area,
we can estimate the total sediment flux attributable to earthflow mass movement.
Dividing this sediment delivery rate (m3fa) by the study area (m2) gives the earthflow
driven erosion rate.
67
Penstock
Earthflow
Sediment flux
caclulati n
Figure 6. Oblique view of the Penstock Earthflow looking southeast across
Kekawaka Creek, illustrating the measurements taken to estim:lte sediment flux
from each earthflow.
4. RESULTS
4.1. Spatial Distribution of Earthflow§
The spatial distribution of active earthflows is presented in Figure 7. Across our
study area along the Eel River, we identified 122 earthflow features that moved during
the interval 1944-2006.
68
Figure 7. Map of the 226 km2 study area, active earthflows, and tree
displacements, from 1944 to 2006. Extent ofLiDAR coverage is shown as
shaded relief, colored by elevation. Background image is 30 m grid shaded relief,
and coordinates are UTM Zone ION.
69
Eel River Canyon
Earthflow Activity 1944-2006
, ~
~Pipe
. Creek
Eel River
(lIoo(j slagel
Tributary
Orain<lges .......r,
Active ~earlhflow
1~lrl"-,,rxl(';
Di5pl<lccmcnt
Elevnlion
lO·,"I.95
Sludy
area
448000
452000
462000
4km
,
0156000
..!tooo
460000
4600eo
464000
d ..OOO
70
4.1.1. Earthflow Shape and Area
Across the study area of 226 km2, we mapped active earthflows covering 16.5
km2, indicating 7.3% of terrain moved between 1944 and 2006. The 122 earthflows have
a median area of 36,500 m2, with an inter quartile range of 12,500-117,000 m2. Figure 8
shows histograms and a cumulative probability plot of earthflow area. The distribution of
earthflow areas appears to display 2 power law relationships, one with an area from 6.0 x
103 to 6.0 X 104 m2, and a second when area exceeds 6.0 x 104 m2. The largest active
earthflow in the study site, the Boulder Creek earthflow (3.1 x 106 m2), is over 3 times
the area of the next largest feature, the Island Mountain slide (0.94 x 106 m\ The group
of earthflows with area less than 6.0 x 104 m2 has an average aspect ratio (width/length)
of 4.2 ± 2.7. In comparison, when earthflow area exceeds 6.0 x 104 m2 the ratio is 7.3 ±
4.2 showing the larger earthflows are typically more elongate.
Many earthflows have intricate planform shapes, often with several small
tributary flows coalescing to form the primary earthflow (Fig. 4). Stable regions can
persist within an earthflow complex, surrounded by moving terrain. The earthflows
generally extend from the channel up to a ridgeline or major break in slope. Earthflows
which do not extend longitudinally along the full slope length are most commonly
restricted the upper parts of the hillslope (Fig. 7).
4.1.2. Earthflow Displacement
In total, we tracked the displacement of 998 features (trees or rocks) distributed
across the 122 earthflows, generating an average density of 1 displaced feature every 1.7
x 104 m2 of earthflow terrain. Median displacement was 23.9 m, with an inter-quartile
range of 14.6-42.3 m (Fig. 9a). Displacements ranged from 4.1 to 175 m, with a mean of
34.6 m. Over the 62 year interval, the median earthflow velocity across all moving
features was 0.4 mla. By comparing the locations of stable features both on the photo and
the LiDAR we estimated the error associated with photo rectification. Rectification error
was small in comparison to the displaced trees, with a median error of 2.1 m (Fig. 9b).
71
Frequency - Maganitude of Earthflow Area
3.2x1061.6x1068.0x105
50
40
C 30
::J
8 20
10
O-P~'"T-"-T--r--"T--.--.--r-r--r-,----.-,-------,----r""""--,---,-----,-----,
0.Ox10o
(a)
Earthflow area (m2)
100
(b) C 10
::J
o
U 1
3.2 3.6 4 4.4 4.8 5.2 5.6 6 6.4
Log(10) earthflow area (m2)
Cumulative Probability of Earthflow Area11"'\
\"1
g
:0
<Ile 0.1
0-
m
>~ 0.01
::J
E
::J
U
... ---------..
6000 < x < 60,000
Y = 18 956x(·O.34)
R2 = 0.98
x> 60,000
Y = 48903x(-1.04)
R2 = 0.98
0.001 -+--'--'-'--TTl""-,j-.-,-"-"",,IT"" -"-,--,,','" T"T''TT',---,---,-"" T"TI,,,,,,
1.0x103 1.0x104 1.0x105 1.0x106
Earthflow area (m2)
!Figure 8. Earthflow areas. a) Histogram of earthflow area. b) Log-log plot of
earthflow magnitude-frequency, c) Cumulative distribution plot of earthflow
area, The distribution exhibits two power laws, with a scale break at
approximately 6 x 104 m2.
72
(b) Rectification error of stable features
12
o 40 80 120 160
Tree displacement (m)
Figure 9.
a) Histogram of 998 tree
displacements.
b) Histogram of
rectification errors,
calculated from the offset
between stable features on
the photo and LiDAR.
Note the scale change
from Figure 9a.
n:: 110
median:: 2.1m
IQR:: 0.9-3.1 m
n:: 998
median:: 23.9 m
IQR:: 14.6-42.3 m
Length of displacement vectors
80
60
C 8
::J
o
u 4
o
-c6 40
U
20
o
(a)
o 2 4 6 8
Rectification error (m)
4.1.3. Slope and Lithology
We calculated the longitudinal slope of each earthflow (headscarp to toe), in
addition to the summary statistics of cells within each earthflow polygon. To compare
earthflow topography to the study area as a whole, we recorded the slope of all cells over
the extent of the study area.
The longitudinal slope has a normal distribution with a mean slope of 31±7%
(mean ± standard deviation). Figure 10 shows the zonal earthflow slope is higher at 36%,
accounting for the increased variability of individual cell slopes within each earthflow
polygon. The mean slope of the 122 earthflows is nearly identical to the mean slope of
73
the study area (36.5%), although the study area slope is more broadly distributed, and
slightly tailed, possibly accounting for gorges, cliffs, and steeper sandstone units.
Our distribution of earthflow slopes are comparable to those compiled by Keefer
and Johnson (1983) for smaller earthflows in the San Francisco Bay area, which clustered
on slightly steeper slopes of 20-22 degrees (36-40%).
Figure 2 illustrates the lithology of the study area, modified from Jayko et al.
(1989) and McLaughlin et al. (2000), and Table 1 documents the number and area of
earthflows in each lithology. Of the 122 earthflows, 98 (82% of the active earthflow area)
occur in the penetratively sheared melange unit (cml).
Regional slope and earth'flow median slope histogram
(4 m grid - 1% bins)
2.5 16
--...
:o? Terrain (grey): Earthflows (red): -0 2 c:'-" 12 ::l......
c Mean =36.5 Mean =36 0
::l 1.5 Q0 Stdev =19.5 Stdev =6.5 ~0 8
.Ec: 1
.c
'cu 4 tl:: 0.5 co(l) W
f-
0 0
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150
Slope (%)
Figure 10. Histogram is slope of all 4 m cells within the study area (grey). Red
histogram is median slope of 4m cells within each earthflow polygon.
74
Lithology Symbol Area (m2) % study No. EF area %EF % lith
area EF (m2) area active
Melange - penetratively cm1 1.32 x 108 58.2 98 1.33 x 107 80.89 10.08
sheared meta-argillite
Broken Fm cb1 5.75 x 107 25.4 16.5 2.10 x 106 12.71 3.65
(Metasandstone)
Melange (Subequal argillite cm2 1.88 x 107 8.3 2.5 5.64 x 105 3.42 3.00
and sandstone)
Serpentinite sp 4.67 x 10
6 2.1 5 1.63 x 105 0.99 3.49
Greenstone gs 4.14x10
6 1.8
Block (unknown lithology) b 2.66 x 106 1.2
Yolla Bolly yb 2.61 x 106 1.2
(metasandstone)
Radiolarian Chert c 1.02 x 106 0.5
White Rock cwr 1.23 x 106 0.5
(meta-sandstone)
Basaltic Rocks bs 6.86 x 10
5 0.3
Broken Fm cb2 4.16 x 105 0.2
(intact metasandstone)
Chert cc 5.57 x 10
5 0.2
Blueschist Blocks m 1.29 x 10
5 0.1
Total 2.26 x 108 100.0 122 16.5 x 107
Table 1. Proportion of lithology and active earthflows (EF) across the study area.
4.1.4. Aspect
Kelsey (1978) noted that earthflows along the Van Duzen River tended to have a
southerly aspect. To test this along the Eel River, we took the median aspect of 4 m cells
within each earthflow polygon (Fig. lla). The mean earthflow aspect is 210° with a mean
resultant vector of 0.42. Applying Rayleigh's test for significance of a mean direction, for
this dataset the critical resultant vector (a 0.05, n> 122) is 0.17, indicating the earthflows
do have a significant mean southwesterly aspect.
75
Given the strong northwest structural control on topography, we also plotted the
aspect of a1l4m cells within the study area to determine whether this biased the aspect of
earthflows. Figure 11b highlights the asymmetry of regional aspect, with an oblate
circular histogram showing a slight predominance of cells facing the southwest. The
mean direction of all topography is 248° with a small mean resultant vector of 0.04. The
critical resultant vector (a =0.05, n »100) is 0.17, so we can't reject the null hypothesis
that there is a preferred orientation of the terrain.
270
(a) Median aspect of earthflows
(Scaled by L09(10j Area (m2))
o
180
90 270
Mean I
resultant
vector
(b) Aspect of field area
o
180
90
Figure 11. Aspect of earthflows and the study area. (a) Median aspect of 4 m
cells within each earthflow polygon, scaled by Log lO area (m2). Mean resultant
vector is 210°. (b) Radial histogram of aspect of 4 m cells across the field area.
Mean resultant vector is 248°.
76
To see if the earthflow aspect trend is significantly different from the terrain as a
whole, we compared the datasets of earthflow and terrain aspect and performed an F-
Test. The two aspect datasets have a calculated F = 53. The critical F value at (a=0.05) is
3.8, so therefore we can reject the null hypothesis that the two populations were drawn
from populations with the same mean, and the aspect preference of earthflows is
significant.
4.2. Temporal Distribution of Earthflows
In addition to mapping the earthflows which moved between 1944 and 2006, we
were able to construct time series of velocity for 16 earthflows. We grouped the
earthflows into three categories based on the temporal pattern of movement. Figures 12-
14 show the temporal velocity of these earthflows, in addition to the outline and
topographic setting of each feature.
Group 1 slides show fast (3-4 rn/a) velocities from 1944 through to the late
1970's (Fig. 12). These velocities were sustained or increased through the 1960s, and
have gradually decayed since 1981. Earthflows in this group, especially the Kekawaka,
Boulder Creek and Rapids features, have some of the most classical earthflow
morphologies in the study area, with a long, well defined transport zone and have the
greatest cumulative displacements. These earthflows all discharge to a major channel.
The Marge earthflow moved a negligible amount after 1968, although this is a small and
recent looking feature, and could arguably be classed differently.
In Group 2 (Fig. 13), the earthflows exhibited fast velocities (3-4 rn/a) in the first
photographic intervals from 1944, and have steadily declined since, with little movement
beyond 1981. Three of the 4 earthflows in this group (Laufer, SE-1 and SE-2) are high on
the hillslopes and have no connection to major channels. White Lower has evidence of
activity all over the earthflow, although due to landscaping during railway operation we
could only track trees in the upper section of the earthflow. We have no information
about transport zone velocity of this feature, and it conceivably has a different temporal
77
pattern of movement than suggested by the trees near the headscarp. The adjacent 201
landslide required 2560 bulldozer hours for landscaping extensive drainage installation
(Dwyer et aI., 1971) which would overwhelm the natural behavior of these earthflows.
The third and largest group of slides has velocities which were generally low «2
m/a), and show steady movement with occasional surges (Fig. 14). A characteristic of
these earthflows is an unusual planform shape. For example, the Island Mountain
earthflow is a large wide feature, which showed a lot of internal variability in the rate and
amount of sliding. Some of the earthflows exhibit surges over 1-2 photo intervals, such
as Lone Pine or SE-3. This group of slides generally had less displacement than the other
groups (40-60 m). Six of the eight earthflows have a tight connection with a major
channel at the toe, and sustained toe erosion may promote a slow steady movement which
overwhelms other factors such as rainfall.
4.3. Climate
Annual rainfall, plotted in Figure 15, is low from the 1920's until the 1940's but
thereafter does not exhibit any obvious long term trends, although the 5 year running
mean highlights an approximately cyclical decadal pattern. Rainfall from approximately
1940 to 1977 does not show the inter-year variability of other periods, and Syvitski and
Morehead (1999) note there is a slight increase in precipitation (and mean annual
discharge) over the mid portion of the century. Similarly, the 25th percentile of annual
mean daily discharge at Scotia is low in the 1920's and 1930's, but increases from 1940
to 1960. Thereafter, increasing variability is superimposed on a gradual decline since
-1980. A second order polynomial fit highlights this pattern (Fig. 15). Conversely, mean
annual daily temperature decreases over the mid century, with a temperature decrease on
the order of 1°C. Again, a second order polynomial fit highlights how mean annual daily
temperatures were depressed from approximately 1940-1980.
78
2000
Marge
Kekawaka
1980
Year
1960
1
~ Boulder'Ck
~
I
------ : /"--..~I-~__/ ···-.~aPidS
.. 1----:--
Sustai ed Ea h low Ve ocity
1940
o
r---,----,----..----..----.,----.,------,
5
8
~6
~
.s
c4
'0
o
Qi
> 2
~ 4
E;:3
'5 2
o
Qi 1
>
o
~4].s3
c2
'0
.Q 1
Q)
> 0
Figure lZ. Temporal velocity and maps of earthflows with sustained velocities
from 1944-1981, and subsequent dec line. The Boulder Creek earthflow is shown
in Figure 4. In Figures 12-14, horizontal error bars represent the interval between
photos, and vertical error bars are the ± the standard deviation of transport zone
tree velocities for each interval.
79
Declinin earthflow velocity
SE-1
Laufer
------;--~
••••_1,
I~
I -----=--_ E-2
o
6
5
~4
E
~3
C
'g 2
Q) 1
>
o
~I Lower White
: li--+---!,~
6
o
1940 1960 1980
Year
2000
Figure 13. Temporal velocity and maps of earthflows with declining velocities
since the J940's-1950's. Lower White was heavily landscaped, and only trees on
the upper region of the flow could be tracked.
80
low, tea y arthflow Velocity
20001980
Year
Penstock
___---'---'-.---.jf:------,
Lake Shore
Wishbone
Lone Pine
1960
Island Mountain
•• I~White /"'-
~ '. ---/, ""-., --l,_:__,_
j----, .----,-::------j--:-
I----,--_~-a-!-m_is_e-........-:.....~
1940
6
o
'Ero~~o-
w.§.. 1
>
o
4 ]£~3o CIl
..Q E 2
Q)-
> 1
o
£~2jo CIl
.2"E 1Q)-
> 0
£~2]o CIl
.2"E 1Q)-
> 0
£~2jo CIl
.2"E 1Q)-
> 0
.£~2 ]o CIl
.2"E 1Q)-
> 0
Figure 14. Temporal velocity and maps of earthflows with a slow, steady
velocity, with occasional surges. The Lone Pine earthflow is shown in Figure 4.
81
__..-/- 5 year running average I
2000
2000
'.
.;V'J '.
. '.
1980 2000
1980
1980
'.'.
... ~ ,. -j":,.
/,~- ..~
'. ...
y =O.0007x2 - 2.9136x + 2879
R2 =0.2296
1960
1960
1960
Year
160 ~
120 ;;;-
m
80 C
40 .~
~+_...ill.QJ.ll.W.t.W¥J!..!W..IWJ1#-WW1ill!i~ill.l!.U4Ja1l.!LI!J.!!IJ.I!.mll.lli!fuI:J.lli.I!.lli.lf.Ul.l.I.I!llli~ 0 .ffi
o
(j);;;-
20000 g
16000 OJ
12000 ~
8000 -D
4000 :5
~-'-'-~---t.:...;...L:..J.U.ll.j-l.JoJ.j.....:..J..i..Lf"-Ul..LJ...I..:..iI.'"i..1.l..l:""::""~f'-U..u......J...:.~l..I..l.J..lJ..'-"""".i...W..~+--,-...u..:.-4..+ 0 ~
ffi
o
2500
E 2000
E
~
m 1500
'+-
e
.~
1000m
:J
e
e 500
<{
0
1920 1940
Q)::§' 25
Ol'" Y=-O.0022x2 +8.8336x - 8681.6'- Em~ 20 R2 =0.147-C OJ0_(j) .-
.- tDm 15>.:J
= IT
m '-D OJ 10
'- ::::
OJ a
>-
iim 5:JQi e
we
<{ 0
1920 1940
15
S
0- 14E
2
m 13:J
e
e
m
e 12m
OJ
:2
11
1920 1940
Figure 15. Plots of annual water year (Oct-Sept) climate statistics recorded on
the Eel River at Scotia (Fig. 1).
a) Annual rainfall (dots) with a 5 year running average. Daily rainfall is
plotted on the right axis.
b) 25 1h Percentile of annual Eel River daily discharge (dots) with 21\d order
polynomial fit (equation shown) and 5 year running average. Daily
discharge is plotted on the right axis - note the peak discharge in the
1964 flood.
c) Average annual mean daily temperature. with 2nd order polynomia I fit
(equation shown) and 5 year running average.
To test the significance of the second order polynomial fits for temperature and
Q25, we randomly re-sampled the annual data individually and fit a second order
polynomial to the shuffled data. Repeating this 10,000 times for the two datasets, we
noted that few of the fit coefficients for the 2nd order term approached the value of the
actual data (Fig. 16), emphasizing the significance of 20th century trends.
Simulation of annual Eel River 025 flow at Scotia 1920-2008
2nd order Polynomial values of 10000 samples
600
- 400c
:::::l
o
o 200
82
-0.0016 o
x2 coefficient
0.0016
Simulation of mean annual Scotia Temperature 1926-2008
2nd order Polynomial values of 10000 samples
1000
800
C 600
:::::l8 400
200
-0.0008 -0.0004 o 0.0004 0.0008
x2 coefficient
Figure 16. Histogram of x2 coefficients of polynomial fits to random sampling of
a) Q25 and b) temperature data, compared to the x2 coefficients of the real data
shown in Figure 15.
83
It can be difficult to discern subtle decadal-scale trends from annual data, so to
better highlight multi-year trends, we calculated the cumulative deviation from the long
term mean (Fig. 17). In these plots, a positive slope indicates a series of years with the
variable consistently above the long term average of the period of record, and vice versa.
Annual rainfall and Q25 discharge show similar trends, consistently above average from
the 1940's through the 1970's. Temperature shows the opposite trend, temperature
cooling abruptly in the late 1940's, and consistently cooler temperatures persist from
1947 to 1978. There is a decrease in the rate of change until 2000, and finally a series of
years with above average temperatures.
One of the major influences on decadal climate in the Pacific northwest is the
Pacific Decadal Oscillation or PDO (Mantua and Hare, 2002; Mantua et al., 1997). The
PDO is driven by sea surface temperatures in the Pacific Ocean, and influences terrestrial
climates which vary between cold-wet and warm-dry phases over a period of several
decades. We include the raw PDO values and recognized phase shifts in Figure 17. The
temperature, precipitation and Q25 trends are approximately bracketed by the changes in
the PDO cycle.
4.4. Earthflow Sediment Flux
We estimate an annual sediment flux of 120 ± 9 x 103 m3/a for 62 earthflows
which discharge directly into a creek or major gully. These earthflows have a combined
area of 1.36 x 106~ (6% of the study area), which equates to an erosion rate for the 62
earthflow features of 8.6 ± 0.6 mm/a. When distributed over the study area of 226 km2,
the active earthflows alone produce a sediment yield of 0.53 ± 0.04 mm/a. This is
approximately 60% of the 0.9 mm/a sediment yield calculated for the Eel River (1950-
2000) by Wheatcroft and Sommerfield (2005), suggesting that 6% of the terrain may
supply the majority of the sediment over that period.
84
Cool Phase Warm Phase Cool Phase Warm Phase
o
50
-200
gj -100
c
~ -150
~
.~ -50
(I) 400
e'
co 200
.I::
U 01Il
U
a -200
ro 400
:::]
-600c
c
«
-800
40
ffi ~ 20(I) :::]
ECii
- lii 0coe.~ E
~ 2 -20
-~o
.......
c::
coQ)
E
E
...
.$
0)
c::
..Q
E
.g
co
8
en
4l '1' ! I
i : ;1. ts' AM '.'f'.A f\ i.' ". ANt l\A ~"1..•.1 ' ."'..... . A . /\. '" .!.~'11'~ ,A..,.' '; '.' [\ I8 'V'~~Vvrrv ~~., .r:1JlVv~v",~rV~R1 'V:
a.. -2 . i ' i ' 'I .
4 I I I
I I II I I
! I
I....-A------;---------+..,--,I---\----=------.P
I
I
40
1900 1920 1940 1960 1980 2000
Water Year
Figure 17. Monthly PD~ values from 1900 with annual running mean. The
recognized warm and cool phases are shown. Annual rainfall, 25th percentile Eel
River discharge, and mean annual temperature cumulative deviation from the long
term mean (%). A positive slope indicates sequential years with records above the
long-term average.
85
In Figure 18, we plot the combined annual sediment delivery of 15 earthflows for
which we have time series data. The Lone Pine earthf10w (Figs. 4, 14) has a large effect
on sediment flux from 1981 to 1998, and we present the annual sediment delivery
including and excluding the Lone Pine contribution. Sediment delivery exceeds 70,000
m
3/a from 1944 to1964, increases to 95,000 m3/a from 1964 to 1968, and (excluding
Lone Pine) thereafter has steadily declined.
Annual sediment delivery from 15 earthflows
100000l,.......
cu
«)
E
'-" 80000~
Q)
>
Q) 60000"0
-CQ)
E
'6 40000Q)
lj)
"0
Q)
c 20000:.0
E
0
0
0
1940 1960
Excluding Lonepine
1980
Year
Including Lonepine
I
I
I
2000
Figure 18. Combined annual sediment delivery for 15 large earthflows, showing the
gradual decrease in sediment delivery, especially when the influence of the Lone Pine
earthflow is removed.
86
5. DISCUSSION
5.1 Erosion and Sediment Delivery
Our results show that large, slow moving earthflows are the primary erosional
process along the Eel River, preferentially occurring in weak melange lithology. Mapping
and sediment delivery calculations show that sediment flux attributable to active
earthflow mass-movement along the Eel River canyon accounts for over half the
estimated average sediment yield of the Eel River catchment during the second half of the
20th Century.
We attribute the sediment yield from the remainder of our study area to gullies,
streambank erosion, soil creep, and isolated shallow landslides. In resistant lithology, the
characteristic ridge and valley morphology indicates debris flows are a primary erosional
process, although such terrain was sparse in our field area. There is potential for fluvial
erosion of active earthflow terrain through localized gullying (Roering et aI., 2009),
although we did not account for this in our estimates of earthflow mass-movement.
Estimates of fluvial erosion of the earthflow material at the Van Duzen and Redwook Ck
sites respectively accounted for 50% and 10% of the total earthflow sediment flux
(Kelsey, 1978; Nolan and Janda, 1995). We predict our main stem Eel Study site would
lie somewhere between these values - the earthflows along the Eel River are not as active
or as fluvially incised as those in the Van Duzen, but have higher rates of activity than
those in Redwood Creek.
Given the extensive melange lithology and the large number of active earthflows,
our study area along the Eel River probably has a higher rate of erosion than the
catchment average, so the proportion of erosion attributable to earthflow processes may
not be as large as our results suggest. We consider Brown and Ritter's (1971) calculation
of a sediment yield of 2 mm/a for the Eel River between Dos Rios and Fort Steward a
maximum constraint. This data was recorded in the 2 years following the devastating
1964 flooding, and likely reflects increased sediment yield due to that event rather than a
long term average rate.
87
The ratio of earthflow area to contribution to regional sediment flux for our Eel
River study site (6% to 60%) approximates that found by Kelsey (1978) in the Van
Duzen watershed (1 % to 10%), suggesting that on average, active earthflows in
Franciscan melange erode an order of magnitude more rapidly than surrounding terrain.
This highlights the highly erosive nature of earthflows, which lower their source area at
8.4 mm/a, 10-20 times greater than the estimated regional long term (103_ 104 year)
erosion rate (Gendaszek et al., 2005). As argued by Mackey et al. (2009), this rapid rate
of erosion indicates earthflows are eroding at unsustainably fast rates over periods
beyond a few hundred years. The legacy of this cycle of rapid erosion followed by
periods of dormancy (or permanent inactivity) is reflected in the ubiquity of inactive
earthflow features in the landscape.
An important finding is that is that 1.3% of the study area, or 17% of the active
earthflow terrain showed activity, but did not reach active channels. Earthflows can
redistribute material on the hillslopes, and may not deliver sediment directly to channels.
5.2 LiDAR and Aerial Photos
Despite the challenges of mapping earthflows in fine grained melange lithology,
we found the combination of LiDAR, orthorectified aerial photographs, and localized
field reconnaissance to be a robust technique to define active features. This approach
allowed us to map both in detail and extensively, effectively bridging the gap between
site specific studies and regional reconnaissance level photo-based mapping. Mapping the
movement of individual trees provided a measure of objectivity in identifying earthflow
activity that was not available from analyzing LiDAR maps or features in the field alone.
5.3 Spatial Characteristics
Our results show that the spatial pattern of earthflow activity across the Eel River
Canyon has a systematic distribution. Although the size of the data set is modest (n =
88
122), earthflow sizes show 2 power law distributions. This is significant, as although
catastrophic landslides have well documented power law behavior, such a relationship
has not been demonstrated for large, slow moving earthflows.
The planform shapes of the earthflows vary greatly, and can diverge significantly
from the classical hourglass planform. Many earthflows have complex and intricate
margins, and multiple small tributary earthflows feeding into a centralized transport zone
(Figs. 4, 7) Earthflows can bifurcate around topography, lower ridges, and capture the
drainage area of adjacent terrain.
Typically active earthflows extend from the channel to the ridge (or a topographic
break in slope), and we rarely observe them in a transient state, limited in extent to a
portion of a hillslope. This suggests that once initiated, earthflows can rapidly expand
longitudinally along a hillslope profile. Where earthflows do not span channel to ridge,
they are limited to the upper sections of hillslope (Fig. 7), and constitute partial
reactivation of older largely dormant earthflows features. This observation supports a top-
down driver of longer term earthflow evolution - earthflows reactivate from the top, and
activity propagates downslope, rather than a toe perturbation propagating upslope. This
contrasts with many other types of bedrock landslide which are very sensitive to base
level lowering (Densmore and Hovius, 2000). We argue the availability of readily
mobilized source sediment that can be incorporated into an earthflow-probably governed
by weathering rates, may be the ultimate control on the location and rate at which
earthflows evolve. Figure 19 conceptually describes how an earthflow may evolve
cyclically when they are governed by the availability of source material.
89
~~-Active earthflow
I
I
Main earthflow
stops moving
Weathering of exposed
. rock in upper basin
/
. ,
Deep axial gully erodes
I' earthflow body
4. Rejuvenation
of earthflow
I
. • I
Gullies cause many small i
earthflows in upper basin i
!
/
Baselevel signal
transmitted to the upper
slopes via gullies
~
Further earthflow
body erosion
I 3. Extensive gully
I
, development
I
Figure 19. Conceptual earthflow evolution model. 1) Active earthflow is being
supplied with material from the upper amphitheatre. 2) Eventually the source
material is exhausted, and gullies begin to erode the earthflow body. 3) Gullies
propagate into the source area, mobilizing the weathered slopes. 4) Small flows
coalesce and reactivate the earthflow transport zone.
Within the lithological variability across the study area, active earthflows
preferentially occur in the sheared argillaceous melange unit cm1, and 10% of this
lithology is subject to active slope instability (Fig. 2, Table 1). However, given the poor
bedrock exposure in this region, much of the lithological mapping was undertaken with
aerial photograph interpretation, and slope morphology is a key metric in distinguishing
between different melange units. For example, McLaughlin et al. (2000) note that unit
cm1 exhibits characteristic earthflow morphology, and use the rounded, poorly incised,
90
lumpy and irregular topography as a diagnostic feature during lithological mapping. This
introduces circularity when comparing earthflow activity across argillaceous melange
units. More robust is the observed process dichotomy between the argillaceous and more
resistant sandstone outcrops, with earthflows absent in the latter (Table 1). Earthflows
bifurcate around the harder melange blocks, leaving them as localized topographic highs
extruding out of the melange. The exposed sandstone blocks either weather away in place
by rockfall processes, or they can fail catastrophically if they are sufficiently undermined
by the surrounding earthflow prone terrain. Larger, resistant sandstone outcrops at the
kilometer scale have well incised drainages, and steeper slopes emphasizing the
contrasting erosional regimes that operate within different lithologies in the same
landscape, literally just meters apart.
The median earthflow slope is very similar to the study area terrain, although
earthflows have a narrower slope distribution. The similarity in mean slopes between
earthflows and all terrain supports the contention much of the landscape has been
modulated by earthflow processes, even if they are not currently active. A slope of
approximately 36% may be an approximate limiting threshold slope for earthflow
activity, and represent the residual shear strength of the melange (e.g., Carson and Petley,
1970; Hutchinson, 1967; Skempton, 1964).
The preferred southwesterly aspect matches the observation of Kelsey (1978),
who suggested that slopes with a southerly aspect are dryer, and do not support conifer
growth. Across our study area lithology appears to have a stronger influence on
vegetation than aspect. Many north facing slopes have the open grassland susceptible to
earthflow activity. The primary difference in northerly and southerly aspect is the sun
exposure, so the preferred southerly aspect is likely attributable to differences in
evaporation and soil moisture. McSaveney and Griffiths (1987) suggest drought may be a
necessary precursor for earthflow activity. They argue that deep desiccation cracks
penetrate into the earthflow mass during drought, allowing easy access for water when
rainfall resumes, whereas when surface moisure levels increase the permeability of the
earthflow decreases markedly. In the Eel River catchment, south facing slopes become
91
highly desiccated over the summer and cracks up to 30 mm wide extend to depths over 1
m. Northerly facing slopes do not dry to the same extent. This aspect governed
asymmetry of drying over the summer months may explain some of the southwesterly
aspect preference we observe in the active earthflows. The northerly facing slopes do
have the morphology of dormant earthflows, so have been active in the past, potentially
under different climatic conditions.
5.4 Temporal Characteristics
The time series of earthflow velocities across 16 features show three distinct
temporal groupings. The variability in the temporal pattern of earthflow movements can
be partially explained by decadal scale climate changes, although earthflow shape and
position in the landscape appear to modulate this pattern. In particular, we consider
sustained wet periods optimal for increased earthflow activity, especially on large
earthflows (Bovis and Jones, 1992). Given the low permeability of earthflow soil
(Iverson and Major, 1987), persistently wet conditions allow pore pressures to build up
within the earthflow over several or more seasons. We can discount land use change,
earthquakes, and base level change as having changed significantly between 1944 and
2006, but acknowledge we have not addressed the cumulative effect of higher frequency
climatic changes such as El Nino.
The first group of earthflows showed rapid movement in the first interval from
1944 (Fig. 12), and then movement gradually slowed or stopped. This group of
earthflows was unable to sustain movement as the wetter conditions persisted through to
the late 1970's, and movement is negligible from 1981. Topographically, these slides are
commonly high on the hillslopes and may have stabilized after initial movement due to
mid-slope toe buttressing.
The second group of earthflows exhibit sustained movement until the late 1970's,
and gradually slowed from approximately 1981 to present. This group of earthflows have
long, narrow transport zones which we argue is the most sensitive recorder of decadal
92
climatic changes (Bovis and Jones, 1992), because they are largely isolated from source
or toe effects. The Boulder Creek and Kekawaka earthflows show an increase in velocity
over the 1960's, and mimic the trends in temperature and Q25 flow, potentially showing a
high sensitivity to the variation in climate. This period also coincides with the December
1964 storm, and we cannot confidently discount the role of extreme rainfall events in
increasing earthflow activity. For example, although we distinguished declining and
sustained temporal behavior (Figs 12-13), if we remove the data point from 1964 to
1968, many of the Figure 12 earthflows would also appear to exhibit declining velocity.
The remaining earthflows are generally slow (1-2 m/a), although some have
isolated surges. Six of these eight earthflows are tightly coupled with major creeks and
channels, and sustained erosion of the toe may dominate earthflow response over the
climatic changes. A clear example of this is the Penstock and Lake Shore earthflows,
whose toes oppose each other across Kekawaka Creek and act as a choke point on the
channel. Other examples, such as the Chamise, Wishbone, Island Mountain, and Lone
Pine landslides feed directly into major channels, which remove the potential buttressing
and topographic stability a compressional toe can provide (e.g. Boulder Creek - Figure 4,
Laufer Earthflow - Figure 12). We also note that this third group of earthflows
commonly feature unconventional shapes, with either small aspect ratios (Lake Edge,
Island Mountain), or are extremely skinny (High White, Penstock).
The climate changes significantly in the 1940's (Figs. 15, 17). From the mid
1940's, rainfall and Q25 discharge are consistently above average, whereas temperatures
are below the average of the period of record (Fig. 15). This time coincides with the PDO
cool phase (1946-1977). As our earliest aerial photos date from 1944, we have no record
of earthflow activity prior to the 1940' s, which impedes our interpretations of long term
behavior. In the decades preceding 1944, rainfall and Q25 river discharge are generally
below the long term average, while median annual daily temperature is above average.
The period from 1924-1947 is recognized as a warm phase of the PD~ (Fig. 17), and if
there was less earthflow activity in from 1920 to 1940, it would support our contention
that decadal scale climate changes influence earthflow activity. In this dryer, warmer
93
climate, we would anticipate less soil moisture, lower groundwater levels, and therefore
we would speculate less earthflow activity, although we stress the amount of pre-1944
earthflow activity remains unconstrained. Although unsubstantiated, it is appealing to
conceive earthflow activity increasing rapidly in response to the climate changes in the
1940's. As such, our study period would document the rapid earthflow adjustment to the
onset of cooler, wetter conditions in 1947.
Some of these earthflows show an interval of rapid movement, such as SE-l in
the mid 1990's. We have little explanation for these events, other than possible natural
variability or possibly triggering from individual storms. Isolated earthflow surges have
been documented (Hutchinson and Bhandari, 1971). One historical example on the Eel
River is the 201 Slide crossing the railway (Fig. 13), reported by Malhase (1938). In
February 1938, this earthflow advanced 13 feet over 24 hours in response to heavy
rainfall.
Generalizing all the temporal earthflow records, with few exceptions, there has
been a decrease in activity since the late 1970's (e.g., Fig. 18). This general reduction in
earthflow movement coincides with more variable climate than persisted from 1947-
1978. Annual rainfall was below average in the period 1983-1992, but has increased
subsequently (Fig. 17). In a departure from the rest of the record, base flow does not
follow the increase in rainfall from 1992, potentially attributable to the increased
temperature (and therefore evaporation) over the past decade. From 1980, we do not
observe the sustained wetter, cooler conditions characteristic of the 1960's, which are
conducive to earthflow activity.
The influence of decadal scale climate changes has been recognized in other
studies. For example, Bovis and Jones (1992) show increased earthflow activity over the
mid 20th Century in British Columbia. Their rainfall cumulative deviation plots are very
similar to those from Scotia, highlighting the widespread effect of PDO changes on
earthflows across the Pacific Northwest. Geertsema et al. (2006) find very similar
temporal patterns of temperature change in British Columbia, although here the change in
94
climate is having the opposite effect, with increasing temperature since the late 1970's
degrading permafrost and increasing landslide activity.
Several earthflows attain maximum velocities in the photo interval 1964-1968
(Fig. 13). This period coincides with both the middle of the PDO cool phase, and the
1964 storm, so it is difficult to distinguish the influence of the extreme rainfall events
from ambient climate conditions. On the photographs, we observe little qualitative
evidence of accelerated earthflow activity or disturbance in response to the December
1964 storm. For example, there is little evidence of substantial toe erosion or disruption
to the gully network on the earthflows, even though the 1964 event had a major effect on
rivers and channels (Kelsey, 1980; Sloan et aI., 2001). This observation also holds for
major storms in 1982 and 1997. Individual rainfall events deliver water too rapidly for
infiltration into the soil, although they may affect gullies on the earthflow surface and
accelerate erosion of the toe.
The PDO cycle is predicted to switch back to a cold phase from 2008, and
measurements of Pacific sea surface temperatures indicate this change is underway (e.g.,
Buis, 2008). If this shift is sustained and the northern California climate returns to that of
the 1940's, we would predict a possible ShOlt term reactivation of earthflows high on the
hillslopes, and gradual increase in activity of the larger earthflows such a Boulder Creek.
Increased earthflow activity over the coming decades will have management implications
for land owners, transport corridors, and aquatic habitat, and management strategies
should be adjusted accordingly.
6. CONCLUSION
A landscape scale theory explaining earthflow movement, sediment flux and the
influence of earthflows on landscape morphology remains a future research objective, but
this study illustrates how earthflows control the evolution of this rapidly eroding
landscape in the northern California Coast Ranges. We use LiDAR and aerial photos to
objectively document earthflow activity and sediment flux at an unprecedented spatial
95
and temporal scale, and with a higher level of accuracy than available with traditional
methods. We find earthflows have a characteristic position in the landscape, and some
earthflows exhibit systematic temporal behavior in response to climatic changes
In fine-grained clay-rich lithology, earthflows can have a major impact on
sediment yield slope morphology. We estimate active earthflows contribute more than
half of the suspended sediment of the Eel River, despite accounting for only 6% of the
terrain area. Given the erosion rate of an active earthflow can exceed the regional rate by
an order of magnitude, we envisage earthflow activity as intermittent, separated by long
periods of dormancy. Most of the argillaceous terrain in our study site shows
topographic evidence of earthflows, suggesting that locations of activity can migrate
through the landscape. We observed 7.3% of the Eel River Canyon to show movement
between 1944 and 2006, and most of this area is represented by large earthflow
complexes spanning channel to ridge, with a southwesterly aspect. In contrast to other
styles of slope failures, earthflows reactivate from the top and this perturbation
propagates downslope, and we suggest long term availability of readily mobilized source
sediment is the ultimate control on earthflow activity and evolution.
The temporal pattern of earthflow movement shows variability which
approximately coincides with changes in decadal scale climate. This is particularly
evident in earthflows with long, well defined transport zones, whereas earthflows with a
less conventional planform behave more stochastically. The lack of information about
pre-1944 earthflow activity tempers some of our interpretations, as we don't know the
historical context of the high rates of movement we document from 1944 to the 1970's
They could either represent an increase attributable to the cooler, wetter climate, or
alternatively we may have witnessed the earthflows in a longer term monotonic decline
of a duration which precedes the photographic record. We argue periods with sustained
rainfall and cool temperatures the most amenable to earthflow activity, and if the
predicted shift to a new PD~ cool phase transpires, earthflow activity in northern
California could again increase in the coming decades.
96
97
APPENDICES
APPENDIX A
14-C DATING OF SKUNK CREEK LANDSLIDE COMPLEX,
SOUTH FORK EEL RIVER, CALIFORNIA
1. Radiocarbon Dating Landslide Movement
In one accessible landslide in the Skunk Creek catchment (Figure 1), we attempted
to constrain the timing of landslide movement. We chose the Skunk Creek landslide
because it has limited disturbance by road construction and historical logging activity,
and appeared to exhibit multiple phases of movement, and the morphology of the
landslide appeared similar to others in the study area. Dating prehistoric landslides is
challenging [Lang et al., 1999], and various techniques including tephrochronology
[Hermanns et al., 2000], cosmogenic exposure age [Ballantyne et al., 1998; Bigot-
Cormier et al., 2005], and radiocarbon dating [Pellegrini et al., 2004] have been utilized.
We focused on dating charcoal in the base of headscarp depressions and graben features
within the landslide mass. Carbonaceous material deposited in the newly opened
depression shortly after slide movement is preserved under the infilling colluvium [Lang
et al., 1999], but due to inheritance issues with detrital carbon (time taken from tree
growth to eventual burial), the age should be considered a maximum constraint [McGill
et al., 2002] and the younger of multiple samples is preferred. We identified a range of
suitable target depressions on the LiDAR derived DEM, and using a combination of hand
auguring and digging, excavated to the base of the infilling colluvium at four locations on
the slide. Two samples of detrital carbon from each of the four sites on the landslide were
collected and analyzed for 14C by accelerator mass spectrometry at Lawrence Livermore
98
National Laboratory, and converted to calendar years using the conversion of Fairbanks
et al [2005].
2. Skunk Creek Landslide Age Constraints
The oldest ages for Skunk Creek landslide activity are late-Pleistocene (-13.7 ka),
and come from an infilled graben below the primary headscarp (Table 1; Figure 1). The
14C ages from Skunk4B are from a deep headscarp depression near the watershed
boundary, just above the headscarp of the younger internal slide described below.
Although stratigraphically consistent, there is a large age difference between the two
Skunk4B samples. Skunk 3B and Skunk 5B come from the margins of a secondary
failure within the parent landslide mass, and show a distinct cluster at 2.5-2.8 ka.
Sample Depth (m) Radiocarbon age Calendar Age*
Mean std deY mean std deY
Skunk2Al-AAA 2.63 (auger) 12600 60 14827 111
Skunk2B-AAA 2.54 (auger) 11840 100 13745 133
Skunk3Bl-AAA 1.5 (soil pit) 2765 45 2858 52
Skunk3B2-AAA 1.5 (soil pit) 2545 35 2691 66
Skunk4B 185 1.85 (auger) 3290 15 3506 31
Skunk4B 195 1.95 (auger) 8790 20 9796 60
Skunk5B 150 1.50 (auger) 2460 15 2563 89
Skunk5B 165 1.65 (auger) 2780 15 2866 19
Table 1. Radiocarbon dates on detrital charcoal from the Skunk Creek Landslide.
*Based on conversion of Fairbanks et al [2005] available at:
http://www.radiocarbon.LDEO.columbia.edu/
99
Sku k k
Landslide
Legend
(\T' LandslideI- Headscarp
( Interior scarp
Slide Margin
Movement
Direction
Catchment
Boundary
Radiocarbon
Samples
High: 76
Low: 0
2m Contour Interval
Figure 1, Slope and contour map of the Skunk Creek landslide complex showing
the headscarp and major morphological features. Numbers 2-5 denote sample
locations for charcoal, described in Table I. Two samples were taken at each
location, with the younger age shown on the map. The 14C ages suggest the
landslide failed -13.5 ka, whereas the morphologically distinct internal feature
reactivated approximated 2.6 ka.
100
3. Climatic Interpretation
The 14C ages from the Skunk Creek landslide support our initial hypothesis of
multiple phases of movement within the Skunk Creek landslide basin. The sample from
the extensional graben near the primary headscarp (Skunk2) suggests a late Pleistocene
age for the landslide. Deep seated landsliding in the Northern Californian Coast Ranges
has often been attributed to the climatic conditions of the late Pleistocene, but has not to
our knowledge yet been confirmed by carbon dating. The significance of the Skunk4 ages
is difficult to interpret, as there is a difference of over 6000 years between samples taken
only lOcm apart. The Skunk 4B site has an age of 9.8 ka, which corresponds with the
earliest Holocene, and materially later than the samples from the headscarp feature at
Skunk2B and 2A, implying different parts of the landslide may have been active at
different times. Given inheritance issues, however, we favor the younger age of 3.5 ka.
We are more confident in the ages for Skunk 3B and 5B, where four ages from 2 different
locations on the inset landslide feature cluster tightly around 2.5-2.9 ka. This landslide
has sharper features than the larger parent slide, which would be expected given the
younger age, and indicates this feature was mobilized within the mass of the older
landslide about 2.6 ka. The multiple phases of movement in the Skunk slide illustrate
how subsidiary failures can occur within a previously failed landslide mass, potentially
many thousands of years later. The older section of the Skunk landslide is
morphologically similar to neighboring slides, and suggests many of the mapped
landslides may have failed in the Late-Pleistocene.
Climatically, the latest Pleistocene conditions in the northern Californian Coast
Ranges are thought to have been wetter than those at the last glacial maximum, and
wetter and colder than modem conditions [Adam and West, 1983; Daniels et al., 2005;
Rypins et al., 1989]. From an offshore pollen core near the mouth of the Eel River,
Barron et al [2003] interpret the latest Pleistocene «16 ka) to exhibit dry but changeable
conditions, with increasing precipitation approaching the Holocene transition. Spikes in
alder pollen indicate ground disturbance, possibly due to widespread landsliding. The
oldest dates from the Skunk Ck landslide therefore correspond with a changing climate
101
with increasing precipitation, conditions that would presumably promote landsliding. The
early Holocene climate is reportedly dryer and warmer [Adam and West, 1983], before a
regional shift to slightly cooler and wetter conditions in the late Holocene [Daniels et al.,
2005], with the establishment of the modem redwood forest and a maritime climate from
about 3.5 ka [Barron et al., 2003]. The radiocarbon dates from 3.5 ka through 2.6 ka
coincide with this transition from the dry mid-Holocene to the establishment of redwood
forest and wetter modem climatic conditions. Given the lack of deep seating landsliding
we observe today, this may reflect a critical period where increased precipitation
generated landsliding, before it was mitigated by the expansion of the dense redwood
forest over the past 3 ka. Both the late Pleistocene (13.5 ka) and Late Holocene (3.5-2.5
ka timeframes at which we constrain landslide movement relate to periods of climatic
change to a wetter climate, and such transitional periods may be instrumental in
generating large slope failures.
References
Adam, D.P., and G.J. West, Temperature and Precipitation Estimates through the Last
Glacial Cycle from Clear Lake, California, Pollen Data, Science, 219 (4581), 168-
170, 1983.
Ballantyne, c.K., J.O. Stone, and L.K. Fifield, Cosmogenic CI-36 dating of postglacial
landsliding at The Storr, Isle of Skye, Scotland, Holocene, 8 (3), 347-351, 1998.
Barron, J.A., L. Heusser, T. Herbert, and M. Lyle, High-resolution climatic evolution of
coastal northern California during the past 16,000 years, Paleoceanography, 18
(1),2003.
Bigot-Cormier, F., R. Braucher, D. Bourles, Y. Guglielmi, M. Dubar, and J.F. Stephan,
Chronological constraints on processes leading to large active landslides, Earth
and Planetary Science Letters, 235 (1-2), 141-150,2005.
102
Daniels, M.L., RS. Anderson, and C. Whitlock, Vegetation and fire history since the
Late Pleistocene from the Trinity Mountains, northwestern California, USA,
Holocene, 15 (7), 1062-1071,2005.
Fairbanks, RG., RA Mortlock, T.C. Chiu, L. Cao, A Kaplan, T.P. Guilderson, T.W.
Fairbanks, AL. Bloom, P.M. Grootes, and M.J. Nadeau, Radiocarbon calibration
curve spanning 0 to 50,000 years BP based on paired Th-230/U-234/U-238 and
C-14 dates on pristine corals, Quaternary Science Reviews, 24 (16-17), 1781-
1796,2005.
Hermanns, RL., M.H. Trauth, S. Niedermann, M. McWilliams, and M.R Strecker,
Tephrochronologic constraints on temporal distribution of large landslides in
northwest Argentina, Journal of Geology, 108 (1), 35-52,2000.
Lang, A, J. Moya, J. Corominas, L. Schrott, and R Dikau, Classic and new dating
methods for assessing the temporal occurrence of mass movements,
Geomorphology, 30 (1-2),33-52, 1999.
McGill, S., S. Dergham, K. Barton, T. Berney-Ficklin, D. Grant, C. Harding, K. Hobart,
R. Minnich, M. Rodriguez, E. Runnerstrom, J. Russell, K. Schmoker, M.
Stumfall,1. Townsend, and J. Williams, Paleoseismology of the San Andreas fault
at Plunge Creek, near San Bernardino, southern California, Bulletin of the
Seismological Society ofAmerica, 92 (7), 2803-2840,2002.
Pellegrini, G.B., N. Surian, and C. Urbinati, Dating and explanation of late glacial -
Holocene landslides: a case study from the southern Alps, Italy, Zeitschrijt Fur
Geomorphologie, 48 (2), 245-258, 2004.
Rypins, S., S.L. Reneau, R. Byrne, and D.R Montgomery, Palynological and
Geomorphic Evidence for Environmental-Change During the Pleistocene
Holocene Transition at Point-Reyes Peninsula, Central Coastal California,
Quaternary Research, 32 (1), 72-87, 1989.
103
APPENDIXB
SUPPLEMENTARY MATERIAL FOR CHAPTER III
Photo Source CFL Fiducials Image ResolutionDate number (nun) Format (dpi)
8-19-1944 DDD 54-46 DoD/USFS 210 (est) Measured Scanned print 800
6-13-1964 VBAG2-81 USGS 151.67 Measured Scanned print 1800
7-22-1968 VBZX 1-163 USGS 152.45 Measured Scanned print 1800
7-08-1981 VEZF2-4 USGS 152.347 Supplied Scanned print 1800
7-26-1991 91CA 10-72 WAC 152.903 Supplied Scanned 1200Diapositive
9-24-2006 - Unfiltered NCALM NA NA 1mDEM 1m9-26-2006 Lidar
Table 1
Orthorectification requires information including camera lens information (focal length,
fiducal distances), a digital elevation model (DEM) of the terrain, and ground control
points (GCP's) to correlate image position to the ground. Where possible, we used
Camera Focal Length (CFL) information from camera calibration reports, although the
1944 photo parameters had to be estimated from a database (Slama 1980). In the earlier
photos (1944-1968) we measured the distance between fiducials manually, and we
assumed a perfect principal point of focus (x,y =0,0). In the 1981 and 1991 photos,
fiducial measurements were supplied in the calibration reports. We selected
approximately 30 GCP's for each image (e.g. buildings, rock outcrops, trees on stable
ridges or terraces) using unfiltered LiDAR as the reference image, and the bare earth
LiDAR 1 m DEM to construct the camera model in Erdas Imagine 9.1 software. (Slama,
C.C., 1980, Manual of Photogrammetry: Falls Church, VA, American Society of
Photogrammetry, p. 1056.)
Table 2. Meteoric lOBe data. For each depth interval, soil samples were oven dried, crushed, and pulverized into a fine powder.
From the pulverized mix, a 0.5 g sample was processed following Stone (1998) (See also:
http://depts.washington.edu/cosmolab/chem/fusionmethod.pdf). Blank samples were processed with every 4-8 soil samples to
constrain the analytical error.
Sample Depth Longitude Latitude Elevation 10-Be Cone Error Inventory Error 10-Be Age
name (em) (UTM) (UTM) (m) (Atomslem3) (Atoms/em3) (atoms/em2) (atoms/em2) age Error
Pit 1 KW1015 0-15 458470 4441020 606 2.9E+07 1.8E+06 1.0E+09 3.9E+07 620 130
KW11530 15-30 2.5E+07 1.5E+06
KW13045 30-45 6.6E+05 2.1E+05
KW14560 45-60 7.8E+05 3.2E+05
KW16075 60-75 2.2E+05 2.0E+05
KW17590 75-90 1.4E+07 1.2E+06
Pit 2 KW2015 0-15 458400 4440759 536 2.1E+07 1.2E+06 4.2E+08 1.9E+07 250 50
KW21530 15-30 4.0E+06 4.0E+05
KW23045 30-45 6.6E+05 2.1 E+05
KW24560 45-60 1.4E+06 2.3E+05
KW26090 60-90 4.0E+05 3.7E+05
KW290120 90-120 -6.4E+03 -2.6E+05
KW2120145 120-145 -1.6E+04 -1.6E+05
KW2145170 145-170 1.7E+05 3.8E+05
Pit 3 KW3015 0-15 458211 4440484 473 2.5E+07 1.9E+06 6.8E+08 3.3E+07 400 80
KW31530 15-30 1.5E+07 1.2E+06
KW33045 30-45 1.5E+06 4.8E+05
KW34560 45-60 8.9E+05 3.3E+05
KW36090 60-90 5.9E+05 3.7E+05
KW390120 90-120 -9.1E+04 -1.4E+05
KW3120160 120-160 2.2E+05 1.3E+05
KW3160200 160-200 3.5E+05 2.9E+05
-0+>-.
Sample Depth Longitude Latitude Elevation 10·Be Cone Error Inventory Error 10·Be Age
name (em) (UTM) (UTM) (m) (Atoms/em3) (Atoms/em3) (atoms/em2) (atoms/em2) age Error
Pit4 KW50010 0-10 458164 4440302 439 7.0E+07 4.1E+06 2.9E+09 2.5E+08 1700 400
KW51020 10-20 4.9E+07 2.9E+06
KW52030 20-30 3.1E+07 1.3E+07
KW53040 30-40 4.9E+07 2.0E+07
KW54050 40-50 3.8E+07 2.2E+06
KW55060 50-60 2.1E+07 1.4E+06
KW56080 60-80 1.8E+07 1.4E+06
Crown KW40010 0-10 458548 4441112 638 2.8E+08 1.7E+07 2.0E+10 4.3E+08 12000 2500
KW41020 10-20 3.2E+08 1.9E+07
KW42030 20-30 3.2E+08 1.8E+07
KW43040 30-40 3.3E+08 1.7E+07
KW44050 40-50 2.9E+08 1.6E+07
KW45060 50-60 2.7E+08 1.6E+07
KW46070 60-70 1.8E+08 9.8E+06
-oVI
106
Figure 1. Earthflow Geomorphology - The earthflow has a steep arcuate
headscarp defining an amphitheatre-like collection or source zone, which
enlarges by retrogressive failure ofthe headscarp and provides the earthflow
material. Failing blocks back rotate in a listric manner and fall into the
hummocky upper amphitheatre or source zone, where they disintegrate, and
themselves become buried by disintegration and raveling from subsequent failure
blocks above. The upper section of the source zone gradually transitions
downslope into a smoother, narrow transportation zone and (although moving)
the surface becomes more stable, allowing shrubs to grow. The transportation
zone transitions into a compressive broad section with thrusting and folding.
Down slope of the road, the flow steepens, and surficial slumps disrupt the
earthflow surface. At the toe the earthflow encounters Kekawaka Creek. A steep,
interlocking boulder cascade (removed by LiDAR processing) inhibits mass flow
movement into the creek, but erosion of the toe occurs by slumping from the toe
face and erosion of fines during high flows. Due to the boulder constriction, the
earthflow toe mass is subject to compression and over-thrusting, and is escaping
laterally onto an alluviated reach downstream. The background image is LiDAR
derived slope (lm resolution).
107
650m
600rn
Alluvial reach
Gully
50m Contour
Boulders
Soil Pit
Kekawaka Ck
Pond I Swamp
Shear Margin
Rock Outcrop
Inlernallhrusl
Linear Feature:
-Sharp
-Faint
Minor scarp
Toe lhrust
Flow Direction
250m
c
4Q0I11_ -
~ Headscarp
F
01
1:-=C-
'iii E
I: :JCII-
.... 1Il
><'0
WI:
til
+
401178°
-123.4908°
108
11968-1981
10
1944-19641
10 ~
------.---. ~e~
-30 -20 -10'~: 10
-lO,ft '
.
U ••
" ~'20'
..
-30
-40
-50
-30
11964-19681
10
0,.
-20
-20
-30
-40
-50
-30 -20
,',
"
1''r.-r-·
-10 :"'~: 10
-to" ~, ,
. . "' .....
" ,
, -30
1981-1 991 1
10 i
-.--r- '"'-'.{-"':f'------,
-30 -20 -10 "~'" 10
.-,i.Q';;~ "
• • ° 0 r 0°
: ...
-20' .
-30
-40 .
-50
-30
1991-2006]
10 i
.~
a" ~
-20 -10, ~ 10
~~'9·. "
"
-20
-30
-40 1
-50 ~
Legend
20m It
Tree on Earthfiow
Feature on Stable Ground
Average Stable Tree
(+/- 2 Standard DeViations)
95% confidence any point outside
the ellipse is earthflow movement
as opposed to orthorectification error
Figm'e 20 Comparison of stable features vs. trees on the earthflow for each
photographic interval. We are 95% confident features plotting outside the circle
(2 standard deviations) reflect earthflow movement, as opposed to
orthorectification errors or vegetation change.
109
APPENDIXC
EVIDENCE FOR A LANDSLIDE DAMMED PALEO-LAKE ALONG
THE MAIN STEM EEL RIVER
During the course of researching earthflows along the Eel River, I have
discovered several lines of evidence suggesting the Eel River was one impounded behind
a large landslide dam.
Many terrace features cluster within a consistent elevation (-240-242m asl), as
shown on Figures 1 and 2. I have visited many of these features in the field, and they are
typically gently undulating surfaces with bioturbated fine grained sediments. Small
creeks and gullies at this elevation commonly exhibit a subtle wedge of fine-grained
sediments, consistent with delta-like deposition.
110
--- , ...._-------------'.
Figure 1. Oblique view up the Eel River canyon. Map is filtered shaded relief
LiDAR. Blue line represents elevation between 240 m and 243 m, showing broad
flat areas defining the edge of the paleolake.
111
Figure 2. Southern section of the LiDAR coverage - filtered shaded relief. Red
shading depicts elevation from 240 to 243 ill. Boxes show where there are flat
terrace features at this elevation, and possibly represent shoreline features from
the landslide dammed paleolake.
112
113
I discovered outcrops of finely laminated sediments along Kekawaka Creek at an
elevation of -230 m (457497 E, 4438948 N - UTM Zone lON). The deposits are
attributable to suspension fallout in standing water (Figs. 3--4). Packages of clast
supported angular gravels separate the fine grained laminated sequences, and I interpret
these deposits as debris flows intermittently entering the lake.
Figure 3. Exposure of finely laminated deposits interlayered with clast supported
angular gravels. Grid is 0.5 m squares.
114
Figure 4. Sample of finely laminated silts. Under a binocular microscope, I
picked fine detrital charcoal for 14C dating. The silts showed extensive faulting
and deformation, probably due to post depositional slumping. Knife handle is 94
mm long.
Sample Lab Lab ID Sample Source
14C Age
Enor Calendar Eo-orID Description (yr) Age (yr)
KW I C LLNL 144860 Bark Paleosol 210 30 230 80
KW 3-1 LLNL 144861 Detrital Laminated 21440 80 25745 154
charcoal silts
KW4-1 LLNL 144862 Detrital Laminated 19020 60 22592 95
charcoal silts
KW2:8 Beta Beta- Detrital Laminated 36050 300 41306 320256979 Charcoal silts
KW2:9 Beta Beta- Wood Laminated 80 40 N/A N/A256980 sil ts
Table 1. 14C ages of fine grained deposits. LLNL is Lawrence Livermore
National Laboratory, and Beta is Beta Analytic. Calibration is from Fairbanks et
al. (2005).
115
Accelerator Mass Spectrometry dating of detrital charcoal in the finely laminated
sediments found in the Kekawaka Creek exposure gives a range of radiocarbon ages
(Table 1). Both intact samples (KW lC and KW2:9) have modern ages, and I consider
these modern roots. The detrital charcoal samples all returned late Pleistocene ages.
Detrital charcoal should be assumed a maximum age given inheritance issues, and the
younger of multiple samples is preferred. As there are samples reasonably close in age at
22,592 and 25,745 years, these dates are preferred to the older date. I interpret the
sediments to have been deposited after the youngest age, possibly just prior to the Last
Glacial Maximum.
We observe a large landslide scar coming off Neafus Pk (Fig. 5), a large
sandstone outcrop east of the Eel River. This landslide scar has the volume, relief and
lithology to conceivably dam the Eel River for a period sufficient for a lake to create
shoreline features, and for lacustrine deposits to accumulate.
116
Figure 4. Oblique view of Nefus Peak showing a large landslide scar just below
the ridgeline. Landslide scar is 454760 E, 4442860 N (UTM Zone 10).
Most pertinent to the emthflow research is that some of these apparent shoreline
features are cut into the toes of huge dormant earthflows flanking the Eel River canyon.
This suggests the earthflows have been dormant at least since the lake was present,
potentially since before the Last Glacial Maximum, and provides a novel relative means
of constraining the timing of earthflow activity
117
References for Appendix C
Fairbanks, R. G., Mortlock, R. A, Chiu, T. C., Cao, L., Kaplan, A, Guilderson, T. P.,
Fairbanks, T. W., Bloom, A L., Grootes, P. M., and Nadeau, M. 1., 2005,
Radiocarbon calibration curve spanning 0 to 50,000 years BP based on paired Th-
230/0-234/0-238 and C-14 dates on pristine corals: Quaternary Science Reviews,
v. 24, no. 16-17, p. 1781-1796.
APPENDIXD
AERIAL PHOTOGRAPHS USED IN CHAPTER IV
118
Date Photo Source CFL Fiducials hnage Resolution
numbers (mm) Format (dpi)
8-19-1944 DDD 54-30:50 DoD 210 Measured Scanned print 800DDD 57-39:59 USFS-6RNF (est.)
7-19-1954 CVL 6N 66:70 HumCo 209.55 Measured Scanned print 1200
6-24-1960 HS 36 07:15 HumCo 303 Measured Scanned print 1200(est.)
6-13-1964 VBAG 1-148, USGS 151.67 Measured Scanned print 18002-81:85
1968 AF1968268 Air Force c/o 151.484 Measured Scanned Print 1600USFS-6RNF
7-22-1968 VBZX 1- USGS 152.45 Measured Scanned print 1800160:163
7-10-1972 F60BP USFS-6RNF 152.57 Supplied Scanned print 16000172 108
7-25-1972 MEND 9-77 Mendocino Co 209.55 Measured Scanned print 1200MEND 10-84
7-22-1976 VEEX 1- UC Berkeley 152.21 Supplied Scanned 240097:104 Diapositive
7-08-1981 VEZF2-4 USGS 152.347 Supplied Scanned print 1800
7-26-1991 10-23:25 WAC (91CA) 152.903 Supplied Scanned 120010-66:72 Diapositive
08-12-1998 10477-194:195 USGS 153.212 Supplied Scanned Print 180010480-256:259
9-24-2006 - Unfiltered NCALM NA NA ImDEM 1m9-26-2006 Lidar
Table 1. Aerial photos used to track earthflow movement in Chapter IV. Est. indicates the camera
focal length (CFL) was estimated.
Sources of aerial photographs in Table 1:
DoD: Department of Defense
HumCo: Humboldt County Department of Public Works, Eureka, CA
USGS: US Geological Survey Earth Explorer (Online store)
USFS-SRNF: US Forest Service, Six Rivers National Forest, Eureka, CA
Mendocino Co: Mendocino County Assessors Office, Ukiah, CA
UC Berkeley: UC Berkeley Earth Science and Map Library
WAC: Western Air Corporation, Eugene, OR
NCALM: National Center for Airborne Laser Mapping
119
120
REFERENCES
Chapter I
Bovis, M. J., and Jones, P., 1992, Holocene History of Earthflow Mass Movements in
South-Central British-Columbia - the Influence of Hydroclimatic Changes:
Canadian Journal ofEarth Sciences, v. 29, no. 8, p. 1746-1755.
Brown, W. M., and Ritter, J. R, 1971, Sediment Transport and Tubidity in the Eel River
Basin, California: Geological Survey Water-Supply Paper, v. 1986, p. 67.
Chorley, R J., Dunn, A J., and Beckinsale, R P., 1991, Historical and regional
geomorphology, 1890-1950, The history of the study oflandforms; or, The
development of geomorphology: New York, Wiley.
Coe, 1. A, Ellis, W. L., Godt, J. W., Savage, W. Z., Savage, 1. E., Michael, J. A, Kibler,
1. D., Powers, P. S., Lidke, D. J., and Debray, S., 2003, Seasonal movement of the
Slumgullion landslide determined from Global Positioning System surveys and
field instrumentation, July 1998-March 2002: Engineering Geology, v. 68, no. 1-
2, p. 67-101.
Davis, W. M., 1899, The Geographic Cycle: The Geographic Journal, v. 14, no. 5, p. 481-
504.
Dietrich, W. E., Bellugi, D. G., Sklar, L. S., Stock, J. D., Heimsath, A M., and Roering,
J. J., 2003, Geomorphic Transport Laws for Predicting Landscape Form and
Dynamics, in Wilcock, P., and Iverson, R M., eds., Prediction in
Geomorphology: Washington DC, American Geophysical Union, p. 103-132.
Dietrich, W. E., and Perron, J. T., 2006, The search for a topographic signature of life:
Nature, v. 439, no. 7075, p. 411-418.
Dwyer, M. J., Scott, R G., and Lorens, P. J., 1971, Landslide Conditions and Related
Sediment Production on the Eel River: California Department of Water
Resources.
Iverson, R M., and Major, J. J., 1987, Rainfall, Groundwater-Flow, and Seasonal
Movement at Minor Creek Landslide, Northwestern California - Physical
121
Interpretation of Empirical Relations: Geological Society of America Bulletin, v.
99,no.4,p.579-594.
Kelsey, H. M., 1978, Earthflows in Franciscan Melange, Van Duzen River Basin,
California: Geology, v. 6, no. 6, p. 361-364.
Malamud, B. D., Turcotte, D. L., Guzzetti, F., and Reichenbach, P., 2004, Landslide
inventories and their statistical properties: Earth Surface Processes and
Landforms, v. 29, no. 6, p. 687-711.
Maquaire, 0., Malet, J. P., Remaitre, A, Locat, J., Klotz, S., and Guillon, J., 2003,
Instability conditions of marly hillslopes: towards landsliding or gullying? The
case ofthe Barcelonnette Basin, South East France: Engineering Geology, v. 70,
no. 1-2, p. 109-130.
Penck, W., 1953, Morphological Analysis of Landforms: London, MacMillan and Co.,
424+ p.
Perron, J. T., Kirchner, J. W., and Dietrich, W. E., 2009, Formation of evenly spaced
ridges and valleys: Nature, v. 460, no. 7254, p. 502-505.
Seidl, M. A, and Dietrich, W. E., 1992, The Problem of Channel Erosion into Bedrock:
Catena Supplement, v. 23, p. 101-124.
Trotter, C. M., 1993, Weathering and Regolith Properties at an Earthflow Site: Quarterly
Journal of Engineering Geology, v. 26, no. 3, p. 163-178.
Chapter II
Abry, P., ONDELETTES ET TURBULENCES - Multiresolutions, algorithmes de
decompositions, invariance d'echelle et signaux de pression, 290 pp., Diderot,
Editeur des Sciences et des Arts, Paris, 1997.
Azanon, J.M., A. Azor, J.V. Perez-Pena, and J.M. Carrillo, Late Quaternary large-scale
rotational slides induced by river incision: The Arroyo de Gor area (Guadix basin,
SE Spain), Geomorphology, 69 (1-4), 152-168,2005.
Barron, J.A., L. Heusser, T. Herbert, and M. Lyle, High-resolution climatic evolution of
coastal northern California during the past 16,000 years, Paleoceanography, 18
(1),2003.
122
Booth, AM., J.J. Roering, and J.T. Perron, Automated landslide mapping using spectral
analysis and high-resolution topographic data: Puget Sound lowlands,
Washington, and Portland Hills, Oregon, Geomorphology, 109 (3-4), 132-147,
2009.
Cendrero, A, and F. Dramis, The contribution of landslides to landscape evolution in
Europe, Geomorphology, 15 (3-4), 191-211, 1996.
Cmden, D.M., and D.J. Varnes, Landslide Types and Processes, in Landslides:
Investigation and Mitigation, edited by AK Turner, and RL. Schuster, pp. 36-
71, National Academy Press, Washington, DC, 1996.
Densmore, AL., RS. Anderson, B.G. McAdoo, and M.A Ellis, Hillslope evolution by
bedrock landslides, Science, 275 (5298), 369-372, 1997.
Densmore, AL., M.A. Ellis, and RS. Anderson, Landsliding and the evolution of
normal-fault-bounded mountains, Journal of Geophysical Research - Solid Earth,
103 (B7), 15203-15219, 1998.
Densmore, AL., and N. Hovius, Topographic fingerprints of bedrock landslides,
Geology, 28 (4),371-374,2000.
Dietrich, W.E., D.G. Bellugi, L.S. Sklar, J.D. Stock, AM. Heimsath, and J.J. Roering,
Geomorphic Transport Laws for Predicting Landscape Form and Dynamics, in
Prediction in Geomorphology, edited by P. Wilcock, and RM. Iverson, pp. 103-
132, American Geophysical Union, Washington DC, 2003.
Dietrich, W.E., and J.T. Perron, The search for a topographic signature of life, Nature,
439 (7075), 411-418, 2006.
Ferguson, R.I., Meander Irregularity and Wavelength Estimation, Journal ofHydrology,
26 (3-4), 315-333, 1975.
Finnegan, N.J., L.S. Sklar, and T.K Fuller, Interplay of sediment supply, river incision,
and channel morphology revealed by the transient evolution of an experimental
bedrock channel, Journal ofGeophysical Research-Earth Suiface, 112 (F3),
2007.
Fuller, T.K, L.A Perg, J.K Willenbring, and K Lepper, Field evidence for climate-
driven changes in sediment supply leading to strath terrace formation, Geology,
37,2009.
Furlong, KP., and S.Y. Schwartz, Influence of the Mendocino triple junction on the
tectonics of coastal California, Annual Review ofEarth and Planetary Sciences,
32, 403-433, 2004.
123
Gangodagamage, c., E. Barnes, and E. Foufoula-Georgiou, Scaling in river corridor
widths depicts organization in valley morphology, Geomorphology, 91 (3-4), 198-
215,2007.
Griffiths, J.S., AB. Hart, AE. Mather, and M. Stokes, Assessment of some spatial and
temporal issues in landslide initiation within the Rio Aguas Catchment, South-
East Spain, Landslides, 2 (3), 183-192,2005.
Harvey, AM., Coupling between hillslopes and channels in upland fluvial systems:
implications for landscape sensitivity, illustrated from the Howgill Fells,
northwest England, Catena, 42 (2-4), 225-250, 2001.
Hasbargen, L.E., and C. Paola, Landscape instability in an experimental drainage basin,
Geology, 28 (12), 1067-1070,2000.
Hilley, G.E., and J.R Arrowsmith, Geomorphic response to uplift along the Dragon's
Back pressure ridge, Carrizo Plain, California, Geology, 36 (5), 367-370,2008.
Hovius, N., C.P. Stark, and P.A Allen, Sediment flux from a mountain belt derived by
landslide mapping, Geology, 25 (3), 231-234, 1997.
Hovius, N., c.P. Stark, H.T. Chu, and J.c. Lin, Supply and removal of sediment in a
landslide-dominated mountain belt: Central Range, Taiwan, Journal of Geology,
108 (1), 73-89, 2000.
Hovius, N., c.P. Stark, M.A Tutton, and L.D. Abbott, Landslide-driven drainage
network evolution in a pre-steady-state mountain belt: Finisterre Mountains,
Papua New Guinea, Geology, 26 (12), 1071-1074, 1998.
Hugenholtz, C., and D. Lacelle, Geomorphic controls on landslide activity in Champlain
sea clays along Green's Creek, Eastern Ontario, Canada, Geographie physique et
Quaternaire, 58 (1), 9-23, 2004.
Iverson, RM., and J.J. Major, Rainfall, Groundwater-Flow, and Seasonal Movement at
Minor Creek Landslide, Northwestern California - Physical Interpretation of
Empirical Relations, Geological Society ofAmerica Bulletin, 99 (4),579-594,
1987.
Jayko, AS., M.C. Blake, RJ. McLaughlin, H.N. Ohlin, S.D. Ellen, and H.M. Kelsey,
Reconnaissance Geologic Map of the Covelo 30- by 60- Minute Quadrangle,
Northern California, US Geological Survey, 1989.
Kelsey, H.M., Earthflows in Franciscan Melange, Van Duzen River Basin, California,
Geology, 6 (6),361-364, 1978.
124
Korup, 0., Geomorphic imprint of landslides on alpine river systems, southwest New
Zealand, Earth Surface Processes and Landforms, 30 (7), 783-800, 2005a.
Korup, 0., Large landslides and their effect on sediment flux in South Westland, New
Zealand, Earth Surface Processes and Landforms, 30 (3),305-323, 2005b.
Korup, 0., Effects of large deep-seated landslides on hillslope morphology, western
Southern Alps, New Zealand, Journal of Geophysical Research - Earth Surface,
111 (Fl), 2006.
Korup, 0., J.J. Clague, R.L. Hermanns, K. Hewitt, AL. Strom, and J.T. Weidinger, Giant
landslides, topography, and erosion, Earth and Planetary Science Letters, 261 (3-
4),578-589,2007.
Lashermes, B., E. Foufoula-Georgiou, and W.E. Dietrich, Channel network extraction
from high resolution topography using wavelets, Geophysical Research Letters,
34 (23),2007.
Lock, l, H. Kelsey, K. Furlong, and A Woolace, Late neogene and quaternary landscape
evolution of the northern California Coast Ranges: Evidence for Mendocino triple
junction tectonics, Geological Society ofAmerica Bulletin, 118 (9-10), 1232-
1246,2006.
Malamud, B.D., D.L. Turcotte, F. Guzzetti, and P. Reichenbach, Landslide inventories
and their statistical properties, Earth Surface Processes and Landforms, 29 (6),
687-711, 2004.
Mather, AE., lS. Griffiths, and M. Stokes, Anatomy of a 'fossil' landslide from the
Pleistocene of SE Spain, Geomorphology, 50 (1-3), 135-149,2003.
McKean, J., and J. Roering, Objective landslide detection and surface morphology
mapping using high-resolution airborne laser altimetry, Geomorphology, 57 (3-4),
331-351,2004.
McKean, J.A, D.l Isaak, and C.W. Wright, Geomorphic controls on salmon nesting
patterns described by a new, narrow-beam terrestrial-aquatic lidar, Frontiers in
Ecology and the Environment, 6 (3), 125-130, 2008.
Meunier, P., N. Hovius, and AJ. Haines, Regional patterns of earthquake-triggered
landslides and their relation to ground motion, Geophysical Research Letters, 34
(20),2007.
Niemi, N.A, M. Oskin, D.W. Burbank, AM. Heimsath, and E.J. Gabet, Effects of
bedrock landslides on cosmogenically determined erosion rates, Earth and
Planetary Science Letters, 237 (3-4),480-498, 2005.
125
Ouimet, W.B., KX. Whipple, L.H. Royden, Z.M. Sun, and Z.L. Chen, The influence of
large landslides on river incision in a transient landscape: Eastern margin of the
Tibetan Plateau (Sichuan, China), Geological Society ofAmerica Bulletin, 119
(11-12), 1462-1476,2007.
Palmquist, R.c., and G. Bible, Conceptual Modeling of Landslide Distribution in Time
and Space, Bulletin of the International Association ofEngineering Geology, 21,
178-186, 1980.
Pisias, N.G., A.c. Mix, and L. Heusser, Millennial scale climate variability ofthe
northeast Pacific Ocean and northwest North America based on radiolaria and
pollen, Quaternary Science Reviews, 20 (14), 1561-1576,2001.
Roering, J.J., J.W. Kirchner, and W.E. Dietrich, Characterizing structural and lithologic
controls on deep-seated landsliding: Implications for topographic relief and
landscape evolution in the Oregon Coast Range, USA, Geological Society of
America Bulletin, 117 (5-6), 654-668, 2005.
Schmidt, KM., and D.R. Montgomery, Limits to Relief, Science, 270 (5236),617-620,
1995.
Seidl, M.A., and W.E. Dietrich, The Problem of Channel Erosion into Bedrock, Catena
Supplement, 23, 101-124, 1992.
Stark, c.P., A self-regulating model of bedrock river channel geometry, Geophysical
Research Letters, 33 (4), 2006.
Stock, J.D., and W.E. Dietrich, Erosion of steepland valleys by debris flows, Geological
Society ofAmerica Bulletin, 118 (9-10), 1125-1148, 2006.
Terzaghi, K, Mechanism of Landslides, in Application of geology to engineering
practice, edited by S. Paige, pp. 83-123, Geological Society of America, New
York, 1950.
Van den Eeckhaut, M., J. Poesen, G. Govers, G. Verstraeten, and A. Demoulin,
Characteristics of the size distribution of recent and historical landslides in a
populated hilly region, Earth and Planetary Science Letters, 256 (3-4),588-603,
2007.
Van Den Eeckhaut, M., J. Poesen, G. Verstraeten, V. Vanacker, J. Moeyersons, J.
Nyssen, and L.P.H. van Beek, The effectiveness of hillshade maps and expert
knowledge in mapping old deep-seated landslides, Geomorphology, 67 (3-4),351-
363,2005.
126
Whipple, KX., Bedrock rivers and the geomorphology of active orogens, Annual Review
ofEarth and Planetary Sciences, 32, 151-185,2004.
Wieczorek, G.P., Preparing a detailed landslide-inventory map for hazard evaluation and
reduction, Bulletin of the Association ofEngineering Geologists, 21 (3),337-342,
1984.
Wills, c.J., and T.P. McCrink, Comparing landslide inventories: The map depends on the
method, Environmental & Engineering Geoscience, 8 (4),279-293,2002.
Zhou, Z.Q., Wavelet energy spectrum for time-frequency localization of earthquake
energy, International Journal of Imaging Systems and Technology, 13 (2), 133-
140,2003.
Chapter III
Baum, RL., Messerich, J., and Fleming, RW., 1998, Surface deformation as a guide to
kinematics and three-dimensional shape of slow-moving, clay-rich landslides,
Honolulu, Hawaii: Environmental & Engineering Geoscience, v. 4, p. 283-306.
Bovis, M.J., and Jones, P., 1992, Holocene History of Earthflow Mass Movements in
South-Central British-Columbia - the Influence of Hydroclimatic Changes:
Canadian Journal of Earth Sciences, v. 29, p. 1746-1755.
Brown, W.M., and Ritter, J.R, 1971, Sediment Transport and Tubidity in the Eel River
Basin, California: Geological Survey Water-Supply Paper, v. 1986, p. 67.
Coe, J.A., Ellis, W.L., Godt, J.W., Savage, W.Z., Savage, J.E., Michael, J.A., Kibler,
J.D., Powers, P.S., Lidke, D.J., and Debray, S., 2003, Seasonal movement of the
Slumgullion landslide determined from Global Positioning System surveys and
field instrumentation, July 1998-March 2002: Engineering Geology, v. 68, p. 67-
101, doi: 1O.1016/S0013-7952(02)00199-0.
Dietrich, W.E., and Perron, J.T., 2006, The search for a topographic signature of life:
Nature, v. 439, p. 411--418, doi: 1O.1038/nature04452.
Fuller, T.K, Perg, L.A., Willenbring, J.K, and Lepper, K, 2009, Field evidence for
climate-driven changes in sediment supply leading to strath terrace formation:
Geology, v. 37, 467--470
Iverson, RM., and Major, J.J., 1987, Rainfall, Groundwater-Flow, and Seasonal
Movement at Minor Creek Landslide, Northwestern California - Physical
127
Interpretation of Empirical Relations: Geological Society of America Bulletin,
v. 99, p. 579-594, doi: 10.1130/0016-7606(1987)99<579:RGFASM>2.0.CO;2.
Keefer, D.K., and Johnson, AM., 1983, Earth Flows: Morphology, Mobilization and
Movement: Geological Survey Professional Paper, v. 1256, p. 1-56.
Kelsey, H.M., 1978, Earthflows in Franciscan Melange, Van Duzen River Basin,
California: Geology, v. 6, p. 361-364, doi: 10.1130/0091-
7613(1978)6<361:EIFMVD>2.0.CO;2.
McKean, J.A, Dietrich, W.E., Finkel, RC., Southon, J.R, and Caffee, M.W., 1993,
Quantification of Soil Production and Downslope Creep Rates from Cosmogenic
Be-lO Accumulations on a Hillslope Profile: Geology, v. 21, p. 343-346, doi:
1O.1130/0091-7613(1993)021<0343:QOSPAD>2.3.CO;2.
Monaghan, M.C., and Elmore, D., 1994, Garden-Variety Be-lO in Soils on Hill Slopes:
Nuclear Instruments & Methods in Physics Research. Section B, Beam
Interactions with Materials and Atoms, v. 92, p. 357-361, doi: 10.1016/0168-
583X(94)96034-8.
Pavich, MJ., Brown, L., Harden, J., Klein, J., and Middleton, R, 1986, Be-l0
Distribution in Soils from Merced River Terraces, California: Geochimica et
Cosmochimica Acta, v. 50, p. 1727-1735, doi: 10.1016/0016-7037(86)90134-1.
Pavich, MJ., and Chadwick, O.A, 1997, lOBe soil inventories along a climatic gradient
on Kohala Mountain: Geological Society of America, Cordilleran Section, 93rd
annual meeting Abstracts with Programs, v. 29, p. 56.
Stone, J., 1998, A rapid fusion method for separation ofberyllium-lO from soils and
silicates: Geochimica et Cosmochimica Acta, v. 62, p. 555-561, doi:
10.1016/S0016-7037(97)00369-4.
Syvitski, J.P., and Morehead, M.D., 1999, Estimating river-sediment discharge to the
ocean: application to the Eel margin, northern California: Marine Geology,
v. 154, p. 13-28, doi: 10.1016/S0025-3227(98)00100-5.
Walstra, J., Dixon, N., and Chandler, J.H., 2007, Historical aerial photographs for
landslide assessment: two case histories: Quarterly Journal of Engineering
Geology and Hydrogeology, v. 40, p. 315-332, doi: 10.1144/1470-9236/07-011.
Wolf, P.R, and Dewitt, B.A, 2000, Elements of Photogrammetry. With Applications in
GIS: Boston, MA, McGraw-Hill.
You, c.F., Lee, T., and Li, Y.H., 1989, The Partition of Be between Soil and Water:
Chemical Geology, v. 77, p. 105-118, doi: 10.1016/0009-2541(89)90136-8.
128
Zhang, X.B., Phillips, c., and Pearce, A, 1991, Surface Movement in an Earthflow
Complex, Raukumara Peninsula, New-Zealand: Geomorphology, v. 4, p. 261-
272, doi: 1O.1016/0169-555X(91)90009-Y.
Chapter IV
Angeli, M. G., Gasparetto, P., Menotti, R. M., Pasuto, A, and Silvano, S., 1996, A visco-
plastic model for slope analysis applied to a mudslide in Cortina d'Ampezzo,
Italy: Quarterly Journal of Engineering Geology, v. 29, p. 233-240.
Baum, R L., Fleming, R W., and Johnson, AM., 1993, Kinematics ofthe Aspen Grove
Landslide, Ephraim Canyon, Central Utah, Landslide Processes in Utah-
Observation and Theory, p. 34.
Bickner, F., 1985, Relative dating and geologic interpretation ofthirteen fluvial terraces
of the Eel River, northern California: Geological Society of America Abstracts
with Programs, v. 17, no. 6, p. 342.
Booth, AM., Roering, 1. J., and Perron, J. T., 2009, Automated landslide mapping using
spectral analysis and high-resolution topographic data: Puget Sound lowlands,
Washington, and Portland Hills, Oregon: Geomorphology, v. 109, no. 3-4, p. 132-
147.
Bovis, M. J., and Jones, P., 1992, Holocene History of Earthflow Mass Movements in
South-Central British-Columbia - the Influence of Hydroclimatic Changes:
Canadian Journal of Earth Sciences, v. 29, no. 8, p. 1746-1755.
Brown, W. M., and Ritter, J. R, 1971, Sediment Transport and Tubidity in the Eel River
Basin, California: Geological Survey Water-Supply Paper, v. 1986, p. 67.
Bruckl, E., and Sehidegger, A E., 1973, Application of Theory of Plasticity to Slow Mud
Flows: Geotechnique, v. 23, no. 1, p. 101-107.
Buis, A, 2008, Larger Pacific Climate Event Helps Current La Nina Linger: NASA JPL,
2008-066.
Burbank, D. W., Leland, J., Fielding, E., Anderson, R S., Brozovic, N., Reid, M. R, and
Duncan, c., 1996, Bedrock incision, rock uplift and threshold hillslopes in the
northwestern Himalayas: Nature, v. 379, no. 6565, p. 505-510.
California Department of Water Resources, 1965, Bulletin No. 136 - North Coastal Area
Investigation: Department of Water Resources, 136.
129
-, 1970, Middle Fork Eel River landslide investigation: California Department of Water
Resources.
Carson, M. A, and Petley, D. T., 1970, The Existence of Threshold Slopes in the
Denudation of the Landscape: Transactions of the Institute of British
Geographers, v. 49, p. 77-95.
Carter, W. E., Shrestha, R. L., and Slatton, K. C., 2007, Geodetic laser scanning: Physics
Today, v. 60, no. 12, p. 41-47.
Chandler, J. H., and Brunsden, D., 1995, Steady-State Behavior of the Black-Ven
Mudslide - the Application of Archival Analytical Photogrammetry to Studies of
Landform Change: Earth Surface Processes and Landforms, v. 20, no. 3, p. 255-
275.
Coe, J. A, Ellis, W. L., Godt, J. W., Savage, W. Z., Savage, J. E., Michael, J. A, Kibler,
J. D., Powers, P. S., Lidke, D. J., and Debray, S., 2003, Seasonal movement of the
Slumgullion landslide determined from Global Positioning System surveys and
field instrumentation, July 1998-March 2002: Engineering Geology, v. 68, no. 1-
2, p. 67-101.
Cruden, D. M., and Varnes, D. J., 1996, Landslide Types and Processes, in Turner, A K.,
and Schuster, R. L., eds., Landslides: Investigation and Mitigation: Washington,
DC, National Academy Press, p. 36-71.
Densmore, A L., Anderson, R. S., McAdoo, B. G., and Ellis, M. A, 1997, Hillslope
evolution by bedrock landslides: Science, v. 275, no. 5298, p. 369-372.
Densmore, A L., Ellis, M. A, and Anderson, R. S., 1998, Landsliding and the evolution
of normal-fault-bounded mountains: Journal of Geophysical Research - Solid
Earth,v. 103,no.B7,p. 15203-15219.
Densmore, A L., and Hovius, N., 2000, Topographic fingerprints of bedrock landslides:
Geology, v. 28, no. 4, p. 371-374.
Dietrich, W. E., Bellugi, D. G., Sklar, L. S., Stock, J. D., Heimsath, A M., and Roering,
J. J., 2003, Geomorphic Transport Laws for Predicting Landscape Form and
Dynamics, in Wilcock, P., and Iverson, R. M., eds., Prediction in
Geomorphology: Washington DC, American Geophysical Union, p. 103-132.
Dietrich, W. E., and Dunne, T., 1978, Sediment budget for a small catchment in
mountainous terrain: Zeitschrift Fur Geomorphologie, v. 29, p. 191-206.
130
Dwyer, M. J., Scott, R G., and Lorens, P. J., 1971, Landslide Conditions and Related
Sediment Production on the Eel River: California Department of Water
Resources.
Fuller, T. K, Perg, L. A, Willenbring, J. K, and Lepper, K, 2009, Field evidence for
climate-driven changes in sediment supply leading to strath terrace formation:
Geology, v. 37.
Furlong, K P., and Govers, R, 1999, Ephemeral crustal thickening at a triple junction:
The Mendocino crustal conveyor: Geology, v. 27, no. 2, p. 127-130.
Furlong, K P., and Schwartz, S. Y., 2004, Influence ofthe Mendocino triple junction on
the tectonics of coastal California: Annual Review of Earth and Planetary
Sciences, v. 32, p. 403-433.
Geertsema, M., Clague, J. J., Schwab, J. W., and Evans, S. G., 2006, An overview of
recent large catastrophic landslides in northern British Columbia, Canada:
Engineering Geology, v. 83, no. 1-3, p. 120-143.
Gendaszek, A, Thompson, N., Balco, G., Todd, C., and Stone, J., 2005, LONG-TERM
EROSION RATES OF COASTAL WATERSHEDS IN THE PACIFIC
NORTHWEST: Geological Society of America Abstracts with Programs, v. 37,
no. 7, p. 200.
Glastonbury, J., and Fell, R, 2008, Geotechnical characteristics of large slow, very slow,
and extremely slow landslides: Canadian Geotechnical Journal, v. 45, no. 7, p.
984-1005.
Glenn, N. F., Streutker, D. R, Chadwick, D. J., Thackray, G. D., and Dorsch, S. J., 2006,
Analysis of LiDAR-derived topographic information for characterizing and
differentiating landslide morphology and activity: Geomorphology, v. 73, no. 1-2,
p. 131-148.
Gonzalez-Diez, A, Remondo, J., de Teran, J. R D., and Cendrero, A, 1999, A
methodological approach for the analysis of the temporal occurrence and
triggering factors of landslides: Geomorphology, v. 30, no. 1-2, p. 95-113.
Guzzetti, F., Malamud, B. D., Turcotte, D. L., and Reichenbach, P., 2002, Power-law
correlations of landslide areas in central Italy: Earth and Planetary Science
Letters, v. 195, no. 3-4, p. 169-183.
Harden, D. R, Colman, S. M., and Nolan, K M., Mass Movement in the Redwood Creek
Basin, Northwestern California: U.S. Geological Survey Profesisonal Paper, v.
1454-G, p. GI-Gll.
131
Hilley, G. E., Burgmann, R, Ferretti, A, Novali, F., and Rocca, E, 2004, Dynamics of
slow-moving landslides from permanent scatterer analysis: Science, v. 304, no.
5679, p. 1952-1955.
Hovius, N., Stark, C. P., and Allen, P. A, 1997, Sediment flux from a mountain belt
derived by landslide mapping: Geology, v. 25, no. 3, p. 231-234.
Hovius, N., Stark, C. P., Tutton, M. A, and Abbott, L. D., 1998, Landslide-driven
drainage network evolution in a pre-steady-state mountain belt: Finisterre
Mountains, Papua New Guinea: Geology, v. 26, no. 12, p. 1071-1074.
Hungr, 0., Evans, S. G., Bovis, M. J., and Hutchinson, J. N., 2001, A review ofthe
classification of landslides of the flow type: Environmental & Engineering
Geoscience, v. 7, no. 3, p. 221-238.
Hutchinson, J. N., 1967, The Free Degredation of London Clay Cliffs, in Proc.
Geotechnical Conf., Oslo, p. 113-118.
Hutchinson, J. N., and Bhandari, R K, 1971, Undrained Loading, a Fundamental
Mechanism of Mudflows and Other Mass Movements: Geotechnique, v. 21, no. 4,
p.353-&.
Iverson, N. R, 1986a, Dynamics of slow landslides: a theory for time-dependent
behavior, in Abrahams, A D., ed., Hillslope Processes: Boston, Mass, Allen and
Unwin, p. 297-317.
Iverson, R M., 1986b, Unsteady, Nonuniform Landslide Motion.1. Theoretical
Dynamics and the Steady Datum State: Journal of Geology, v. 94, no. 1, p. 1-15.
-, 1986c, Unsteady, Nonuniform Landslide Motion .2. Linearized Theory and the
Kinematics of Transient-Response: Journal of Geology, v. 94, no. 3, p. 349-364.
-, 2005, Regulation of landslide motion by dilatancy and pore pressure feedback: Journal
of Geophysical Research - Earth Surface, v. 110, no. F2.
Iverson, R M., and Major, J. 1., 1987, Rainfall, Groundwater-Flow, and Seasonal
Movement at Minor Creek Landslide, Northwestern California - Physical
Interpretation of Empirical Relations: Geological Society of America Bulletin, v.
99, no. 4, p. 579-594.
Jayko, A S., Blake, M. c., McLaughlin, R J., Ohlin, H. N., Ellen, S. D., and Kelsey, H.
M., 1989, Reconnaissance Geologic Map ofthe Covelo 30- by 60- Minute
Quadrangle, Northern California: US Geological Survey, scale 1:100,000.
Keefer, D. K, and Johnson, AM., 1983, Earth Flows: Morphology, Mobilization and
Movement: Geological Survey Professional Paper, v. 1256, p. 1-56.
132
Kelsey, H. M., 1977, Landsliding, channel changes, sediment yield and land use in the
Van Duzen River basin, north coastal California, 1941 - 1975 [Ph.D. thesis]:
University of California, Santa Cruz, 370 p.
-, 1978, Earthflows in Franciscan Melange, Van Duzen River Basin, California: Geology,
v. 6, no. 6, p. 361-364.
-,1980, A Sediment Budget and an Analysis of Geomorphic Process in the Van-Duzen
River Basin, North Coastal California, 1941-1975 - Summary: Geological Society
of America Bulletin, v. 91, no. 4, p. 190-195.
Kelsey, H. M., and Carver, G. A, 1988, Late Neogene and Quaternary Tectonics
Associated with Northward Growth of the San-Andreas Transform-Fault,
Northern California: Journal of Geophysical Research-Solid Earth and Planets, v.
93, no. B5, p. 4797-4819.
Korup, 0.,2005, Large landslides and their effect on sediment flux in South Westland,
New Zealand: Earth Surface Processes and Landforms, v. 30, no. 3, p. 305-323.
-,2006, Rock-slope failure and the river long profile: Geology, v. 34, no. 1, p. 45-48.
Korup, 0., Clague, J. l, Hermanns, R L., Hewitt, K., Strom, A L., and Weidinger, J. T.,
2007, Giant landslides, topography, and erosion: Earth and Planetary Science
Letters, v. 261, no. 3-4, p. 578-589.
Lehre, A K., and Carver, G. A, 1985, Thrust faulting and earthflows: Speculations on
the sediment budget of a tectonically active drainage basin, in Savina, M. E., ed.,
American Geomorphological Field Group Field Trip Guidebook 1985
Conference,p.169-183.
Leprince, S., Berthier, E., Ayoub, E, Delacourt, c., and Avouac, J. P., 2008, Monitoring
Earth Surface Dynamics With Optical Imagery: EOS, Transactions, American
Geophysical Union, v. 89, no. 1, p. 1-2.
Lins, H. E, and Slack, l R, 1999, Streamflow trends in the United States: Geophysical
Research Letters, v. 26, no. 2, p. 227-230.
Lock, J., Kelsey, H., Furlong, K, and Woolace, A, 2006, Late neogene and quaternary
landscape evolution of the northern California Coast Ranges: Evidence for
Mendocino triple junction tectonics: Geological Society of America Bulletin, v.
118, no. 9-10, p. 1232-1246.
Luce, C. H., and Holden, Z. A, 2009, Declining annual streamflow distributions in the
Pacific Northwest United States, 1948-2006: Geophysical Research Letters, v. 36.
133
Mackey, B. H., Roering, J. J., and McKean, J. A, 2009, Long-term kinematics and
sediment flux of an active earthflow, Eel River, California: Geology, v. 37, no. 9,
p.803-806.
Malamud, B. D., Turcotte, D. L., Guzzetti, F., and Reichenbach, P., 2004, Landslide
inventories and their statistical properties: Earth Surface Processes and
Landforms, v. 29, no. 6, p. 687-711.
Malet, J. P., Maquaire, 0., and Calais, E., 2002, The use of Global Positioning System
techniques for the continuous monitoring of landslides: application to the Super-
Sauze earthflow (Alpes-de-Haute-Provence, France): Geomorphology, v. 43, no.
1-2, p. 33-54.
Malhase, J., 1938, Landslide at Mile Post 201 Telegraph Poles 9 to 12, Northwestern
Pacific Railroad.
Mantua, N. J., and Hare, S. R, 2002, The Pacific decadal oscillation: Journal of
Oceanography, v. 58, no. 1, p. 35-44.
Mantua, N. J., Hare, S. R, Zhang, Y, Wallace, J. M., and Francis, R c., 1997, A Pacific
interdecadal climate oscillation with impacts on salmon production: Bulletin of
the American Meteorological Society, v. 78, no. 6, p. 1069-1079.
Maquaire, 0., Malet, J. P., Remaitre, A, Locat, J., Klotz, S., and Guillon, J., 2003,
Instability conditions of marly hillslopes: towards landsliding or gullying? The
case of the Barcelonnette Basin, South East France: Engineering Geology, v. 70,
no. 1-2, p. 109-130.
McKean, J., and Roering, J., 2004, Objective landslide detection and surface morphology
mapping using high-resolution airborne laser altimetry: Geomorphology, v. 57,
no. 3-4, p. 331-351.
McLaughlin, R. J., Ellen, S. D., Blake, M. C. J., Jayko, A S., Irwin, W. P., Aalto, K. R,
Carver, G. A, and Clark, S. H. J., 2000, Geology ofthe Cape Mendocino, Eureka,
Garberville, and Southwestern part of the Hayfork 30 x 60 Minute Quadrangles
and Adjacent Offshore Area, Northern California: U.S. Geological Survey.
McLaughlin, R. J., Kling, S. A, Poore, R. Z., McDougall, K., and Beutner, E. c., 1982,
Post-Middle Miocene Accretion of Franciscan Rocks, Northwestern California:
Geological Society of America Bulletin, v. 93, no. 7, p. 595-605.
McSaveney, M. J., and Griffiths, G. A, 1987, Drought, Rain, and Movement of a
Recurrent Earthflow Complex in New-Zealand: Geology, v. 15, no. 7, p. 643-646.
134
Merritts, D., and Bull, W. B., 1989, Interpreting Quaternary Uplift Rates at the
Mendocino Triple Junction, Northern California, from Uplifted Marine Terraces:
Geology, v. 17, no. 11, p. 1020-1024.
Merritts, D. J., 1996, The Mendocino triple junction: Active faults, episodic coastal
emergence, and rapid uplift: Journal of Geophysical Research-Solid Earth, v. 101,
no. B3, p. 6051-6070.
Muhs, D. R., Thorson, R. M., Clague, J. J., Mathews, W. H., McDowell, P. F., and
Kelsey, H. M., 1987, Pacific Coast and Mountain System, in Graf, W. L., ed.,
Geomorphic systems of North America: Boulder, Geological Society of America,
p.517-581.
Nittrouer, C. A., 1999, STRATAFORM: overview of its design and synthesis of its
results: Marine Geology, v. 154, no. 1-4, p. 3-12.
Nolan, K. M., and Janda, R. J., 1995, Movement and Sediment Yield of Two Earthflows,
Northwestern California, in Survey, U. S. G., ed., p. FI-FI2.
Ohlmacher, G. c., 2007, Plan curvature and landslide probability in regions dominated
by earth flows and earth slides: Engineering Geology, v. 91, no. 2-4, p. 117-134.
Ouimet, W. B., Whipple, K. X., and Granger, D. E., 2009, Beyond threshold hillslopes:
Channel adjustment to base-level fall in tectonically active mountain ranges:
Geology, v. 37, no. 7, p. 579-582.
Perron, J. T., Dietrich, W. E., and Kirchner, J. W., 2008, Controls on the spacing of first-
order valleys: Journal of Geophysical Research-Earth Surface, v. 113, no. F4.
Pettinga, J. R., 1987, Waipoapoa Landslide - a Deep-Seated Complex Block Slide in
Tertiary Weak-Rock Flysch, Southern Hawkes Bay, New-Zealand: New Zealand
Journal of Geology and Geophysics, v. 30, no. 4, p. 401-414.
Roering, J. J., Stimely, L. L., Mackey, B. H., and Schmidt, D. A., 2009, Using DInSAR,
airborne LiDAR, and archival air photos to quantify landsliding and sediment
transport: Geophysical Research Letters, v. 36.
Savage, W., and Wasowski, J., 2006, A plastic flow model for the Acquara-Vadoncello
landslide in Senerchia, Southern Italy: Engineering Geology, v. 83, no. 1-3, p. 4-
21.
Schmidt, K. M., and Montgomery, D. R., 1995, Limits to Relief: Science, v. 270, no.
5236, p. 617-620.
Schulz, W. H., 2007, Landslide susceptibility revealed by LIDAR imagery and historical
records, Seattle, Washington: Engineering Geology, v. 89, no. 1-2, p. 67-87.
135
Schulz, W. H., McKenna, J. P., Kibler, J. D., and Biavati, G., 2009, Relations between
hydrology and velocity of a continuously moving landslide-evidence of pore-
pressure feedback regulating landslide motion?: Landslides, v. 6, no. 3, p. 181-
190.
Scott, R G., 1973, Geology and sediment production for ten Eel River landslides:
California Department of Water Resources - Northern District.
Skempton, A W., 1964, Long-term stability of clay slopes: Geotechnique, v. 14, p. 77-
102.
Slama, C. c., 1980, Manual of Photogrammetry: Falls Church, VA, American Society of
Photogrammetry, p. 1056.
Slatton, K c., Carter, W. E., Shrestha, R L., and Dietrich, W., 2007, Airborne Laser
Swath Mapping: Achieving the resolution and accuracy required for geosurficial
research: Geophysical Research Letters, v. 34, no. 23.
Sloan, J., Miller, J. R, and Lancaster, N., 2001, Response and recovery of the Eel River,
California, and its tributaries to floods in 1955, 1964, and 1997: Geomorphology,
v. 36, no. 3-4,p. 129-154.
Smith, B. J., Tilton, W. c., Elko, J. M., and Jacobus, C. D., 1974, Eel River Basin
Environmental Studies: California Department of Water Resources
Sommerfield, C. K, Drake, D. E., and Wheatcroft, R A, 2002, Shelf record of climatic
changes in flood magnitude and frequency, north-coastal California: Geology, v.
30, no. 5, p. 395-398.
Sommerfield, C. K, and Nittrouer, C. A, 1999, Modern accumulation rates and a
sediment budget for the Eel shelf: a flood-dominated depositional environment:
Marine Geology, v. 154, no. 1-4, p. 227-241.
Sommerfield, C. K, and Wheatcroft, R A, 2007, Late Holocene sediment accumulation
on the northern California shelf: Oceanic, fluvial, and anthropogenic influences:
Geological Society of America Bulletin, v. 119, no. 9-10, p. 1120-1134.
Swanson, F. 1., and Swanston, D. N., 1977, Complex mass-movement terrains in the
Western Cascade Range, Oregon: Geological Society of America Reviews in
Engineering Geology, v. 3, p. 113-124.
Syvitski, J. P., and Morehead, M. D., 1999, Estimating river-sediment discharge to the
ocean: application to the Eel margin, northern California: Marine Geology, v.
154, no. 1-4, p. 13-28.
136
van Asch, T. W. J., Van Beek, L. P. H., and Bogaard, T. A, 2007, Problems in predicting
the mobility of slow-moving landslides: Engineering Geology, v. 91, no. 1, p. 46-
55.
Van de Grift, J. W., and Sack, D., 2004, Monitoring slump-earthflow complex
movement: A southeastern Ohio case study: Ohio Journal of Science, v. 104, no.
5, p. 126-135.
VanDen Eeckhaut, M., Poesen, J., Verstraeten, G., Vanacker, V., Moeyersons, J.,
Nyssen, J., and van Beek, L. P. H., 2005, The effectiveness of hillshade maps and
expert knowledge in mapping old deep-seated landslides: Geomorphology, v. 67,
no. 3-4,p. 351-363.
Vulliet, L., and Hutter, K., 1988, Viscous-Type Sliding Laws for Landslides: Canadian
Geotechnical Journal, v. 25, no. 3, p. 467-477.
Wahrhaftig, c., and Curry, R R, 1967, Geologic implications of sediment discharge
recorded from Northern Coast Ranges, California, in Man's effect on California
watersheds: Report ofthe California Assembly Committee on Natural Resources,
Planning, and Public Works, Subcommittee on Forestry and Watershed
Management, Part III, p. 35-58
Wheatcroft, R A, and Sommerfield, C. K., 2005, River sediment flux and shelf sediment
accumulation rates on the Pacific Northwest margin: Continental Shelf Research,
v. 25, no. 3, p. 311-332.
Wieczorek, G. E, 1984, Preparing a detailed landslide-inventory map for hazard
evaluation and reduction: Bulletin of the Association of Engineering Geologists,
v. 21, no. 3, p. 337-342.
Wills, C. J., and McCrink, T. P., 2002, Comparing landslide inventories: The map
depends on the method: Environmental & Engineering Geoscience, v. 8, no. 4, p.
279-293.
Wolf, P. R, and Dewitt, B. A, 2000, Elements of Photogrammetry. With Applications in
GIS: Boston, MA, McGraw-Hill.
Zhang, X. B., Phillips, c., and Pearce, A, 1991, Surface Movement in an Earthflow
Complex, Raukumara Peninsula, New-Zealand: Geomorphology, v. 4, no. 3-4, p.
261-272.