Content uploaded by M. Parise
Author content
All content in this area was uploaded by M. Parise on Feb 07, 2019
Content may be subject to copyright.
Ž.
Geomorphology 39 2001 171–188 www.elsevier.nlrlocatergeomorph
Wildfire-related debris-flow initiation processes,
Storm King Mountain, Colorado
Susan H. Cannon a,), Robert M. Kirkhamb,1, Mario Parise c,2
aGeologic Hazards Team, U.S. Geological SurÕey, Box 25046, DenÕer Federal Center, MS 966, DenÕer, CO 80225, USA
bColorado Geological SurÕey, P.O. Box 172, Monte Vista, CO, USA
cNational Research Council of Italy— CERIST, cro Istituto di Geologia Applicata e Geotecnica, Via Orabona, 4, 70125 Bari, Italy
Received 15 March 2000; received in revised form 21 November 2000; accepted 25 November 2000
Abstract
A torrential rainstorm on September 1, 1994 at the recently burned hillslopes of Storm King Mountain, CO, resulted in
Ž.
the generation of debris flows from every burned drainage basin. Maps 1:5000 scale of bedrock and surficial materials and
Ž.
of the debris-flow paths, coupled with a 10-m Digital Elevation Model DEM of topography, are used to evaluate the
processes that generated fire-related debris flows in this setting. These evaluations form the basis for a descriptive model for
fire-related debris-flow initiation.
The prominent paths left by the debris flows originated in 0- and 1st-order hollows or channels. Discrete soil-slip scars do
not occur at the heads of these paths. Although 58 soil-slip scars were mapped on hillslopes in the burned basins, material
derived from these soil slips accounted for only about 7% of the total volume of material deposited at canyon mouths. This
fact, combined with observations of significant erosion of hillslope materials, suggests that a runoff-dominated process of
progressive sediment entrainment by surface runoff, rather than infiltration-triggered failure of discrete soil slips, was the
primary mechanism of debris-flow initiation. A paucity of channel incision, along with observations of extensive hillslope
erosion, indicates that a significant proportion of material in the debris flows was derived from the hillslopes, with a smaller
contribution from the channels.
Because of the importance of runoff-dominated rather than infiltration-dominated processes in the generation of these
fire-related debris flows, the runoff-contributing area that extends upslope from the point of debris-flow initiation to the
drainage divide, and its gradient, becomes a critical constraint in debris-flow initiation. Slope-area thresholds for fire-related
Ž.
3
debris-flow initiation from Storm King Mountain are defined by functions of the form Atan
u
sS, where Ais the
cr cr
critical area extending upslope from the initiation location to the drainage divide, and tan
u
is its gradient. The thresholds
vary with different materials. q2001 Elsevier Science B.V. All rights reserved.
Keywords: Wildfire; Erosion; Debris flow; Initiation
)Corresponding author. Tel.: q1-303-273-8600; fax: q1-303-273-8600.
Ž. Ž . Ž.
E-mail addresses: cannon@usgsg.gov S.H. Cannon , rmk@amigo.net R.M. Kirkham , cerimp06@area.ba.cnr.it M. Parise .
1Tel.: q1-719-587-0139; fax: q1-719-587-2187.
2Tel.: q39-80-5428137; fax: q39-80-5567944.
0169-555Xr01r$ - see front matter q2001 Elsevier Science B.V. All rights reserved.
Ž.
PII: S0169-555X 00 00108-2
()
S.H. Cannon et al.rGeomorphology 39 2001 171–188172
1. Introduction and approach
During the evening of September 1, 1994, debris
flows originating in response to a heavy rainstorm
occurred on Storm King Mountain west of Glen-
wood Springs, CO. These flows traveled down chan-
nels on the south flank of the mountain and emptied
onto or next to Interstate Highway 70 in several
locations. The flows originated in drainage basins
recently burned by the July 1994 South Canyon fire.
Every drainage basin burned by this fire, and even
some that were not burned, produced debris flows.
The primary focus of this study is to examine pro-
cesses that lead to the generation of fire-related
debris flows in this setting and to develop a descrip-
tive model for their initiation.
The process most commonly associated with the
generation of debris flows from unburned hillslopes
in the United States is that of failure of a landslide as
an intact block which then mobilizes into a fluid
debris flow with travel downslope. When this pro-
cess occurs, the debris-flow path can be traced up a
channel to a discrete landslide-scar source. Campbell
Ž.
1975 referred to this process as soil slip–debris
flow; we adopt this terminology here because it
reflects both the transition from landslide to debris
flow and the observation that the landslide involved
primarily the surficial soil. Considerable research
effort has gone into the development of hypotheses
to explain both the failure and mobilization of soil
slips into debris flows, many of which are discussed
Ž.
in Iverson et al. 1997 .
Two initiation processes specifically for fire-re-
lated debris flows have been identified in the litera-
ture: infiltration-triggered soil slip, as described
above, and runoff-dominated erosion by surface
runoff. These two processes are reported in widely
disparate environments. The process of soil slip-de-
bris flow has been documented in burned areas in
Ž. Ž.
southern California by Wells 1987 , Morton 1989 ,
Ž. Ž.
Booker 1998 , and Cannon 1999 ; the occurrence
of soil slips on the hillslopes points to failure trig-
gered by rainfall infiltration. Debris-flow generation
by failure of a discrete landslide in burned areas has
also been attributed to reduced evapotranspiration
rates and the consequent increase in soil moisture
ŽKlock and Helvey, 1976; Helvey, 1980; Swanson,
.
1981; Megahan, 1983 and decay of roots that an-
Ž.
chor colluvium e.g., Swanson, 1981; DeGraff, 1997 .
These processes are generally thought to occur a few
years after the fire.
An alternative process for debris-flow initiation
based on significantly decreased rainfall infiltration
rates has also been proposed for burned areas. John-
Ž.
son 1984 , working in Big Sur, Monterey County,
Ž.
CA, and Wells 1987 , working in the San Gabriel
Mountains of southern California, traced debris-flow
deposits directly upslope through small gullies and
into a series of rills. These workers concluded that
the debris flows initiated high on the hillslopes from
material eroded by surface runoff, and that the debris
flows increased in volume by entraining larger mate-
Ž.
rial from the channels. In Wells’ model 1987 ,
debris flows initiate by failure of a saturated layer of
soil a few mm thick above a subsurface water-repel-
lent zone as miniature soil slips. Material from these
tiny soil-slips forms rills and travels downslope as
shallow, narrow debris flows.
Ž.
More recently, Meyer and Wells 1997 , working
in Yellowstone National Park, observed the first
appearance of debris-flow features such as levees
and mud coatings in the middle reaches of the main
basins. These workers concluded that debris flows
would thus initiate through a process of progressive
sediment bulking of surface runoff and rill erosion in
steep upper basin slopes, followed by deep incision
Ž.
as flows progressed down channels. Parrett 1987
also noted the lack of landslide scars in a burned
area that experienced debris flows in Montana and
suggested a similar mechanism. Meyer and Wells
Ž. Ž.
1997 and Parrett 1987 emphasize that both hills-
lope sediment input from rills and gullies, as well as
material entrained by extensive channel incision are
important in the bulking process that led to the
Ž.
formation of debris flows. Meyer and Wells 1997
further hypothesized that addition of fine-grained
sediment eroded from hillslopes to the generally
coarser-grained channel material was important in
both the development of debris-flow conditions and
in maintaining the mobility of the flow.
On Storm King Mountain, we observed evidence
of both runoff- and infiltration-dominated initiation
processes. In this paper, we examine the relative
contributions of materials generated by both pro-
cesses to the debris-flow deposits, and the effects of
topographic configuration and lithology on fire-
()
S.H. Cannon et al.rGeomorphology 39 2001 171–188 173
related debris-flow initiation. We then explore the
idea that the runoff-contributing area extending ups-
lope from the initiation location to the drainage
divide, and its gradient, constrained debris-flow initi-
ation during the September events. These evalua-
tions, coupled with field observations and measure-
ments, are used to build a descriptive model for the
generation of fire-related debris flows from Storm
King Mountain.
The analyses are based on digital compilations of
1:5000-scale geologic mapping of the south flank of
Storm King Mountain and of the debris-flow paths
from the September 1, 1994 events. We used a 10-m
Ž.
resolution Digital Elevation Model DEM and Digi-
Ž.
tal Line Graph DLG with 20-ft contours obtained
from survey-controlled, 1:8000-scale aerial pho-
tographs taken on November 10, 1994 for topo-
graphic information.
2. Study area
The study area lies on the south flank of Storm
King Mountain, north of the Colorado River and
about 5–10 km west-northwest of downtown Glen-
Ž.
wood Springs Fig. 1
2.1. The South Canyon fire, July 1994
On July 2, 1994, lightning from a summer storm
started a fire on the south flank of Storm King
Mountain. For the next few days the fire slowly
spread across the mountain, and firefighters were
brought in to protect homes in the Canyon Creek
area west of the fire. On July 6, a strong cold front
accompanied by a sudden shift in the direction and
intensity of the wind passed through the area and
caused the fire to spread rapidly. The fire quickly
grew from around 1 to about 7 km2, threatening
West Glenwood Springs and resulting in the evacua-
tion of parts of the city. Tragically, 14 firefighters
lost their lives while battling this fire.
2.2. Geomorphic and geologic setting
The study area consists of seven major intermit-
tent stream drainages with basin areas from 0.31 to
2.46 km2, labeled A through G in Fig. 1. All seven
drainage basins are direct tributaries of the Colorado
River and have steep stream channels and precipi-
Ž.
tous side slopes Table 1 . This topographic configu-
ration is conducive to a rapid concentration of runoff,
and when combined with intense rains, can lead to
high peak discharge and erosion rates, even without
the exacerbating effects of wildfire. Drainage basin
A experienced only 3% burn; basins E, F, and G
experienced between 48% and 75% burn; and basins
Ž
B, C, and D were nearly completely burned Table
.
1.
Both the burned and unburned hillslopes of the
Storm King Mountain watershed are characterized
by an average gradient of 158, with some hillsides,
particularly in the southern portion, having slopes
greater than 358. Before the fire, southeast-facing
hillslopes supported a sparse pinyon–juniper vegeta-
tive community. The northern two-thirds to one-half
of the burned area supported a nearly impenetrable
thicket of oak brush. Soils in the lower one-third to
one-half of the burned area are generally very shal-
low, poorly developed, and contain a high percent-
age of gravel- and larger-sized material. Using the
Ž.
Unified Soil Classification System Craig, 1987 ,
Ž.
Cannon et al. 1995 classified two samples of un-
burned soil as silty fine sand and silty sand and one
sample of burned soil as silty sand. The climate is
semiarid; and the majority of precipitation occurs in
July, August, and September as convectional thun-
derstorms.
Erosion following wildfires is often attributed to
the development, or enhancement, of a water-repel-
Ž
lent layer in the soil by the fire e.g., DeBano and
Letey, 1969; DeBano, 1981; Morris and Moses, 1987;
.
Wells, 1987; Robichaud, 1996 . The presence and
extent of water-repellent soils at Storm King Moun-
tain was assessed by digging small pits with clean,
inclined sides in areas where burned soil and ash
were in place. Pits were, in general, 10 cm deep, 20
cm wide, and with one side inclined at about 3:1.
Water from a squirt bottle was dripped along the
incline. Water repellency was identified if the water
beaded on the surface and did not infiltrate for at
least 30 s. Where water-repellent material was found,
its lateral extent was evaluated by dripping more
water along the 20-cm wide inclined surface.
Water-repellent soils were detected in approxi-
mately 35% of the test pits in the months following
()
S.H. Cannon et al.rGeomorphology 39 2001 171–188174
Fig. 1. Storm King Mountain study area. Drainage basins are labeled A through G and are delineated by thin solid lines. Small watershed
fronts along the Colorado River between the major drainage basins are outlined but not labeled. Heavier black solid lines show paths of the
debris flows that occurred during the September 1, 1994 event. Soil-slip scars are shown as solid black polygons. Heavy dashed line marks
the extent of the South Canyon fire of July 1994.
()
S.H. Cannon et al.rGeomorphology 39 2001 171–188 175
Table 1
Area, height, and relief ratio of drainage basins, burned area, and
percent of drainage basin burned
Drainage Area Relief Relief Burned Percent
22
Ž.Ž. Ž. Ž.
basin km m ratio area km burned %
A 2.03 842 0.32 0.06 3.0
B 2.23 950 0.27 2.08 93.5
C 2.46 941 0.30 2.27 92.3
D 0.77 772 0.38 0.73 95.3
E 0.46 449 0.33 0.29 62.6
F 2.11 853 0.30 1.02 48.2
G 0.31 526 0.32 0.23 75.3
Relief ratio is calculated as basin relief divided by length. Basin
length is measured from the drainage mouth along the length of
the longest channel extended to the drainage divide.
Ž.
the fire Cannon et al., 1995, 1998 . The water-repel-
lent soils that were observed were not extensive nor
laterally continuous. Field evidence of variations in
fire temperature, soil materials, and vegetation could
not explain their presence or absence. At one un-
burned site located under a juniper tree, a discontinu-
ous water-repellent soil was detected, probably
caused by hydrophobic compounds in the unburned
organic materials.
In the days following the fire, residents of Glen-
wood Springs reported seeing huge clouds of white
dust flying above the mountain. Presumably, the ash
and loose, friable, and exposed burned mineral soil
on the hillsides were being transported by wind and
redistributed on hillslopes, deposited in the tributary
drainages, and removed from the area. In addition,
the processes of dry ravel both during and after the
fire resulted in the downslope transport and accumu-
lation of material. Dry-ravel deposits are formed by
the particle-by-particle transport of material downs-
lope by gravity. Dry ravel has been described as an
important post-fire process in southern California
where channels are loaded with sediment, increasing
available sediment for transport in large runoff events
Ž.
e.g., Wells, 1987; Florsheim et al., 1991 . After the
fire on Storm King Mountain, accumulations of ash
and dry-ravel material up to about 1 m deep along
the sides of most tributary drainages were observed
Ž.
Fig. 2 . This material was at its angle of repose,
measured between 268and 328. The dry-ravel mate-
rial was primarily well-sorted, silty sand and lacked
Ž
the larger clasts present in the in-place soils Cannon
.
et al., 1995, 1998 . Aprons of this loose material
were also observed mantling many sideslopes. In
addition, larger material in the form of loose boul-
ders, cobbles, and channel alluvium had been de-
posited in the channels prior to the fire by either
gravity-driven colluvial processes or by fluvial and
Ž.
debris-flow processes Fig. 2 .
Field and aerial photographic geologic mapping
of the study area at a scale of 1:5000 was completed
Ž.Ž.
in 1995 Kirkham et al., 1999 Fig. 3 Permian and
Pennsylvanian red beds of the Maroon Formation
underlie most of the study area and evaporitic rocks
of the Pennsylvanian Eagle Valley Evaporite crop
out in the northern part of the mapped area. The
Fig. 2. Photograph of loose, noncohesive material transported by
dry ravel and wind that forms aprons along tributary channels and
larger material stored in channel prior to the September 1994
debris-flow event. Photograph was taken on August of 1994.
These deposits supplied material for debris-flow events in Septem-
ber of 1994. Photograph by Roger Pihl, Colorado Geological
Survey.
()
S.H. Cannon et al.rGeomorphology 39 2001 171–188176
Ž.
Fig. 3. Geologic map of the south flank of Storm King Mountain simplified from Kirkham et al., 1999 .
Eagle Valley Formation, which is mapped between
the Maroon Formation and Eagle Valley Evaporite,
is transitional between the red bed and evaporitic
formations and contains rock types found in both of
the formations. A late Tertiary conglomerate mantles
the upper south shoulder of Storm King Mountain.
The areal extent of these units within the study area
is shown in Table 2.
()
S.H. Cannon et al.rGeomorphology 39 2001 171–188 177
Table 2
Areal distribution of geologic units within the Storm King Moun-
tain study area
Geologic unit Area Percent of study
2
Ž. Ž.
km area %
Artificial fill 0.01 0.1
Eagle Valley Formation 0.93 7.7
Eagle Valley Evaporite 0.81 6.8
Maroon Formation 4.91 41.0
Conglomerate of 0.23 1.9
Storm King Mountain
Younger colluvium and 1.67 13.9
sheetwash deposits
Older colluvium and 0.95 7.9
sheetwash deposits
Recent debris-flow deposits 0.02 0.2
Younger debris-flow deposits 0.16 1.3
Older debris-flow deposits 0.10 0.1
Recent landslide deposits 0.20 0.2
Landslide deposits 0.13 1.1
Older landslide deposits 2.13 17.8
Terrace alluvium 0.01 0.1
Structurally, the study area lies astride part of the
Grand Hogback Monocline, and a west-northwest-
trending structural terrace within the monocline ex-
tends across the study area. Several folds and faults
are associated with the hinge zones on either side of
the structural terrace, which may be a collapse fea-
ture related to dissolution or flowage of underlying
Ž.
evaporitic rocks Bryant et al., 1998 . Within this
collapsed structural terrace, bedrock is intensely frac-
tured.
Bedrock is locally well exposed in bold outcrops,
notably on the steep slopes adjacent to the Colorado
River and on the southwest side of Storm King
Mountain. Within the structural terrace, however,
outcrops are scarce. Bedrock in this area, most no-
tably the Maroon Formation, weathers rapidly, prob-
ably due to the intense fracturing associated with the
collapsed structural terrace. Colluvium, sheetwash,
and landslide deposits, locally as much as 38 m
Ž.
thick, and deeply weathered bedrock residuum
mantle the bedrock in most of the study area. Prior to
the rainstorm on September 1, 1994, many of the
steep hillslopes underlain by the Maroon Formation
within the structural terrace had covers of residuum
up to about 1.2 m thick.
Surficial deposits other than Maroon Formation
Ž
residuum cover about 43% of the study area Table 2
.
and Fig. 3 . These units include the following
Ž.Ž.
Kirkham et al., 1999 : i Younger colluvium and
Ž
sheetwash deposits Holocene and late Pleistocene
.
age consisting primarily of poorly sorted, poorly to
moderately well bedded, matrix-supported gravelly
silty sand and sandy silt. These units cover approxi-
mately 14% of the study area, mantling hillslopes
Ž.
and infilling drainage channels. ii Older colluvium
Ž.
and sheetwash deposits Pleistocene age on hill-
slopes, ridge crests, and basin floors that are ero-
sional remnants of formerly more extensive deposits
Ž.
that once filled basins. iii Recent debris-flow de-
Ž.
posits latest Holocene associated with the Septem-
ber 1, 1994 storm are poorly sorted, matrix- and
clast-supported, and range between a silty sand, a
bouldery, cobbly, and pebbly gravel, and a sandy
Ž. Ž
silt. iv Younger debris-flow deposits Holocene
.
age , deposited prior to the September 1, 1994 storm
occur in the channels of the seven major drainage
Ž.
basins and on fans at the mouths of some. v Older
Ž
debris-flow deposits early Holocene and late Pleis-
.
tocene age include remnants of formerly more ex-
tensive deposits that lie up to about 15 m above
Ž.
modern channels. vi Recent landslide deposits
Ž.
latest Holocene age include active and recently
active landslides having morphological features that
suggest movement during the previous few years.
Ž. Ž .
vii Landslide deposits Holocene age exhibit dis-
tinctive landslide morphology but do not appear to
Ž.
have moved during the last few decades. vii The
Ž.
older landslide deposits Pleistocene age include a
very large landslide complex that heads on the south
side of Storm King Mountain in the upper reaches of
basins C, D, E, F, and G and covers 17.8% of the
study area. This old landslide complex appears to
have been stable for thousands of years and is locally
dissected by stream channels up to a depth of a few
Ž.
tens of meters. ix Minor amounts of artificial fill
and Quaternary terrace deposits also occur in the
study area.
3. September 1–2, 1994 debris-flow event
On September 1, 1994 at approximately 10:30
p.m., in response to a torrential downpour, debris
flows originated on the burned hillslopes on Storm
Ž
King Mountain. These flows consisting of mud,
()
S.H. Cannon et al.rGeomorphology 39 2001 171–188178
.
rocks and burned vegetation emptied down onto or
Ž
next to Interstate Highway 70 from 15 channels Fig.
.
1 . Thirty cars traveling on the highway at the time
of the debris flows were engulfed or trapped by the
mud. At least two of the people travelling in these
vehicles were swept into the river by the debris
flows. Although some travelers were seriously in-
jured, fortunately no deaths resulted from this event.
Unfortunately, the only available rainfall records
are daily totals recorded at a site 5–10 km southeast
of Storm King Mountain in downtown Glenwood
Springs. This record does not reflect an unusual
event. However, a motorist traveling on the highway
at the time of the storm described the rain as A...so
hard you almost couldn’t see.B
According to Colorado Department of Transporta-
tion personnel, burned logs and branches and up to
boulder-sized material in a very fluid, muddy matrix
continued to flow out of the canyons as a series of
pulses throughout the night of September 1 and the
early morning hours of September 2. Material was
deposited at the mouth of all of the burned basins. A
total area of approximately 0.13 km2was inundated
with approximately 68,000 m3of debris-flow de-
Ž
posits generated from the burned areas Cannon et
.
al., 1995, 1998 . Material deposited at canyon mouths
was generally flat-lying, indicating low shear strength
Ž.
perhaps due to high water content , and was com-
prised primarily of silty sand with extremely variable
Ž.
amounts of gravel, cobbles, and boulders Fig. 4
The passage of the debris flows down channels
was marked by a distinct path on the channel side-
walls consisting of a muddy veneer up to 2.5 cm
Ž.
thick Fig. 5 Material was also deposited locally in
the channels. This material consisted of cobble- and
boulder-sized material in an abundant silty sand ma-
Ž.Ž.
trix Cannon et al., 1995, 1998 Fig. 6 These
deposits were in the form of levees and lobes, indi-
cating a more significant shear strength than those
deposited at the canyon mouths. The boulders and
cobbles in the deposits from the September 1994
event probably came from material that had been
deposited in the channels prior to the fire by either
gravity-driven colluvial processes or by fluvial and
Ž.
debris-flow processes Cannon et al., 1995, 1998 .
A particular focus of the field mapping was to
identify the upper extent of flow paths where surface
Ž
runoff could be characterized as debris flow Kirk-
.
ham et al., 1999 . These points, which we refer to as
initiation locations, are where debris-flow features
consisting of nearly continuous levees comprised of
primarily matrix-supported material and a mud ve-
Fig. 4. Photograph of very fluid debris-flow deposits at mouth of canyon. Deposits consisted of abundant silty sand with variable amounts of
gravel, cobbles, and boulders.
()
S.H. Cannon et al.rGeomorphology 39 2001 171–188 179
Fig. 5. Photograph of debris-flow path flow through high-order channel. Note thin, muddy veneer deposited by debris flow on channel
sidewalls. Photograph by Roger Pihl, Colorado Geological Survey.
neer lining the path persist down channel. We identi-
fied 84 debris-flow initiation locations, shown as the
upper ends of the debris-flow paths in Fig. 1. All of
the initiation locations occurred in pre-existing
Ž.
0- and 1st-order channels Fig. 7 . Nearly complete
removal of burned soil and ash from the drainages
immediately upslope of the initiation locations indi-
cated the occurrence of concentrated overland flow.
Fig. 6. Photograph of debris-flow deposits in channel axis. Deposits consisted of cobble-to boulder-sized material in a silty sand matrix.
()
S.H. Cannon et al.rGeomorphology 39 2001 171–188180
Fig. 7. Photograph of extensively eroded hillslopes in basin C showing debris-flow paths and initiation locations in 0- and 1st-order
drainages. Before the September rainstorm, the hillslopes were mantled with dark-colored burned soil and ash.
Abrupt headcuts, like those described for channel
Ž.
initiation by Montgomery and Dietrich 1994 , or
small headcuts, like those attributed to very small-
scale landslides or Coulomb failure at the onset of
Ž. Ž.
rill initiation by Johnson 1984 and Wells 1987 ,
were not observed at the initiation locations.
With the exception of some localized lateral
stream erosion and bank caving and flushing of the
dry-ravel material and loose material stored in the
stream channels, there did not appear to be signifi-
cant amounts of channel incision during the Septem-
ber 1994 event. The channels did not exhibit exten-
sive areas of freshly exposed, continuous, steep to
near vertical walls with dangling tree roots that
usually indicate incision. Rather, even at steep gradi-
ents, the channels were generally V- or U-shaped
and coated with a thin veneer of debris-flow de-
posits, as seen in Fig. 5. Bedrock was exposed in
places in the channel, but the lack of incised channel
banks indicated that there likely was not more than
about 0.5 m of material mantling the surface before
the passage of the debris flows. In addition, U.S.
Bureau of Land Management personnel familiar with
the canyons reported that bedrock was exposed in
places in the main channels prior to the September
events, and they also saw little evidence for exten-
sive channel erosion or incision following the
Ž.
September event. Cannon and Reneau 2000 also
observed fire-related debris-flow deposits in a chan-
nel that did not exhibit extensive incision.
Field estimates made following the September
events suggested that approximately 15% of the sur-
face of the mineral soil in the burned area was
removed to an approximate average depth of 4 cm
by erosive sheetwash, rilling, and raindrop impact
Ž.
Cannon et al., 1995, 1998 . Rill networks developed
in both the in-place, burned mineral soil and the
aprons of dry ravel material on the sideslopes. Rills
started high on the hillslopes; and, in general, their
frequency, depth, and width increased with distance
down slope to a maximum width of 30–40 cm and
maximum depth of approximately 5 cm. Field obser-
vations also indicated that erosion by sheetwash and
rilling was particularly severe on the steep slopes cut
into the toe of the older landslide deposits, in the
older colluvium and sheetwash deposits, and in the
residuum developed on the Maroon Formation
Ž.
Kirkham et al., 1999 .
Fifty-seven soil-slip scars were mapped on steeper
Ž.
slopes in the area Fig. 1 . The soil slips typically
involved a 0.3- to 1.0-m thick veneer of surficial
2Ž
deposits over areas of 29–828 m Kirkham et al.,
.
1999 . Material mobilized from the soil slips appar-
ently was very fluid in that only a few traces of the
()
S.H. Cannon et al.rGeomorphology 39 2001 171–188 181
deposits were observed on the hillsides downslope
from the soil slip scars. This contrasts with debris
flows in the main channels that deposited a nearly
continuous, up to 2.5-cm thick, veneer of mud on the
channel banks and sidewalls. Most material from the
soil slips apparently traveled over the hillsides and
into adjacent channels, and no significant deposits
from the soil slips were observed remaining in the
channels. The soil exposed in the soil-slip scars was
not burned, indicating that they formed after the fire
and most likely during the September storm.
It is important to note that the mapped soil-slip
scars did not coincide with the mapped debris-flow
initiation locations, nor could the debris-flow paths
Ž.
be traced up channel to the scars Fig. 1 .
4. Relative contribution of materials from hills-
lope soil slips to debris-flow deposits
We calculated the volume of material mobilized
from the soil-slip scars using an estimated average
Ž.
thickness of 0.6 m Table 3 to compare the volume
of material contributed from the soil slips to the
volume of deposits at the canyon mouths estimated
Ž.
by Cannon et al. 1995, 1998 . With this approach,
we assume that the bulk densities of the material
mobilized from the soil-slip scars and the deposits
are similar. This assumption seems reasonable given
the accuracy of the method. The comparison indi-
cates that, with the exception of basin F, only be-
tween 5% and 12% of the deposits could have come
from the soil-slip scars. In basin F, nearly one quar-
ter of the material could have come from discrete
Ž.
soil slips Table 3 . In this basin, however, the great
majority of the soil-slip scars are located outside the
burned area, suggesting that these failures occurred
in response to heavy rainfall, independent of the
effects of the fire.
The fact that the majority of the debris-flow
material deposited at the mouths of the canyons at
Storm King Mountain did not originate as soil slips
on the hillslopes and the observation that the debris-
flow paths did not originate at the soil-slip scars
suggests that mass movements, or infiltration-tri-
Ž
ggered processes as described by Morton, 1989;
.
Booker, 1998; Cannon, 1999 , were a minor compo-
nent in debris-flow initiation on Storm King Moun-
tain. However, observations of extensive erosion by
rainsplash, sheetwash, and rilling on the hillslopes
indicate that a process of progressive sediment en-
Ž
trainment by surface runoff as suggested by Parrett,
.
1987; Meyer and Wells, 1997 was the primary
mechanism of debris-flow initiation on Storm King
Mountain.
At Storm King Mountain, channel erosion was
laterally discontinuous, extremely variable, and the
pre-event channel configuration was not known.
These conditions make any evaluation of the relative
contribution to deposits at channel mouths from
channel incision beyond the scope of this study.
5. Lithologic and topographic controls on debris-
flow initiation
We determined the geologic unit at the point of
origin of each of the 84 mapped debris-flow paths.
Most of the debris-flows originated within pre-exist-
ing drainages underlain by either Maroon Formation
Table 3
Evaluation of contribution of soil-slip scars to fire-related debris-flow deposits
Drainage Area of soil-slip Estimated volume of Volume of deposit at Ratio of scar volume to deposit
233
Ž. Ž. Ž.
basin scars m soil-slip scars m basin mouth m volume at basin mouth
B 1878 1127 20,824 0.05
C and D 4031 2419 39,064 0.06
E 262 157 1368 0.12
F 1486 892 4256 0.21
G 159 95 1064 0.09
Total 8232 4690 67,944 0.07
Average depth of each soil-slip scar assumed to be 0.6 m. Note that drainages C and D were combined because the deposits merged.
()
S.H. Cannon et al.rGeomorphology 39 2001 171–188182
Ž.
or older landslide deposits Table 4 . Considerably
fewer debris flows initiated in younger colluvium
and sheetwash deposits, older colluvium and sheet-
wash deposits, and the Eagle Valley Evaporite. The
number of debris-flow initiation locations within in a
particular geologic unit depends at least in part upon
the areal extent of that unit within the study area. To
compensate for this effect, we divided the percent of
the total number of debris-flow initiation locations in
each geologic unit by the percent exposure of the
host unit within the study area to obtain a dimension-
Ž.
less index of relative susceptibility Table 4 . These
susceptibility indices indicate that the Maroon For-
mation and older landslide deposits were the most
susceptible geologic units to debris-flow initiation.
The remaining geologic units that produced debris
flows were considerably less susceptible. Field ob-
servations of abundant sheetwash and rill erosion on
steep slopes cut into the toe of the older landslide
deposits and the residuum on the Maroon Formation
support the high susceptibility of these units.
Although the hillslope gradient at the location of
failure is a primary control on the initiation of
Ž
rainfall infiltration-triggered debris flows e.g., Ellen
.
et al., 1988; Wieczorek et al., 1988 , the evaluations
presented above indicate the importance of runoff-
dominated rather than infiltration-triggered process
in the generation of the fire-related debris flows from
Storm King Mountain. With this consideration, we
explored the idea that the runoff-contributing area
extending upslope from the initiation location to the
drainage divide, and its gradient, constrained debris-
flow initiation during the September events. We
assume that runoff and material eroded from the
contributing areas were necessary to generate the
debris flows, which then propagated downslope. The
upslope contributing area is similar to the critical
support area defined by Montgomery and Foufoula-
Ž.
Georgiou 1993 and Montgomery and Dietrich
Ž.
1992, 1994 in their work on the runoff-controlled
generation of channels and to the topographic index
used in the hydrologic model TOPMODEL for runoff
Ž.
generation Beven and Kirkby, 1979 .
Contributing areas were delineated on the
1:5000-scale, 20-ft contour DLG generated from sur-
vey-controlled aerial photographs taken after the
Ž.
September events Fig. 8 Lateral boundaries of con-
tributing areas were defined as the pair of flow lines
essentially perpendicular to the contour lines that
converge at the debris-flow initiation location. The
areas and slopes of the contributing areas were mea-
sured using a planimeter and scale. Although at-
tempts were made to use automated commercially
available watershed definition tools, these tools did
not satisfactorily delineate the subtle 0- and 1st-order
channels or hollows occupied by the debris-flow
initiation locations.
Ž.
Montgomery and Foufoula-Georgiou 1993 show
theoretically and empirically that channels are initi-
ated with smaller drainage areas on steeper slopes.
An empirical relation between the upslope contribut-
ing area and its gradient can be defined for the
debris-flow initiation locations on Storm King
Ž.
Mountain Fig. 9 A regression analysis of the area
and gradient data yields the relation
y2.90
As1726 tan
u
,1
Ž. Ž.
where Ais the upslope contributing area in square
meters and tan
u
is its gradient. In this form, 1726
Table 4
Distribution of debris-flow initiation locations within the geologic units and debris-flow susceptibility index for each geologic unit
Geologic unit Number of initiation Percent of initiation Index of relative
Ž.
locations locations % susceptibility
Eagle Valley Evaporite 1 1 0.15
Maroon Formation 55 65 1.60
Younger colluvium and sheetwash 6 7 0.51
Older colluvium and sheetwash 4 5 0.60
Older landslide deposits 18 21 1.20
The index is calculated as the percentage of the total number of debris flows in each geologic unit divided by the percentage area of the unit
Ž.
within the study area see Table 2 .
()
S.H. Cannon et al.rGeomorphology 39 2001 171–188 183
Fig. 8. Example of mapping of upslope contributing areas to debris-flow initiation locations in basin C. Topographic base is from
1:5000-scale DLG generated from survey-controlled aerial photographs. Solid circles mark the debris-flow initiation locations, and solid
black lines are the debris-flow paths. Shaded polygons denote the contributing areas above each initiation location. Dashed line shows the
drainage divides between basins C, B, and H.
m2is the value of the upslope contributing area at
Ž.
tan
u
s1458slope . A correlation coefficient, R,of Ž.
0.58 indicates that Eq. 1 describes the relation
between the independent and dependent variables
()
S.H. Cannon et al.rGeomorphology 39 2001 171–188184
Fig. 9. Graph of upslope contributing area as a function of its average gradient for all 84 debris-flow initiation locations.
Ž.
with some scatter. The F-statistic Fis greater
calc
than F, indicating that the independent vari-
1,69;0.05
able contributes significantly in predicting the depen-
Ž.
dent variable Table 5 . A Pvalue of -0.001
indicates that there is little probability of being wrong
in concluding that there is a true relation between the
independent and dependent variables.
Ž.
Eq. 1 can also be written as
2.90
Atan
u
s1726, 2
Ž. Ž.
or generalized into the form
3
Atan
u
sS,3
Ž. Ž.
cr
where Ais the critical contributing area, and Sis
cr
the value of upslope contributing area at tan
u
s1.
Table 5
Analysis of variance for relation between upslope contributing
area and gradient
df Sum of Mean FF P
calc 1,69;0.05
squares square
Regression 1 10.487 10.487 35.309 4.00 -0.001
Residual 69 20.493 0.297
Total 70 30.980 0.443
The physical significance of the cubed slope term in
unknown; this is simply an empirical result.
On Storm King Mountain, we found that the
geologic unit underlying the contributing area affects
Ž.
its gradient and area characteristics. Fig. 10 shows
the relation between area and slope for contributing
areas underlain entirely by either Maroon Formation
or older landslide deposits, the two geologic units
that hosted the most debris flows. This graph shows
a distinct cluster of contributing areas within the
Ž
Maroon Formation on slopes between 278tan
u
s
.Ž .
0.51 and 428tan
u
s0.90 , while the contributing
areas within the older landslide deposits are larger
than those in the Maroon Formation and form on
Ž.Ž.
slopes between 148tan
u
s0.25 and 378
u
s0.75 .
Some of this distinction may be the result of overall
lower slopes on the older landslide deposits. Note
also that the ambiguous inverse slope–area relation
for the data from the Maroon Formation alone may
indicate that other factors, such as variations in
rainfall intensity and sediment availability, may also
affect initiation locations.
We calculated the lower bounds for debris-flow
initiation on Storm King Mountain using values
measured for each contributing area and its gradient
()
S.H. Cannon et al.rGeomorphology 39 2001 171–188 185
Ž.
Fig. 10. Graph of contributing area as a function of slope for contributing areas underlain entirely by either the Maroon Formation vor
Ž.
the older landslide deposits `. Thin solid line marks the lower initiation threshold for the Maroon Formation; thick line marks the
threshold for the older landslide deposits. Dashed line marks the upper limit of contributing areas that produced fire-related debris flows.
Ž.
in Eq. 3 . With the elimination of the lowest, and
errant, value for each unit, a slope–area threshold for
debris-flow initiation locations in the Maroon Forma-
tion is
3
Atan
u
s200 4
Ž. Ž.
cr
for values of
u
between 278and 428. The threshold
for the older landslide deposits is
3
Atan
u
s300 5
Ž. Ž.
cr
Ž.
for values of
u
between 148and 378Fig. 10 .
Although the initiation threshold lines are similar
for the two units, debris flows were produced from
the Maroon Formation from contributing areas that
were generally smaller and with steeper gradients
than those that produced debris flows from the older
landslide deposits.
Ž. Ž.
Eqs. 4 and 5 define slope-dependent thresholds
for fire-related debris-flow initiation for different
materials. The equations are similar in form to
thresholds for critical support area defined by Mont-
Ž.
gomery and Dietrich 1994 for channel initiation by
overland flow, the exception being that in their paper
tan
u
is squared rather than cubed. Note also that an
upper limit of slope–area characteristics exists for
Ž
the contributing areas on Storm King Mountain Fig.
.
10 , indicating that specific combinations of large,
steep contributing areas will not produce runoff-
dominated debris flows.
6. Conclusions and discussion
From field observations and measurements, and
the evaluations above, we suggest that the generation
of debris flows at Storm King Mountain started with
significant sheetwash, rill, and rainsplash erosion and
transport of burned mineral soil and dry-ravel mate-
rials from the hillslopes high within the contributing
areas. Surface runoff bulked with material eroded
from the hillslopes converged into small, 0- and
1st-order hollows and channels that were mantled
with dry-ravel material. The flowing water easily
incorporated this material. At the point within the
drainages defined by a threshold value of upslope
()
S.H. Cannon et al.rGeomorphology 39 2001 171–188186
contributing area and its gradient, sufficient eroded
material had been incorporated, relative to volume of
contributing surface runoff, to generate debris flows.
The dependence of the threshold on gradient sug-
Ž.
gests three possibilities: i The amount of post-fire
sediment mobilized depended on the erosive ability
of runoff at each point within the contributing area,
which in turn depended on shear stress and hence on
Ž.
flow depth and gradient. ii Within the contributing
areas, the down-gradient increase in available sedi-
ment was greater than the down-gradient increase in
surface runoff, which resulted in a progressive in-
Ž.
crease in the sedimentrwater ratio. iii A combina-
tion of both. The location of the threshold thus
reflects variations in sediment supply related to gra-
dient, in that steeper slopes provide more erodible
material per unit area. The slope–area threshold also
varies with geologic materials.
As the flows traveled through higher-order chan-
nels, discharges increased as runoff and additional
eroded material was contributed from the sideslopes
and tributary channels and from soil slips on the
hillslopes. Larger material stored within the channel
was incorporated into the flows and subsequently
flushed out of the canyon mouths as cobble- and
boulder-sized material in an abundant fine-grained
matrix. The material deposited at the canyons mouths
was more fluid and contained relatively less large
material than deposits within the canyons, indicating
that the downchannel contribution of sediment might
have decreased somewhat relative to the downchan-
nel increase in discharge. However, the sedimentr
water ratio was sufficient to maintain debris-flow
conditions.
It is also plausible that the debris-flow material
exiting the channel mouths was of lower strength
and contained less large material due to segregation
of the flows within the channel into less mobile
bouldery flow fronts, with more mobile and gravel-
poor surges and midsections and dilute tails, as
Ž.
documented by Pierson 1986 . Other debris-flow
studies document how boulders and cobbles are typi-
cally deposited at higher gradients and flow depths
Ž.
e.g., Sharp and Nobles, 1953 , while the debris
flows minus these coarser materials continue down-
Ž.
channel. Meyer and Wells 1997 also document this
depositional pattern, and based on a sediment budget
infer that the large volume of gravel-poor debris-flow
deposits observed in fire-related events resulted from
a significant contribution of fine sediment from the
hillslopes to debris flows.
Although erosion following wildfires is frequently
attributed to the development of a water-repellent
soil, water-repellent soils were not extensive at Storm
King Mountain. Thus, the process of debris-flow
initiation described above does not depend on the
presence of such soils. This is consistent with con-
clusions reached in southern California by Cannon
Ž. Ž.
1999 , and Meyer and Wells 1997 who found that
the presence of water-repellent soils were not a
prerequisite for debris-flow occurrence.
In contrast to the mechanism for debris-flow initi-
Ž.
ation proposed by Meyer and Wells 1997 , the
majority of the debris flows from Storm King Moun-
tain initiated in 0- and 1st-order channels and hol-
lows, in contrast to Yellowstone where Meyer and
Ž.
Wells 1997 described the first recognition of de-
bris-flow features within higher-order channels. The
origination of debris flows on Storm King Mountain
in 0- and 1st-order channels and hollows also differs
Ž.
from the mechanism proposed by Wells 1987 and
Ž.
Johnson 1984 , where debris flows are thought to
initiate as rills high on burned hillslopes.
Ž.
In addition, while Meyer and Wells 1997 con-
cluded that a significant volumetric contribution to
the flow by erosion of channel material was impor-
tant to the initiation process, observations at Storm
King Mountain indicate that extensive erosion of the
channels did not occur, and thus, contribution of
material from the channels was of secondary impor-
tance. Interestingly, the total volume of sediment
produced per unit basin area for basins B, C, and D
at Storm King Mountain is 1.5–2 times greater than
for a debris-flow producing basin in Yellowstone of
Ž
comparable size and near-total burn A12 kmBbasin,
.
Meyer and Wells, 1997 . This would suggest perhaps
significant differences in degree of weathering and
thus considerably greater erodibility of the Storm
King Mountain hillslopes relative to the thin mantle
of colluvium and soil in glaciated Yellowstone. An-
other possibility for this contrast might be that large
intense fires may have been more common in the
Ž.
late Holocene at Yellowstone Meyer et al., 1995
than at Storm King Mountain, leaving smaller vol-
umes of erodible materials available for entrainment
into individual debris-flow events.
()
S.H. Cannon et al.rGeomorphology 39 2001 171–188 187
Acknowledgements
We greatly appreciate the efforts of W. Pat Rogers
of the Colorado Geological Survey for initiating and
coordinating this work and for reviewing a previous
version of the paper. The U.S. Geological Survey
Landslide Hazards Program, the Colorado Geologi-
cal Survey, the U.S. Bureau of Land Management,
and the Colorado Department of Transportation—
Region 3, provided funding for this cooperative pro-
gram. John Michael provided invaluable expertise in
the GIS Lab, and Steven Reneau, Alan Chleborad,
Grant Meyer and an anonymous reviewer con-
tributed thorough and constructive criticisms of this
paper.
References
Beven, K.J., Kirkby, M.J., 1979. A physically based, variable
contributing area model of basin hydrology. Hydrol. Sci. Bull.
24, 43–69.
Booker, F.A. 1998. Landscape and management response to wild-
fires in California. MS Thesis, Univ. of California, Berkeley,
436 pp.
Bryant, B., Shroba, R.S., Harding, A.E., 1998. Revised prelimi-
nary geologic map of the Storm King Mountain quadrangle,
Garfield County, Colorado. U. S. Geol. Surv. Open-File Rep.,
98-472.
Cannon, S.H. 1999. Debris-flow response of watersheds recently
burned by wildfire. PhD Thesis, Univ. of Colorado, Boulder,
200 pp.
Cannon, S.H., Reneau, S.L., 2000. Conditions for generation of
fire-related debris flows, Capulin Canyon, New Mexico. Earth
Surf. Processes Landforms 25, 1103–1121.
Cannon, S.H., Powers, P.S., Pihl, R.A., Rogers, W.P., 1995.
Preliminary evaluation of the fire-related debris flows on
Storm King Mountain, Glenwood Springs, Colorado. U. S.
Geol. Surv. Open-File Rep., 95-508, 38 pp.
Cannon, S.H., Powers, P.S., Savage, W.Z., 1998. Fire-related
debris flows on Storm King Mountain, Glenwood Springs,
Colorado, USA. Environ. Geol. 35, 210–218.
Campbell, R.H., 1975. Soil slips, debris flows, and rainstorms in
the Santa Monica Mountains and vicinity, southern California.
U. S. Geol. Surv. Prof. Pap. 851, 51 pp.
Craig, R.F., 1987. Soil Mechanics. 4th edn. Van Nostrand-Rein-
hold, UK.
DeBano, L.F., 1981. Water repellant soil: state-of-the-art. USDA
For. Serv., Gen. Tech. Rep. PSW 46, 21 pp.
DeBano, L.F., Letey, J., 1969. Water-repellent soils. Univ. of
California, Riverside, May 6–10, 1968, Proceedings. 354 pp.
DeGraff, J.V., 1997. Geologic investigation of the Pilot Ridge
debris flow, Groveland Range District, Stanislaus National
Forest. USDA For. Serv. 20 pp.
Ellen, S.D., Cannon, S.H., Reneau, S.L., 1988. Distribution of
debris flows in Marin County. In: Ellen, S.D., Wieczorek,
Ž.
G.F. Eds. , Landslides, Floods, and Marine Effects of the
Storm of January 3–5, 1982, in the San Francisco Bay Region,
California. U. S. Geol. Surv. Prof. Pap. 1434, pp. 113–131.
Florsheim, J.L., Keller, E.A., Best, D.W., 1991. Fluvial sediment
transport following chaparral wildfires, Ventura County,
southern California. Geol. Soc. Am. Bull. 103, 504–511.
Helvey, J.D., 1980. Effects of a north central Washington wildfire
on runoff and sediment production. Water Res. Bull. 16,
627–634.
Iverson, R.M., Reid, M.E., LaHusen, R.G., 1997. Debris-flow
mobilization from landslides. Annu. Rev. Earth Planet. Sci.
25, 85–138.
Johnson, A.M., 1984. Debris flow. In: Brundsen, D., Prior, D.B.
Ž.
Eds. , Slope Instability. Wiley, NY, pp. 257–361with contri-
butions from J.R. Rodine.
Kirkham, R.M., Parise, M., Cannon, S.H., 1999. Geology of the
1994 South Canyon fire area, and a geomorphic analysis of the
September 1, 1994 debris flows, south flank of Storm King
Mountain, Glenwood Springs, Colorado. Colo. Geol. Surv.
Spec. Publ. 46, 35 pp.
Klock, G.O., Helvey, J.D., 1976. Debris flows following wildfires
in north central Washington. Proc. 3rd Federal Inter-Agency
Sedimentation Conference, Denver, CO. pp. 91–96.
Megahan, W.F., 1983. Hydrologic effects of clearcutting and
wildfire on steep granitic slopes in Idaho. Water Resour. Res.
19, 811–819.
Meyer, G.A., Wells, S.G., 1997. Fire-related sedimentation events
on alluvial fans, Yellowstone National Park, U.S.A. J. Sedi-
ment. Res. 67, 776–791.
Meyer, G.A., Wells, S.G., Jull, A.J.T., 1995. Fire and alluvial
chronology in Yellowstone National Park: climatic and intrin-
sic controls on Holocene geomorphic processes. Geol. Soc.
Am. Bull. 107, 1211–1230.
Montgomery, D.R., Dietrich, W.E., 1992. Channel initiation and
the problem of landscape scale. Science 225, 826–830.
Montgomery, D.R., Dietrich, W.E., 1994. Landscape dissection
Ž.
and drainage area–slope thresholds. In: Kirkby, M.J. Ed. ,
Process Models and Theoretical Geomorphology. Wiley, NY,
pp. 221–246.
Montgomery, D.R., Foufoula-Georgiou, E., 1993. Channel net-
work source representation using digital elevation models.
Water Resour. Res. 29, 3925–3934.
Morris, S.E., Moses, T.A., 1987. Forest fire and the natural soil
erosion regime in the Colorado Front Range. Ann. Assoc. Am.
Geogr. 77, 245–254.
Morton, D.M., 1989. Distribution and frequency of storm-gener-
ated soil slips on burned and unburned slopes, San Timoteo
Badlands, Southern California. In: Sadler, P.M., Morton, D.M.
Ž.
Eds. , Landslides in a Semi-Arid Environment with Emphasis
on the Inland Valleys of Southern California. Pubs. Inland
Geol. Soc. 2, pp. 279–284.
Parrett, C., 1987. Fire-related debris flows in the Beaver Creek
drainage, Lewis and Clark County, Montana. U. S. Geol. Surv.
Water-Supply Pap. 2330, 57–67.
Pierson, T.S., 1986. Flow behavior of channelized debris flows,
()
S.H. Cannon et al.rGeomorphology 39 2001 171–188188
Ž.
Mount St. Helens, Washington. In: Abrahams, A.D. Ed. ,
Hillslope Processes. Allen and Unwin, Boston, pp. 269–296.
Robichaud, P.R. 1996. Spatially-varied erosion potential from
harvested hillslopes after prescribed fire in the interior north-
west. PhD thesis, Univ. of Idaho, Moscow, 219 pp.
Sharp, R.P., Nobles, L.H., 1953. Mudflow of 1941 at Wright-
wood, southern California. Geol. Soc. Am. Bull. 64, 547–560.
Swanson, F.J., 1981. Fire and geomorphic processes. In: Mooney,
H.A., Bonnicksen, T.M., Christensen, N.L., Lotan, J.E., Rein-
Ž.
ers, W.A. Eds. , Fire Regimes and Ecosystem Properties.
USDA For. Serv. Gen. Tech. Rep. WO, vol. 26, pp. 401–420.
Wells, W.G., 1987. The effects of fire on the generation of debris
flows in southern California. In: Costa, J.E., Wieczorek, G.F.
Ž.
Eds. , Geol. Soc. Am. Rev. Eng. Geol. 7, pp. 105–114.
Wieczorek, G.F., Harp, E.L., Mark, R.K., Bhattacharyya, A.K.,
1988. Debris flows and other landslides in San Mateo, Santa
Cruz, Contra Costa, Alameda, Napa, Solano, Sonoma, Lake,
and Yolo Counties, and factors influencing debris-flow distri-
Ž.
bution. In: Ellen, S.D., Wieczorek, G.F. Eds. , Landslides,
Floods, and Marine Effects of the Storm of January 3–5,
1982, in the San Francisco Bay Region, California. U. S. Geol.
Surv. Prof. Pap. 1434, pp. 133–161.
